Analytics September/October 2015 by Lionheart Publishing Inc.

h t t p : / / w w w. a n a ly t i c s - m a g a z i n e . O R G

Driving Better Business Decisions

septem ber/ o c tober 2015 Brought to you by:

ALSO INSIDE: • The magic of managed autonomy • Build global business collaboration • Predictive analytics & publishing

Why Customer knowledge is king •C ognitive computing for automating customer knowledge • Managing customer experience, delivering real-time results

Executive Edge Arvind Purushothaman, senior director of analytics at Virtusa, on benefits of updating your data architecture

Ins ide story

When TMI is never enough TMI (too much information) may be a popular dismissive remark for online texting chatterboxes, but there’s no such thing as TMI for marketing mavens when it comes to current and potential customers. Marketers, it seems, can never get enough information, and in today’s big data world where every move you make, every bond you break, every step you take (apologies to Sting) is seemingly captured electronically, marketers find themselves swimming in info. The problem, of course, is converting all of the information into something useful, such as deep customer knowledge, so marketers can predict what a particular customer will want to buy next, and offer them a deal on that very product right now, whether the customer is online or in the store. Behind all of the marketing magic, particularly at big organizations, you’ll typically find a team of data scientists, mathematicians, statisticians, operations researchers, computer scientists and assorted high-end analysts, most of them with advanced degrees in their respective fields. Given the enormous market demand for these quants and the skills they bring to the table, how do you build an effective, productive, collaborative analytics team and how do you retain the team members and keep them engaged when they all have good reason to believe that they are the smartest person in the room? 2

a n a ly t i c s - m a g a z i n e . o r g

These two related problems – creating deep customer knowledge and building and retaining analytics teams that make company mission-critical goals such as deep customer knowledge possible – are the focus of several articles in this September/October issue of Analytics magazine. To the first point, Guy Mounier, co-founder and CEO at CustomerMatrix, extols the virtues of “Cognitive computing for automating customer knowledge.” Swaroop Johnson, a consultant at Blueocean Market Intelligence, follows by walking readers through the process of understanding the role of customer intelligence, managing the customer experience and delivering real-time results in his article on “Customer intelligence.” Turning to the issue of analytics teambuilding, senior data scientist Vinod Cheriyan outlines how online lender Enova International built a successful analytics team by balancing talent engagement with business priorities in “The magic of managed autonomy.” Next, Rasesh Shah, CIO of Fractal Analytics, and Aliasgar Rajkotwala, Fractal’s global head of IT, discuss “Building global business collaboration” based on security, freedom and mobility. Too much information? LOL.

– Peter Horner, editor peter.horner@ mail.informs.org w w w. i n f o r m s . o r g

C o n t e n t s

DRIVING BETTER BUSINESS DECISIONS

September/october 2015 Brought to you by

Features 34 Cognitive computing, customer knowledge How to automatically provide a clear means to make the most of your company’s proprietary data By Guy Mounier

38 The keys to customer intelligence Replacing data silos with “datamarts” and the impetus behind real-time business intelligence By Swaroop Johnson 42 The magic of managed autonomy Online lender builds successful analytics team by balancing talent engagement, business priorities By Vinod Cheriyan 48 Global business collaboration The ideal infrastructure relies on the three best practice pillars: security, freedom and mobility By Rasesh Shah and Aliasgar Rajkotwala 54

Predictive analytics in publishing Transformation of traditional print media and the growing intelligence behind monetization strategies By Arvid Tchivzhel

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

S d A upp pa or ch ts P eS o pa we rk r B Analytics in Excel Bi I gD at a

ANALYTIC SOLVER PLATFORM ®

Easy to Use, Industrial Strength Predictive

How can you get results quickly for business decisions, without a huge budget for “enterprise analytics” software, and months of learning time? Here’s how: Analytic Solver Platform does it all in Microsoft Excel, accessing data from PowerPivot and SQL databases.

Data Mining, Text Mining, Predictive Analytics – with Surprising Performance on Large Datasets. Point and click to analyze a mix of free-form text and structured data, with text processing, latent semantic analysis, Zipf’s, Scree and Term-Concept scatter plots. Use feature selection and principal components to identify important variables, and k-means clustering or hierarchical clustering to simplify and cluster your data. Use exponential smoothing and ARIMA models for forecasting; multiple regression, k-nearest neighbors, and ensembles of regression trees and neural networks for prediction; discriminant analysis, logistic regression, naïve Bayes, k-nearest neighbors, and ensembles of classification trees and neural nets for classification; and association rules for affinity analysis.

Simulation, Optimization, Prescriptive Analytics. Analytic Solver Platform also includes decision trees, Monte Carlo simulation, and powerful conventional and stochastic optimization for prescriptive analytics.

Help and Support to Get You Started. Analytic Solver Platform can help you learn while getting results in business analytics, with its Guided Mode and Constraint Wizard for optimization, and Distribution Wizard for simulation. You’ll benefit from User Guides, Help, 30 datasets, 90 sample models, and new textbooks supporting Analytic Solver Platform.

Find Out More, Download Your Free Trial Now. Visit www.solver.com to learn more, register and download a free trial – or email or call us today.

The Leader in Analytics for Spreadsheets and the Web Tel 775 831 0300 • info@solver.com • www.solver.com

DRIVING BETTER BUSINESS DECISIONS

Departments 2 Inside Story 8 Executive Edge 12 Analyze This! 18 Healthcare Analytics 22 INFORMS Initiatives 24 Viewpoint 30 Forum 60 Corporate Profile 68 Conference Preview 74 Five-Minute Analyst 78 Thinking Analytically

Analytics (ISSN 1938-1697) is published six times a year by the Institute for Operations Research and the Management Sciences (INFORMS), the largest membership society in the world dedicated to the analytics profession. For a free subscription, register at http://analytics.informs.org. Address other correspondence to the editor, Peter Horner, peter.horner@mail.informs.org. The opinions expressed in Analytics are those of the authors, and do not necessarily reflect the opinions of INFORMS, its officers, Lionheart Publishing Inc. or the editorial staff of Analytics. Analytics copyright ©2015 by the Institute for Operations Research and the Management Sciences. All rights reserved.

INFORMS Board of Directors President L. Robin Keller, University of California, Irvine President-Elect Edward H. Kaplan, Yale University Past President Stephen M. Robinson, University of Wisconsin-Madison Secretary Brian Denton, University of Michigan Treasurer Sheldon N. Jacobson, University of Illinois Vice President-Meetings Ronald G. Askin, Arizona State University Vice President-Publications Jonathan F. Bard, University of Texas at Austin Vice President Sections and Societies Esma Gel, Arizona State University Vice President Information Technology Marco Lübbecke, RWTH Aachen University Vice President-Practice Activities Jonathan Owen, CAP, General Motors Vice President-International Activities Grace Lin, Institute for Information Industry Vice President-Membership and Professional Recognition Ozlem Ergun, Georgia Tech Vice President-Education Jill Hardin Wilson, Northwestern University Vice President-Marketing, Communications and Outreach E. Andrew “Andy” Boyd, University of Houston Vice President-Chapters/Fora David Hunt, Oliver Wyman

INFORMS Offices www.informs.org • Tel: 1-800-4INFORMS Executive Director Melissa Moore Meetings Director Laura Payne Director, Public Relations & Marketing Jeffery M. Cohen Headquarters INFORMS (Maryland) 5521 Research Park Drive, Suite 200 Catonsville, MD 21228 Tel.: 443.757.3500 E-mail: informs@informs.org Analytics Editorial and Advertising Lionheart Publishing Inc., 506 Roswell Street, Suite 220, Marietta, GA 30060 USA Tel.: 770.431.0867 • Fax: 770.432.6969 President & Advertising Sales John Llewellyn john.llewellyn@mail.informs.org Tel.: 770.431.0867, ext. 209 Editor Peter R. Horner peter.horner@mail.informs.org Tel.: 770.587.3172 Assistant Editor Donna Brooks donna.brooks@mail.informs.org Art Director Jim McDonald jim.mcdonald@mail.informs.org Tel.: 770.431.0867, ext. 223 Advertising Sales Sharon Baker sharon.baker@mail.informs.org Tel.: 813.852.9942

S d A upp pa or ch ts P eS o pa we rk r B in Excel Bi I gD at a

ANALYTIC SOLVER PLATFORM ®

From Solver to Full-Power Business Analytics

Solve Models in Desktop Excel or Excel Online.

Plus Forecasting, Data Mining, Text Mining.

From the developers of the Excel Solver, Analytic Solver Platform makes the world’s best optimization software accessible in Excel. Solve your existing models faster, scale up to large size, and solve new kinds of problems. Easily publish models from Excel to share on the Web.

Analytic Solver Platform samples data from Excel, PowerPivot, and SQL databases for forecasting, data mining and text mining, from time series methods to classification and regression trees and neural networks. And you can use visual data exploration, cluster analysis and mining on your Monte Carlo simulation results.

Conventional and Stochastic Optimization. Fast linear, quadratic and mixed-integer programming is just the starting point in Analytic Solver Platform. Conic, nonlinear, non-smooth and global optimization are just the next step. Easily incorporate uncertainty and solve with simulation optimization, stochastic programming, and robust optimization – all at your fingertips.

Find Out More, Download Your Free Trial Now. Analytic Solver Platform comes with Wizards, Help, User Guides, 90 examples, and unique Active Support that brings live assistance to you right inside Microsoft Excel. Visit www.solver.com to learn more, register and download a free trial – or email or call us today.

Fast Monte Carlo Simulation and Decision Trees. Analytic Solver Platform is also a full-power tool for Monte Carlo simulation and decision analysis, with 50 distributions, 40 statistics, Six Sigma metrics and risk measures, and a wide array of charts and graphs.

The Leader in Analytics for Spreadsheets and the Web Tel 775 831 0300 • info@solver.com • www.solver.com

exe cu tive e d g e

It pays to modernize your data architecture The key lies in building a modern data architecture that is open, flexible and scalable, something that can accommodate your existing data assets as well as potential new ones.

By Arvind Purushothaman 8

In today’s world where data is collected at every interaction, be it over the phone, mobile, PC, sensors, with or without us knowing, it becomes important to have a strategy around data. Traditionally, data has been seen as something to “run the business,” but, in today’s context, it can actually “be the business” if monetized well. An example of Internet of Things (IoT) data in a customer context is the wristband one wears at amusement parks that provides real-time data about customer interaction at all times, and this data can be processed in near real time to push out relevant offers and alerts to enhance the customer experience. The question is: How do organizations prepare themselves to take advantage of data? The key lies in building a modern data architecture that is open, flexible and scalable, something that can accommodate your existing data assets as well as potential new ones. Before we talk about specific steps to modernize data architecture, let’s look at typical challenges: 1. Many applications within the organization have been around for 20 or more years. While the usage for some of them is known, it is still not clear who is

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

ANALYTICS IN YOUR BROWSER

Optimization, Simulation, Statistics in Excel Online, Google Sheets – or Your Own Web or Mobile Application with RASON™ Software

Advanced Analytics built on Business Intelligence is moving to the Web, and getting easier! Spreadsheet users can get results quickly in Excel Online or Google Sheets. Analysts and developers can build Web and mobile apps quickly using our new RASON™ language.

Use our Analytics Apps in your Web Browser Solve linear, integer and nonlinear optimization models with Frontline’s free Solver App, for desktop Excel Solver features in Excel Online and Google Sheets. Run Monte Carlo simulation models with Frontline’s Risk Solver App – free for limited-size models, and free without size limits for desktop Risk Solver Pro, Risk Solver Platform, and Analytic Solver Platform users. Use our free XLMiner Analysis ToolPak App for statistical analysis in Excel Online and Google Sheets, matching the Analysis ToolPak in desktop Excel – and our free XLMiner Data Visualization App to explore your data with multivariate charts in Excel Online.

With support for linear, nonlinear and stochastic optimization, array and vector-matrix operations, and dimensional tables linked to external databases, the RASON™ language gives you all the power you need.

Application Deployment Made Easy The magic begins when you deploy your model to the Web. Unlike other languages, RASON™ models are valid in JavaScript. Your browser or mobile apps can access our cloud servers via a simple REST API, to easily solve the largest models on the smallest devices. And your server-based applications can easily solve RASON™ models with our Solver SDK object-oriented library.

Find Out More, Sign Up for a Free Trial Now. Visit www.solver.com/apps to learn more, and visit rason.com to sign up for a free trial of RASON™ and our REST API. Or email or call us today.

Build Your Own Apps with RASON™ Software RASON™ – RESTful Analytic Solver™ Object Notation – is a new modeling language for optimization and simulation that’s embedded in JSON (JavaScript Object Notation).

The Leader in Analytics for Spreadsheets and the Web Tel 775 831 0300 • info@solver.com • www.solver.com

exe cu tive e d g e

CIOs are concerned about having to find an army of programmers for populating Hadoop-based data repositories.

leveraging the data in each application and for what purpose. How do we find out? 2. To meet their reporting needs, organizations have built multiple data assets including data warehouses and data marts. Additionally, they have power users collating data from multiple sources and creating reports using MS Excel. Numbers are inconsistent and vary based on who is preparing them and the intended purpose. 3. Organizations have multiple applications and data assets starting with mainframe-based ones, client-server, Web applications and some newer cloud-based applications, all co-existing. They struggle to find the right people to support the applications, especially the older ones. 4. Organizations are aware of the new developments in the big data space including NoSQL databases and the Hadoop ecosystem, and have typically embarked on some initiatives to get started on this. The main challenge is around integrating this with the traditional data warehouse technologies. 5. People, and by extension, their skills, are the biggest assets of any organization. CIOs are concerned about having to find an army of programmers for populating Hadoop-based data repositories. The other big concern is how to leverage existing SQL skills, which people have acquired over the years. These are valid concerns, and some are more applicable than others based on the context. Nonetheless, given the inevitable need to be able to better monetize data and modernize

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

technology platforms, it is important to have a strategy. I recommend the following approach: 1. Data asset inventory: Create a complete list of data assets – legacy, data warehouses, data marts, data islands. Identify the data flows between these assets and the usage patterns. It might be particularly hard for some legacy systems, but this serves as the starting point for any consolidation and modernization. 2. Data asset rationalization: Based on the list of data assets and the usage, it is important to rationalize them. What this means is to identify if the same data is coming from multiple applications, and if so, which is the authoritative source, and which ones can be retired. This is a very important exercise and can help consolidate the number of data assets to a manageable few. In this context, master data management is critical to ensure you have good quality data. 3. Data lineage: Undertaking a data lineage exercise to identify data flows – creating detailed documentation especially for the legacy applications – is a must. This greatly reduces the risk of dependency on key personnel and also makes it easier to migrate to a future state architecture. 4. Data infrastructure: Have a big data and cloud strategy in place to bring a na l y t i c s

in newer technologies in a pilot mode. Start with a non-legacy application to understand the technology, and move applications over in conjunction with data asset rationalization. The “data on cloud” is going to be an important component of modern architecture especially when dealing with IoT data. 5. Data technology: It pays to understand the different options available in a very crowded and rapidly evolving marketplace, and to select the right technologies that fit into your architecture from a technology standpoint as well as a people standpoint. For example, using a data integration tool with big data connectors will eliminate the need for people who can write MapReduce code. Creating a holistic data strategy in light of changes in the business, and taking a structured approach, will definitely help lay a solid foundation that will be the basis for monetizing data. ❙ Arvind Purushothaman is the practice head and senior director of Information Management & Analytics at Virtusa’s Chennai ATC of Virtusa, an information technology services provider with global reach. He has 19 years of industry experience, with a focus on planning and executing data management and analytics initiatives. Prior to taking on this role, he was involved in architecting and designing centers of excellence, as well as service-delivery functions focused on information management encompassing traditional data warehousing, master data management and analytical reporting. He holds an MBA from Georgia State University.

s e p t e m b e r / o c t o b e r 2 015

ANALYZE THIS

Analytics-based program has the ‘write’ stuff Too many newly minted graduates have too little capacity to express themselves clearly in writing, with negative consequences for both their current employment prospects and their future employers.

By Vijay Mehrotra

As a business school faculty member, I have been hearing for years that the ability to write well still matters mightily [1], even in our increasingly data-driven business world. Most recently, the 2014 Graduate Management Admissions Council’s annual survey of more than 550 companies from 44 different countries around the world found that employers valued communication skills more highly than anything else [2]. And while many courses include opportunities to make oral presentations and receive feedback, MBA students typically receive little training on how to write effectively [3]. Of course, effective writing is a skill that can and should be taught long before graduate school. And so much of learning to write is about coaching, feedback and revision. But in today’s world of shrinking educational budgets, growing class sizes, and an increasingly large population of poorly paid and overworked writing instructors, too many students are too often provided with too little of this type of personal attention. The results are all too predictable: too many newly minted graduates with too little capacity to express themselves clearly in writing, with negative consequences for both their current employment prospects and their future employers.

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

THE

NATION’S FIRST

Associate in Applied Science (A.A.S.) degree in Business Analytics on campus or online.

Credential options • Enroll in one or several: • AAS degree

Why Study Business Analytics?

The Business Analytics curriculum is designed to provide students with the knowledge and the skills necessary for employment and growth in analytical professions. Business Analysts process and analyze essential information about business operations and also assimilate data for forecasting purposes. Students will complete course work in business analytics, including general theory, best practices, data mining, data warehousing, predictive modeling, project operations management, statistical analysis, and software packages. Related skills include business communication, critical thinking and decision making.The curriculum is hands-on, with an emphasis on application of theoretical and practical concepts. Students will engage with the latest tools and technology utilized in today’s analytics fields.

Accelerated Executive Program

Our accelerated learning options allow students to complete certificate credentials in two semesters part time or one semester full time. Accelerated options are available for the Business Intelligence and the Business Analyst certificates.

Questions? Tanya Scott

Director, Business Analytics

919-866-7106 tescott1@waketech.edu

• Certificates: Business Intelligence, Business Analyst, Finance Analytics, Marketing Analytics, and Logistics Analytics

Flexibility • Open-door enrollment • Courses are offered in the fall and spring • Courses can be taken online or on campus • Competitively priced tuition

Gain skills in: • Data gathering • Collating • Cleaning • Statistical Modeling • Visualization • Analysis • Reporting • Decision making

Use data and analysis tools: • Advanced Excel • Tableau • Analytics Programming • SAS Enterprise Guide • SAS Enterprise Miner • SPSS Modeler • MicroStrategy

• Visualization • Presentation

Funded in full by a $2.9 million Dept. of Labor Trade Adjustment Assistance Community College & Career Training (DOLTAACCCT) grant.

businessanalytics.waketech.edu

ANALYZE THIS

Question: H ow do we address this vexing problem? Answer: Analytics! This is a story about a small company with a big mission. WriteLab, a two-year-old start-up that emerged from U.C.-Berkeley, has developed a software platform that aims to radically improve the way that students learn and practice the craft of effective written communication. Founded by Matthew Ramirez and Donald McQuade, WriteLab’s software application ingests written documents and quickly annotates them with notes and suggestions intended to improve the quality of the user’s writing. The story begins with Ramirez, a Ph.D. student in English at U.C.-Berkeley. Early in his graduate school career, he found himself teaching a writing class for the first time, spending a huge amount of time grading student papers. Moreover, he soon discovered that the feedback loop was too slow; by the time that students received the papers that he had so carefully annotated and graded, their attentions were focused on other things. Soon thereafter, his eclectic interests (which include both a passion for the structure of language and a keen curiosity about mathematics and computer science) led him to courses in natural language processing and machine learning 14

a n a ly t i c s - m a g a z i n e . o r g

at Berkeley’s School of Information. After learning about syntactic parsers, he began to explore how such software might be used to improve the lives of writing students and writing teachers. Around this time, he met McQuade, a professor of English with nearly 30 years of teaching writing to college students. McQuade’s interest was piqued immediately. In December 2013, Ramirez and McQuade launched WriteLab with a vision of using technology to cost-effectively help users to write effectively. Since then, the company has developed a SaaS application that coaches students on their writing and revising, and more than 500 students around the country participated in its spring 2015 beta test. WriteLab plans to rollout its product during the current academic year while continuing to add features and improve functionality. Like many cloud-based analytics applications, WriteLab sits on top of a complex tech stack. Documents are read in and parsed into sentence-level records that are stored in a Dynamo DB hosted on Amazon Web Services, after which more than 20,000 tags are attached to each sentence (this is clearly a memory-intensive application). From here, a series of proprietary predictive algorithms (written in Python and based on WriteLab’s Style Guide) are used to identify potentially valuable pieces of feedback. The writer/ w w w. i n f o r m s . o r g

user then receives this feedback within an annotated version of the original document that includes color-coded labels and comments in seven different areas (clarity, cohesion, logic, concision, emphasis, elegance and coherence), and proceeds by either incorporating or ignoring each piece of feedback in subsequent drafts. In a sense, the WriteLab platform has tapped into McQuade’s deep knowledge and extensive teaching experience while keeping in mind what the company’s data scientists and software engineers believe

can be successfully implemented. In addition, its machine-learning algorithms were initially trained on data sets that are handclassified by expert writing teachers, and continue to be refined as the amount of data processed by the system increases. Les Perelman is a believer in the WriteLab system. Perelman, a research associate and former director of undergraduate writing at MIT, is a longtime critic of automated essay scoring systems. The Boston Globe has labeled him “The Man Who Killed the SAT Essay” [4].

71%

of data management professionals admitted they “have yet to begin planning” their big data strategies. SOURCE: “Big Data in Big Companies,” International Institute for Analytics, May 2013

75% STATISTICS + 25% BUSINESS = 100% COMPETITIVE analytics.stat.tamu.edu

1415_Analytics_Stat.indd 1

a na l y t i c s

2/12/2015 5:03:36 PM

s e p t e m b e r / o c t o b e r 2 015

ANALYZE THIS

It’s a noble and honorable undertaking, and with tens of millions of students as prospective customers, it’s also a potentially profitable one.

He first heard of WriteLab prior to the annual meeting of the National Council of Teachers of English in 2014, and claims that he arrived as a skeptic. “I was planning to trash it,” he later admitted to me, “until I realized that it had immense potential.” Perelman, who subsequently joined WriteLab’s Advisory Board, cited several reasons for the shift in his perspective. First of all, he pointed out that WriteLab’s focus is on writing style at the sentence level, which greatly simplifies the analytical challenge by providing a smaller set upon which to do statistical inference. Also, WriteLab has been built from the ground up as a coaching/teaching tool, unlike many other similar systems that were initially designed to automate the grading of essays (which he vehemently believes they also do quite poorly [5]). Finally, to Perelman, WriteLab’s founders’ first-hand experience as writing teachers is very visible in the product, particularly in the way it asks good questions and gently encourages writers to consider different choices. In these heady days of analytics’ ascendancy, there are countless start-up companies out there promising to use big data to somehow change the world. At their core, however, the vast majority of these applications are about increasing efficiency, pumping up profits and/ or strengthening managerial control. Having spent my entire adult life in this profession, there are days that I wearily wonder if this is really what it’s all about. Writing is about something else entirely. Two of my favorite contemporary authors, Jonathan Franzen and the late David Foster Wallace, have characterized their writing as “a way out of loneliness” [6], and indeed this same characterization is at the heart of why I write this column. Given all of this, it is gratifying to find an

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

analytics-based platform like WriteLab, whose core mission is helping people more effectively connect with others through the written word. It’s a noble and honorable undertaking, and with tens of millions of students as prospective customers, it’s also a potentially profitable one. Vijay Mehrotra (vmehrotra@usfca.edu) is a professor in the Department of Business Analytics and Information Systems at the University of San Francisco’s School of Management and a longtime member of INFORMS. Full disclosure: This column has been significantly improved through the use of the WriteLab software.

REFERENCES 1. For example, see http://www.wsj.com/ articles/SB100014240527487034099045761746 51780110970 and http://www.washingtonpost. com/business/capitalbusiness/careercoach-are-writing-skills-necessaryanymore/2011/05/18/AFJLUF9G_story.html 2. As reported in http://poetsandquants. com/2014/05/19/what-employers-want-fromthis-years-graduates/ 3. Some exceptions are chronicled in http:// www.bloomberg.com/bw/articles/2013-07-03/bschools-get-serious-about-writing 4. https://www.bostonglobe.com/ opinion/2014/03/13/the-man-who-killed-satessay/L9v3dbPXewKq8oAvOUqONM/story. html 5. http://chronicle.com/article/WritingInstructor-Skeptical/146211 6. http://www.amazon.com/Farther-AwayEssays-Jonathan-Franzen/dp/1250033292

Organized by the INFORMS Analytics Section, INFORMS Chicago Chapter, and the University of Chicago

September 16, 2015 University of Chicago Gleacher Center

This is a half-day event with speakers from industry focused on the practice of analytics. The program is envisioned to include: a plenary talk, four breakout sessions on topics such as healthcare, finance and insurance, methods and techniques, and implementation and innovation of analytics initiatives. The conference will conclude with a panel session with industry leaders, followed by a networking reception.

QUESTIONS? EMAIL RogerLMoore@gmail.com

www.informs.org/Community/Conferences/ The-Practice-of-Analytics-Third-Annual-Conference-in-Chicago

a na l y t i c s

s e p t e m b e r / o c t o b e r 2 015

healt h car e a na ly t i c s

Healthcare industry continues to change â&#x20AC;Ś which is good news for data scientists Data is becoming more important in healthcare both for care delivery and for payment organizations.

By Rajib Ghosh

Recently, the healthcare industry experienced developments that are both interesting and thoughtprovoking. Some leading health insurance companies have announced merger and acquisition deals, IBM added new capabilities to its Watson Expert System, and Google separated its newly formed life sciences unit from its core business. The healthcare provider market continues to consolidate as big health systems get bigger with the acquisitions of hospitals or provider practices. Clearly, the industry is undergoing rapid shifts in the post Affordable Care Act (ACA) world. This is not uncommon when both demand and supply in an industry undergo rapid transformation. On the demand side, the ACA has introduced millions of newly insured people who bought insurance either via health information exchanges or through the Medicaid expansion program. On the supply side, health systems, driven by payment reform, are forced to focus on care outside of the hospital walls. Meanwhile data is becoming more important in healthcare both for care delivery and for payment organizations.

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

Will data scientists emerge as key figures in efficient healthcare delivery? Consolidation in the Insurance Industry The recent merger announcement of Anthem and Cigna made headlines. Prior to that Aetna and Humana also made headlines with merger talks. Those four and United Healthcare together cover 40 percent or more of all Americans. The merger will shorten the list to three. While the antitrust division of the Department of Justice will scrutinize all deals â&#x20AC;&#x201C; and the closing of those deals, if approved, will be almost a year away â&#x20AC;&#x201C; arguments on both sides of the proposals are becoming louder. It is clear that such mega mergers will leave consumers with fewer options in many markets. If insurance companies create monopolies like

a na l y t i c s

utility companies, the impact on product pricing can be significant. However, less competition is not necessarily bad. Less competition definitely gives more pricing power to the seller, but healthcare is a complex industry and consolidated entities are actually welcome. The Affordable Care Act encouraged healthcare delivery and payer organizations to join hands and create accountable care organizations (ACO) for better-coordinated care for patients at a lower cost. Bigger health systems can negotiate higher reimbursement from the commercial insurance payers. The latter will surely pass along the cost to the end consumers. When payers become bigger they negotiate lower costs and force providers to s e p t e m b e r / o c t o b e r 2 015

healt h car e a na ly t i c s

Healthcare is a data play, and very soon medical data scientists will become as important if not more important as front-end clinicians in delivering better patient care.

move to value-based care rather than volumebased care. To put this into perspective, other than Medicareâ&#x20AC;&#x2122;s modest success in value-based purchasing, the commercial insurance market so far has minimally treaded the value path. Besides, insurance regulators have checks and balances in place and can intervene if insurance companies do unwarranted price hikes or pocket huge profits from the lower costs. The other benefit of getting bigger is access to more data than before. When data is fragmented among various insurance payers and provider organizations within a given region, population health analytics becomes less effective. Consolidation of the payer industry will enable more analytics and hopefully better health management insights within different geographical markets. Image Analytics in Healthcare IBM recently purchased Merge Healthcare, a company focused on X-ray, CAT scan and other medical image storage and analysis software. IBM so far has made some headway into the healthcare market via various partnerships with Medtronic, Johnson & Johnson and Apple, but it still has not proven the value of a medical expert system. On paper, the potential is enormous, and Watson is a harbinger of what can be expected in healthcare within a decade, but IBM wants to create more value for its customers now. Adding the ability to analyze images in conjunction with structured or unstructured text data and detect with precision what a manual assessment might have missed will be a huge value proposition for Watson.

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

IBM claims that 90 percent of medical data comes from images, so clearly a big chunk of the medical data is currently absent from the purview of data analytics and automation. It will be interesting to see if Merge Healthcare’s solution seamlessly integrates with the knowledge engine of Watson, and if so, whether this can give rise to a new genre of analytics companies. Google Life Sciences In a recently announced corporate restructuring effort, Google has made its Life Sciences organization into an independent company and part of Alphabet, the newly formed holding company or “house of brands.” Google has been working on its contact lens glucometer solution for quite some time. Larry Page, the CEO of Alphabet, wants to scale his aspiration higher. Data and analytics drive Google. The Life Sciences division was created to make healthcare more proactive, predictive and preemptive rather than reactive. Google (or Alphabet) will continue to invest heavily in this data and its analyticsfocused healthcare unit to build smarter data mining solutions that will take input from various bio-signals. Supporting ecosystems such as body sensors, ambient sensors and gene sequencing technologies are developing rapidly. These sources are on the verge of a na l y t i c s

creating a data tsunami that only large data-focused corporations like Google with massive infrastructure and talent pool will be able to absorb and leverage. The contact lens glucometer is only the beginning of Google’s journey. Turning this division into an independent business unit with its own CEO is a game changer. In the future, healthcare data scientists will find lucrative job opportunities and ample scope to create impact in this outfit. Overall, the changes we see now have the potential to create a far-reaching impact on the U.S. healthcare system. As a consumer and a data analytics professional, I feel excited rather than intimidated by such changes. It is reinforcing my core belief that healthcare is a data play, and very soon medical data scientists will become as important if not more important as front-end clinicians in delivering better patient care. Rajib Ghosh (rghosh@hotmail.com) is an independent consultant and business advisor with 20 years of technology experience in various industry verticals where he had senior-level management roles in software engineering, program management, product management and business and strategy development. Ghosh spent a decade in the U.S. healthcare industry as part of a global ecosystem of medical device manufacturers, medical software companies and telehealth and telemedicine solution providers. He’s held senior positions at Hill-Rom, Solta Medical and Bosch Healthcare. His recent work interest includes public health and the field of IT-enabled sustainable healthcare delivery in the United States as well as emerging nations. Follow Ghosh on twitter @ghosh_r.

s e p t e m b e r / o c t o b e r 2 015

INFO RMS I ni t i at i ve s

CAP program rebrands, launches new website The Certified Analytics Professional (CAP®) program, the premier global professional certification program for analytics professionals, recently unveiled it’s new brand identity package, including a new website and URL, explainer video and other material. CAP’s new website can be seen at www.certifiedanalytics.org. There, visitors will have easy access to key information about becoming a Certified Analytics Professional. In addition, the website explains how organizations can add value to their workforce and enhance business outcomes, while individuals who are just beginning their analytics careers can learn more about the Associate CAP designation expected to launch this fall. As part of CAP’s rebranding, a new animated explainer video has been created that can be viewed on the CAP website as well as on YouTube. CIO Magazine named CAP one of “13 big data certifications that will pay off.” The Institute for Operations Research and the Management Sciences (INFORMS) launched CAP in 2013. Those who meet 22

a n a ly t i c s - m a g a z i n e . o r g

CAP’s high standards and pass the rigorous exam distinguish themselves and create greater opportunities to enrich their careers. For organizations seeking to enhance their ability to transform data into actionable insights that create greater value, CAP provides a trusted means to identify, recruit and retain top analytics talent. For more information, click here. CAP exams are available through a computer-based testing format, INFORMS continues to host paper-andpencil exams at selected sites, notably at INFORMS conferences. The next such exam will be held in Philadelphia on Nov. 4 in conjunction with the INFORMS Annual Meeting. To take this or any CAP exam, you must first apply at https:// www.certifiedanalytics.org/apply.php and be approved for the CAP examination. As always, applications remain open for the continuously available computer-based examinations. Test on your schedule and put CAP after your name. For more information, see www.certifiedanalytics.org.

w w w. i n f o r m s . o r g

Master of Science in

EnginEEring ManagEMEnt Delivered 100% Online

A Technical Alternative to the MBA Fast track option available â&#x20AC;&#x201C; Finish in 12 months! b 30-credit-hour curriculum with admission offered in Fall, Spring and Summer semesters b Gain skills in demand by industrial, research and commercial firms

Now Accepting Applications

DistanceEd.uncc.edu 704-687-1281

b Concentrations and graduate certificates available: Logistics and Supply Chains, Energy Systems, Lean Six Sigma and Systems Analytics

vie w poin t

Not taught in school (but useful in the ‘real world’) All models are wrong, but some are useful. – George Box The goal is almost always to develop the most accurate model possible.

By Scott Nestler, CAP, (top) and Sam Huddleston 24

Someone recently commented (and asked), “It’s a great time to be a quant grad, but what didn’t they teach you in school that you really need to know?” The first three things that occurred to us are: 1. The need to fully embrace the second half of the famous quote from George Box [1]. 2. How to communicate the results of your analysis to decision-makers. 3. The importance of creating a good visualization of your data/model (a subset of the previous task). With regard to models, one of the most important keys “in the real world” is understanding the difference between a great-fitting model and a useful model and the need to structure your research and modeling toward improving the decision-maker’s ability to make decisions rather than simply getting a high R2 value. In academic environments, students are often introduced to a series of methods and then provided data sets with which to practice. The goal is almost always to develop the most accurate model possible from the available methods. In the real world, it is often more important to focus on developing models that are useful and then

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

Source: http://www.sophia.org/tutorials/accuracy-and-precision

Figure 1: The authors recently saw an attempt to explain the terms precision and accuracy, and the difference between systematic and reproducibility errors. In lieu of three paragraphs of text, they suggest the graphic shown here. to improve the accuracy of the models over time under the constraint that their utility isn’t compromised. Examining the use of blackjack card-counting systems provides a good example of how model utility rather than performance is important in the real world. Blackjack Example Blackjack counting systems are not very “accurate” in the sense that even when the “deck is hot” (when the odds have swung in favor of the players vs. the dealer) there is still a very high probability that the dealer will win a given hand (and the bettor will lose his money to the

a na l y t i c s

house). These systems aren’t great predictors of the outcome of individual hands, as the model is often wrong. However, card counts are useful in that they can be employed in a systematic manner to win money, as documented in numerous recent popular books and movies. These counting systems (models) provide information that is useful for making decisions about what size bet to make, given the current state of the system. While many different card-counting systems have been developed, the most useful of these card-counting systems are relatively simple because more complex card counting systems (which may be more accurate) s e p t e m b e r / o c t o b e r 2 015

vie w poin t

are almost impossible to employ effectively for betting decisions in a chaotic casino environment. This is just one example of how, in the real world, there is often a trade-off space between model accuracy and model utility, with decisionmaking utility carrying more weight. The point about communication skills was hammered home when the first author was teaching at the Naval Postgraduate School in Monterey, Calif. In the biennial (every two years) program review, the number one comment was that the graduates came with all of the technical skills that they needed, but their ability to communicate to senior leaders and decision-makers, whose time is limited and span of involvement is high, was lacking. One common mistake made by many analysts is a failure to make a distinction between a technical report or presentation and an executive summary or decision brief. A technical report is a document written to record how an analysis was done (so that it can be replicated) and is designed to make a scientific and logical argument to support a set of conclusions. Therefore, a common outline for such a report might be: introduction, literature review, problem definition, methodology, results and conclusions. A presentation to a decision-maker using this format is likely to produce impatience and frustration – “Just get to the bottom line.” 26

a n a ly t i c s - m a g a z i n e . o r g

Executive Summary & Decision Brief The structure of a good executive summary or decision brief relies on the logical argument of the technical report (and should only be written once this logic is firmly established) but presents the logic in reverse order. An executive summary should lead with a brief statement of purpose to orient the reader, and then summarize the conclusions and recommendations, i.e., the bottom line up front, the results of the analysis (preferably in an easy-to-read chart), and briefly highlight the methodology and data used. One way to highlight the distinction between the logic of a technical report and an executive summary is that the logic of a technical report can be summarized with a series of “Therefore . . .” statements, while the logic of an executive summary should rely on a series of “Because . . .” statements. These ideas appeared in a blog post [2] by Polly Mitchell-Guthrie, chair of the INFORMS Analytics Certification Board (ACB), which oversees the Certified Analytics Professional (CAP®) [3] program, in 2013. She writes, “Much as we lament the shortage of graduates from the STEM disciplines (science, technology, engineering and math), it is arguably more difficult to find within that pool graduates who also have the right ‘soft skills.’” Polly points out that “selling” – yourself w w w. i n f o r m s . o r g

and your skills as an analyst – to convince others that you can solve their problems and improve their decision-making is critical. She suggests Daniel Pink’s book, “To Sell Is Human: The Surprising Truth About Moving Others” [4]. While “hard math” is critical in many instances, convincing someone that you have the technical skills to solve their problem is often more difficult. This is further highlighted in the seven domains of the CAP Job Task Analysis: business problem

framing, analytics problem framing, data, methodology selection, model building, deployment and lifecycle management. Not surprisingly, many of the supporting 36 tasks and 16 knowledge statements involve communications skills. These shortcomings among analysts are nothing new. In 2011, an Analytics magazine article [5] by Freeman Marvin, CAP, and Bill Klimack highlighted six “soft” skills every analyst needs to know: partnering with clients, working with teams,

SEARCHING? HIRING?

CAREER CENTER The INFORMS Career Center offers employers expanded opportunities to connect to qualified O.R. and analytics professionals. INFORMS career center offers a complete line of services to be used alone or in conjunction with the Job Fair at the 2015 Annual Meeting. Both give applicants and employers a convenient venue to connect. The Career Center is free to INFORMS members.

Access INFORMS CAREER CENTER

• More analytics jobs • More O.R. jobs • Preferred jobs • Featured jobs • Job alerts • Anonymous career profiles • Powerful searching capabilities

Learn more about the 2015 INFORMS Annual Meeting Job Fair, http://meetings.informs.org/philadelphia2015/jobfair.html

a na l y t i c s

s e p t e m b e r / o c t o b e r 2 015

vie w poin t

problem framing, interviewing experts, collecting data from groups and communicating results. Failure to effectively communicate results can lead to a project that is a technical success but has no impact. They propose that instead of dragging the decision-maker through the entire chronology of an analysis, tell a compelling story with a beginning, middle and end. One of the best ways to tell a compelling story is to use pictures (or graphics) to communicate the results of an analysis. Unfortunately, methods and principles for visually communicating the results of an analysis are often not taught in technical programs even though, as Mike Driscoll asserts in a popular online presentation [6], the ability to “munge, model and visually communicate data” are “the three core skills of data geeks.” Reviewing the work of Edward Tufte [7] and William S. Cleveland [8] provides an excellent foundation for visually communicating quantitative information. “Choosing a Good Chart” [9] by Abela is also useful, as it suggests an appropriate type of graphic for nearly any type of data and purpose. Summing Up In summary, first focus on developing useful models. Second, when communicating with decision-makers, start by describing the utility of those models – how can they be used and what difference will 28

a n a ly t i c s - m a g a z i n e . o r g

it make. Only after communicating the practical effects of the employment of the model/analysis should you communicate how you arrived at your conclusions (follow the logic of the technical report backwards). Finally, the most compelling way to communicate these ideas is through developing graphical products that clearly communicate the key results of your analysis. As they say, “A picture is worth a thousand words.” ❙ Scott Nestler (snestler@nd.edu), Ph.D., CAP, is an associate professional specialist in the Department of Management, Mendoza College of Business, University of Notre Dame, and a longtime member of INFORMS. Sam Huddleston (shh4m@virginia.edu), Ph.D., is an operations research analyst in the U.S. Army. Disclaimer: The views expressed in this article are those of the authors and do not necessarily reflect the official policy or position of the Army, the Department of Defense or the U.S. government. REFERENCES 1. https://en.wikiquote.org/wiki/George_E._P._Box 2. http://blogs.sas.com/content/ subconsciousmusings/2013/02/01/why-soft-skillsare-so-important-in-analytics-and-how-to-learnthem/ 3. www.certifiedanalytics.org 4. http://www.danpink.com/books/to-sell-is-human/ 5. http://www.analytics-magazine.org/januaryfebruary-2011/76-special-conference-sectionpeople-to-people.html 6. http://igniteshow.com/videos/mike-driscoll-threesexy-skills-data-geeks 7. http://www.edwardtufte.com/tufte/ 8. http://www.stat.purdue.edu/~wsc/ 9. http://extremepresentation.typepad.com/ blog/2015/01/announcing-the-slide-chooser.html

w w w. i n f o r m s . o r g

Become More.

Earn an

aaCSB-aCCrEditEd MBa froM thE

UnivErSity of SoUth dakota

Advance your career with an online Master of Business Administration with a specialization in Business Analytics through the Beacom School of Business at the University of South Dakota. Solve key business problems using big data in areas of:

4 Sales 4 Finance

4 Marketing 4 Economics

4 Operations Management 4 Strategy

Grad Performance in to 3% -

DIVISION OF CONTINUING & DISTANCE EDUCATION

Our students score in the 97th percentile in a national exam for graduate business students.

Real Professors –

Learn from the same faculty who teach on campus.

Be Ready to Lead –

Our programs emphasize decision making, problem solving and social responsibility.

Flexible Schedules –

Work full-time while earning your advanced degree. Our courses are built around manageable time commitments.

Learn more: www.usd.edu/onlinemba

414 E. Clark St. | Vermillion, SD 57069 800-233-7937 | 605-658-6140 grad@usd.edu

The University of South Dakota’s Beacom School of Business has been accredited since 1949. USD’s M.B.A. Program has been accredited by AACSB since 1965.

FORUM

How to run a cross-functional team Four principles for speedy, actionable, data-driven decision-making

By Rishi Padhi, Ankur Uttam and Achal Asawa

Organizations are very complex and have many self-governing systems that work in close coordination and interdependence with each other. These linkages lead to differing magnitude of decisions, often unexpected, that ripple across seemingly unrelated elements. Organizations often create cross-functional teams to align units with the hopes of creating synergies. However, creating and running a successful crossfunctional team has many challenges including but not restricted to: conflict management, prioritization, lack of accountability, missing decision-makers, group think, anchoring bias, intuition- and gut-based decisions and information asymmetry. This article focuses on speedy and actionable datadriven decision-making in such cross-functional teams. The authors discuss principles that have been tried and tested. Broadly, these are “set the right team and process,” “align organization and customer,” “make data-driven decisions” and “close the loop.” 1. Set the right team and process Occasionally, we hear functional stakeholders say clichés like, “I will get back to you on this” or “let me chew on this and revert” or “I have not thought about this and

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

Team meetings should focus on reaching solutions to maximize benefits for the organization rather than a function maxima. need more time” and there you lose another week, fortnight or maybe even a month. These are cases of decision-makers not being in the room, no prior agenda circulation or no prior preparation/brainstorming before coming to the meeting. The meeting owner should ensure that agendas are tight and pre-circulated, should follow-up on action items multiple times between two meetings and should have an update before the meeting, calling out laggards in the meeting and holding them accountable. The decision-making also needs to be collaborative and inclusive. Its objective (more on this later) and efforts should be a na l y t i c s

spent discussing and convincing the quorum rather than, “I know my function best and this is the right choice.” The focus should be to reach solutions to maximize benefits for the organization rather than a function maxima. 2. Align organization and customer The customer is the center of all organizational initiatives: impress and excite the customer and business will do well. Customers are all different, have different needs, are buying different things, undergo different stages of the purchase cycle and have very unique behaviors. The s e p t e m b e r / o c t o b e r 2 015

Forum

same efforts can emote different responses – making someone feel good and another bad. Also, remember that needs are not independent; they are a complex mesh of intertwined portfolios. When the committee decides initiatives for customers, it should ensure that none of the actions are in contradiction to its mission and vision in the short run or the long run. Contradictions and conflicts can inflict long-term setbacks that can take years to rebuild. A simple example could be creating a non-biodegradable packaging when the mission is to go-green or offering a perfectly shaped, sized and polished genetically modified fruit instead of organic food. 3. Make data-driven decisions In today’s environment of hypergrowth, hyper-diversity and hypersensitivity it is important to understand customers through their demographic, behavior and attitudinal traits. Once they are determined, the next important step is asking the right questions. Answering the right question incorrectly is more beneficial than answering the wrong question correctly. This needs a lot of front-load thinking, preemptive reasoning and buy-in from all functions. Questions should be forward-looking, hypothesisdriven or backed by results from previous experiments. 32

a n a ly t i c s - m a g a z i n e . o r g

The multi-dimensional explosion of data has almost overwhelmed today’s systems and has pushed the limits of computation and imagination. It is important that you have all the data, and if something is missing either procure it externally or setup systems to create that data internally or do both for long-term gains. Once this data beast has been tamed, the focus should be on correct and consistent use of data through a single source of truth. All functions should use this source for all decisions. Post data comes the key performance indicators (KPIs) and measurement framework. All functions have to be aligned on the KPIs – their definitions and the way they are calculated – to ensure that the business outcomes can be measured consistently. Once common and consistent KPIs are cemented, the measurement framework should be capable of viewing results in real time and concerned only with KPIs that lead to action (no trivial KPIs). The framework should be more like cockpits used to measure and drive business rather than dashboards that measure a specific metric or a group of functional metrics. Inquisitive and deep-dive analysis should be done in an agile fashion by each function and intermediate results reported and discussed in each crossfunctional meeting. The end outcome w w w. i n f o r m s . o r g

should incorporate both macro insights for understanding the overall strategy and micro insights to ensure short-term actionability, consumption and validation. The focus of learning should be on rapid iterations and quick experimentation. Similarly, any product development should happen in small and quick cycles and built in incremental fashion rather than long cycles. 4. Close the loop Teams need to ensure accountability with each other. All action items should be discussed (material pre-circulated so that only objections, clarifications and next steps discussions occur rather than the long monotonous presentation of the output) and functional leads should be responsible for the outcome of their experiments (especially if they are very different from hypothesized results). It is important to understand that a non-performing function can be penalized by reducing budgets, but this also leads to a function not being performed well. Once the results are vetted, the focus should be on consumption and ensuing discussions should include questions such as: What does the rollout mean for different functions? How can the negative side effect be minimized? Are there synergies with any other initiative in a different function? a na l y t i c s

Once the team is onboard then a clear roadmap with milestones, risks and contingencies should be created for implementation. Another key aspect is to incorporate learning (especially from failed experiments) into the next cycle of decision-making. Making mistakes is OK but repeating mistakes is expensive. Final Thoughts To set up your cross-functional team for success, ensure presence of decisionmakers, ruthlessly drive accountability, understand customers and be aligned to the organizational mission and vision. In addition, use all the data, use it consistently, judiciously and create actionable measures and outcomes. Finally, perform quick iterations and experiments, understand downstream dependencies on all functions and idolize past learning. ❙ Rishi Padhi is director of vertical marketing initiatives for eBay North America. He has more than 15 years of experience in CPG, e-commerce and strategy and holds an MBA from the Kellogg School of Management, Northwestern University. Ankur Uttam is a senior engagement manager and client partner at Mu Sigma. He has more than 10 years of industry experience in products and services companies solving a wide range of problems and holds an MBA from IIM Bangalore. Achal Asawa is a senior associate at Mu Sigma. He has more than three years of management and analytics consulting experience in e-commerce, retail and managed services. He holds a master’s degree in technology management from the University of Illinois, Urbana-Champaign.

s e p t e m b e r / o c t o b e r 2 015

cu stomer r e lat i o n s h i p m an age me n t

Cognitive computing for automating customer knowledge By Guy Mounier f your customer relationship management (CRM) system could actually think, would Elon Musk and other AI detractors want to kill it? Much has been made lately of the fearful admonitions about machine learning or AI by technology luminaries such as Musk, Stephen Hawking and Bill Gates. But would a CRM system â&#x20AC;&#x201C; employing cognitive computing to do a better job helping customers get what they need faster â&#x20AC;&#x201C; be dangerous? Would a CRM work flow that tells the advisor that a customer needs to adjust his portfolio to avoid a coming risk be a candidate to become emperor of the world? Of course not. Businesses are already using cognitive computing to make

a n a ly t i c s - m a g a z i n e . o r g

the bond with their customers stronger by making themselves more valuable, and they are gaining new revenue with that strengthened bond. Now to add some perspective, a cognitive computing platform is not a sentient being, but it is two generations beyond business analytics, and even one generation beyond machine learning. So what is it that cognitive computing delivers that makes it superior to previous generations of analytics and why should we embrace it? In three words, automated customer knowledge. In todayâ&#x20AC;&#x2122;s big data environment, companies are gathering and storing vast amounts of information about their customers, but it is rarely translated into w w w. i n f o r m s . o r g

actionable customer knowledge. Businesses know this information can help them better understand their customer, innovate products and services, and improve revenues, but they have not been able to accomplish that with previous technologies. What they require is a technology smart enough to make sense of all that data and transform it into actionable and measureable customer outcomes. Cognitive computing does just that automatically, providing a clear means to make the most of a company’s proprietary data. Some innovators are already disrupting their industries with this technology. They use it to intuit what customers need and to enhance customer insights in order to hone their cross-sell offers and increase revenues. At a time when corporations are trying harder than ever to keep and grow their customer base, this technology can offer a sustainable competitive advantage by capitalizing on existing relationships. Cognitive Computing Implications for CRM There are many applications for cognitive computing, but one of the most compelling is its ability to make sense of the great volumes of customer data – both static and dynamic – to learn, think and recommend. Global banks are using the technology to connect external and a na l y t i c s

internal information to accelerate revenue across corporate accounts managed by a diminishing number of bankers. Wealth management organizations are enriching their internal data with information about a customer’s key life events obtained from external public social and professional data sources, helping their advisors proactively up-sell to existing clients. For customer care operations, cognitive computing platforms help companies substantially reduce the time and resources they spend achieving customer issue resolution. Even more interesting is that these organizations are finding that the technology can be used to convert their customer care centers into profit centers. This nifty trick is accomplished by resolving customer issues quickly and proactively, making the customer feel known and appreciated, and then using the resulting “wow” moments to make up-sell recommendations during the very same client interaction. The Cognitive Computing Advantage Automated data science is the key to winning the battle for customer loyalty at scale. Companies can learn from their data and enrich their learnings with information from external sources. Unlike traditional business intelligence or big data projects that take months or years to execute, cognitive computing helps companies attain s e p t e m b e r / o c t o b e r 2 015

Cog n itive c om p u t i ng

Not only does cognitive computing offer a competitive advantage, it also helps unlock the value of data to counter emerging threats from nontraditional market entrants and competitors.

payback within weeks. Businesses are using this technology to leverage existing human and technology investments in CRM, customer experience tools, big data analytics and more. In essence they paid for all that IT infrastructure, and cognitive computing allows them to fully capitalize on it. Global corporations took some time to understand the benefits of cognitive computing, but that is now changing. Technology giants like IBM have used cognitive computing to help companies in a number of ways, e.g., the new “Chef Watson” app for Bon Appétit, utilization management decisions in lung cancer treatment and developmental assistance in Africa. Companies that are truly innovative don’t throw away or waste the vast amounts of data that they already have; instead, they are using cognitive computing to extract predictive value, i.e., answers to important business questions, and leverage their data for business gains in a more rapid fashion. Not only does cognitive computing offer a competitive advantage, it also helps unlock the value of data to counter emerging threats from non-traditional market entrants and competitors. It is one of the key technology solutions used today by businesses across the world to fend off competition, increase revenue from existing customers and thrive in fastchanging market conditions. The Technology Customer-facing employees often have to resolve issues in split seconds. They might struggle to identify the best course of action, which impacts customer experience. Cognitive computing can help these front-line professionals by providing suggested

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

actions or best practice recommendations. The technology does this by extracting actionable insights from structured and unstructured customer data. Sales agents and customer relationship professionals can use these best practice recommendations to identify up-sell and cross-sell opportunities. With this technology, companies can get a 360-degree view of their customer. Cognitive computing technology can be used to create a customer knowledge layer that enriches data collected over years of customer interaction and domain experience. The platform combines data (which resides in the CRM, care or account management system) with files from various internal and external sources. Once the information has been enriched, the technology continuously applies data enrichments, predictive recommendation algorithms and unsupervised semantic learning. The process is both continuous and dynamic. Unlike other predictive analytics that are rules-based and static, this technology is self-learning, real-time and contextual. Every interaction and result educates the platform, helping it become even more effective over time. Since the technology is lightweight and quick to deploy, cognitive computing can impact business revenues within weeks. This is achieved using a compressed platform-based methodology. By choosing a a na l y t i c s

technology partner that has relationships with key consulting firms and systems integrators, businesses can potentially get first use cases developed in weeks. Also, the cognitive computing technology is offered via a private cloud-based solution, making it easy for companies to deploy on their internal cloud or an industry standard external offering. The Potential We have just begun addressing the potential use cases of cognitive computing in the business world. Those leading the way know this to be true because they are continually discovering new revenuegenerating applications with customers across multiple industries. In an environment where businesses are finding it hard to gain new customers and even harder to retain existing customers, cognitive computing offers a chance to differentiate. Innovative companies understand that deeper customer knowledge is the key to surviving and thriving in this competitive landscape, and that cognitive computing not only enhances customer experience, but also delivers new revenue. ❙ Guy Mounier is co-founder and CEO at CustomerMatrix. Before that, he co-founded and ran BA Insight – a leader in agile information integration. Mounier holds a master’s degree in mathematics from Harvard University and a master’s degree in computer science and electrical engineering from École Centrale Paris.

s e p t e m b e r / o c t o b e r 2 015

Com petitive Adva ntag e

Customer Intelligence The impetus behind real-time business intelligence

By Swaroop Johnson ith the marketplace and product lifecycle operating under a constant threat of rapid innovation and change, customer intelligence holds the key to establishing transient competitive advantage. By enabling efficient and speedy marketing decisions, customer intelligence provides insights for products and marketing channels to focus on. It also provides faster, data-driven information on the satisfaction levels of different sales and operating channels, as well as financial and strategic information on various geographies.

a n a ly t i c s - m a g a z i n e . o r g

The emergence of big data and consumersâ&#x20AC;&#x2122; willful generation of unstructured data have led to the evolution of organizations from profit centers to data centers, transforming decision-making from an intuition-driven process to a data-driven approach. Understanding the Role of Customer Intelligence Customer intelligence is essentially a combination of customersâ&#x20AC;&#x2122; demographic data (pertaining to the individualâ&#x20AC;&#x2122;s age/ gender/location/income, etc.) and social and behavioral intelligence (pertaining to w w w. i n f o r m s . o r g

Most businesses tend to view social intelligence as a virtual and unruly dimension of customer intelligence. customer interactions and social behavior). It has become a critical ingredient in making effective strategic decisions, and it’s the foundation of building a comprehensive business intelligence system. Most businesses tend to view social intelligence as a virtual and unruly dimension of customer intelligence because they cannot control the amount of unstructured data being generated. However, when combining both the traditional and virtual dimensions, insights can be generated into customer preferences and evolving trends. Business intelligence encompasses all spheres of customer interaction and has evolved in recent years from basic dashboards and scorecards to a holistic a na l y t i c s

view of customers including customer preferences and real-time buying patterns. Customer intelligence, procured through a vast network of customer “datamarts,” has widespread applications from demand creation and customer acquisition to understanding why products were not purchased. Managing the Customer Experience Customer intelligence has evolved from its traditional dimension of only including demographic information to a vital tool in managing the customer experience. It provides insights into a consumer’s shopping behavior, product preferences and probability of purchase. Marketing s e p t e m b e r / o c t o b e r 2 015

Cu stomer I n t e lli g e n c e

strategies such as product bundling and cross-sell/up-sell â&#x20AC;&#x201C; which are heavily dependent on consumer preference and probability of purchase â&#x20AC;&#x201C; are enabled on a real-time basis today, thanks to the evolution of customer intelligence. To augment traditional customer intelligence efforts, organizations now need to introduce social intelligence to make strategic decisions and obtain holistic insights. The quality of outputs and insights generated from customer interaction can be improved if social intelligence is inclusive of customer intelligence, as online reviews and customer feedback have an added element of genuineness and willingness. Organizations can design regular updates and newsletters to create awareness and interest in their dynamic customer base with the use of social intelligence.

therefore improve return on marketing investments and customer response rates. For example, a major retailer in the Middle East wanted to increase revenue and profitability using personalized and targeted campaigns during the festive period of Ramadan. While using the business intelligence aspect of the data, the team was able to identify the best product affiliations to devise strategies around product bundling, up-selling, cross-selling and assortment planning. However, when customer intelligence (demographic data) and business-level data (transaction data) were used in combination, the team could build real-time campaigns throughout the promotional season by identifying the right marketing channels, depth of discount and assortment planning for each store segment, which helped the retailer achieve double their initial season targets.

Delivering Real-Time Results

Replacing Information Silos with Datamarts

Data privacy concerns have long been a major obstacle in utilizing customer intelligence for effective marketing and promotional activities. However, the advent of machine learning and abilities to predict parts of personally identifiable information (gender, age, etc.) through a customerâ&#x20AC;&#x2122;s social behavior (websites browsed, apps used) can help organizations map campaigns and messaging to respective customer preferences and 40

a n a ly t i c s - m a g a z i n e . o r g

While the case just described demonstrates the benefits of customer intelligence, in hindsight it is a classic example of benefits that can be drawn from a customer datamart. A well-planned datamart helps organizations integrate data from disparate source applications and operating systems for analysis, thereby giving companies a holistic picture of the business. Building a datamart appears simple, w w w. i n f o r m s . o r g

at least in terms of definition and using pre-defined steps, but it can be complicated, tiresome and irrelevant (to the business purpose) if important data and tool considerations are not taken into account. While cost, ease of use and familiarity of the tool are important organizational factors, other factors such as functionality, end reports and dashboards, as well as the amount of data that has to be handled, are important considerations when selecting the tools for building the datamart. Keep in mind the following: • Data points and definitions should be normalized to at least the third normalized factor to avoid redundancy and duplicate values in the datamart. • Schema definitions should take into account the granularity of data so they are adequately defined. Only transformed and cleaned data should be included in the datamart. • Most important, when separate datamarts are established, they should not be antagonistic in nature, i.e., affect the performance of the other datamarts already established in the organization and other departments. Enhancing Functional Benefits of the Datamart From a functional perspective, datamarts are built with the primary aim of building statistical models to aid a na l y t i c s

business decisions. However, important data considerations need to be assessed to achieve the expected and most optimum outcome from this exercise. The three important data considerations are: • Data horizon: how often or quickly the new data becomes part of the training model. • Data relevance: provides answers on whether the right assumptions have been made regarding the relevance of data in the model. • Data obsolescence: the time period before the data becomes irrelevant to the model. The age of data silos are disappearing, and it would be wise if organizations move from traditional modes of data collection to integrating all sources to a centralized repository to generate the best insights from all possible sources. There are incremental insights to be generated from unstructured data sources, which come to an organization free of cost. ❙ Swaroop Johnson is a consultant for analytical solutions at Blueocean Market Intelligence, a global analytics and insights provider. In his consultant role, Johnson proposes optimized analytical solutions across retail, banking, insurance, life sciences, utilities, entertainment and technology industries. For more information, visit www.blueoceanmi.com.

s e p t e m b e r / o c t o b e r 2 015

Hu m an Reso u rc e M a n ag e me n t

The magic of managed autonomy How online lender Enova International built a successful analytics team by balancing talent engagement with business priorities

By Vinod Cheriyan

nova International, a technology and analytics-driven online lender based in Chicago, was in the swing of rapid growth in 2013. The analytics team more than doubled to 51 in a matter of two years in order to support five new products in four countries. Since the business was entirely online, it meant we had to run as many as 10 models and provide loan applicants with a decision within seconds after they clicked â&#x20AC;&#x153;submit.â&#x20AC;? We hired a team with diverse backgrounds 42

a n a ly t i c s - m a g a z i n e . o r g

and specializations, from biostatistics to mechanical engineering, in order to keep improving our analytics capabilities and maintain our competitive edge. With all of this highly intelligent, intellectually curious talent, one of the key questions that was staring Enova in the face was this: How do we keep these highly talented folks engaged while at the same time drive maximum business value? This challenge is nearly universal in todayâ&#x20AC;&#x2122;s business environment. In order to survive, companies need to leverage w w w. i n f o r m s . o r g

analytics – not just linear and logistic regressions, but advanced algorithms such as deep machine learning. This requires attracting and retaining people with advanced degrees and specializations that were once purely in the purview of academia. It’s no secret that these talented team members are hard to come by and even more difficult to keep. Therefore, it is all the more important to keep them engaged. Looking back on how Enova was able to successfully balance talent engagement with business priorities, we have identified several approaches that worked for us and which we think could work for other growing tech companies in a variety of industries. 1. Create an environment that values engagement Quite ironically, the groundwork for addressing the question of team engagement vs. business priorities must take place much before an answer is needed. It is important to have an environment that values team members’ engagement and builds loyalty. At this point, the rationalist in you (or your CEO) might ask, why care about team engagement? After all, business success is the top priority, right? Well, yes. At the end of the day, work that drives business value is essential. However, a na l y t i c s

when thinking about the projects to assign to analytics team members, a quick reflection shows that completely ignoring their wishes is not a good idea. The cost of acquiring talent is very high. And turnover can be difficult to manage, especially for a hot area like analytics, where there is always a competitor trying to recruit out of your team. It should not be difficult to show leadership how a short-term gain in revenue may hurt in the long run if employee engagement is ignored. Once you’ve achieved buy-in, consistently communicate across your organization that while driving the business forward is the number one priority, people have a say in the types of projects they would like to work on and have the freedom to achieve business targets using their own unique strengths and methods. 2. Implement a “managed autonomy” organizational structure Once you have achieved a culture that values engagement, it is essential to put in place an organizational structure that is conducive to effectively allocating projects and resources in order to meet business goals and drive engagement. The solution of managed autonomy has two components: strategic alignment and tactical/resource allocation. s e p t e m b e r / o c t o b e r 2 015

man ag ed au to n omy

Strategic alignment. Strategic alignment sometimes requires big changes, even across departments. For instance, one of the first things that Joe DeCosmo, our CAO, did in his role was bring out the business intelligence and data services teams from under the technology department and put them under the analytics team so that all of our analytics operations could be streamlined. By its very nature, strategic alignment requires someone at the C-suite level with necessary leverage to make these highimpact changes. The second part of strategic alignment is investing in good hiring practices. As was mentioned earlier, it is important to hire the right people with specific skill sets that, when combined, form a balanced team. To facilitate this, Enova has a hiring manager dedicated to the analytics team who works closely with the CAO to understand the current and upcoming resource needs and helps with recruiting specific analytics team members. Proper resource allocation. Once the high-level strategic alignment is in place, the tactical resource allocation and project assignment can take place under the purview of the division managers. (In smaller analytics teams, the CAO and the division manager can be the same person). 44

a n a ly t i c s - m a g a z i n e . o r g

The first step in resource allocation should always be, “ask your team members.” This might sound like an obvious piece of advice, but it is often overlooked amidst day-to-day work. It is critical to have a good understanding of not only what employees’ specialty skill sets are but also what else they would like to work on. This information can be obtained through a direct one-on-one conversation or baked into annual goal-setting processes. For example, at Enova we support employees to drive their own career development. We provide clear expectations for each level on our internal wiki. Each team member can design their own goals and skill-development plan; they are even encouraged to recommend one or more projects that bring them closer to their next career step. The next step is to identify projects based on business goals and to give autonomy to the team member on how to implement them. (Hence the term “managed autonomy.”) 3. Give employees autonomy to leverage their strengths Properly assigning projects to meet business goals while keeping team members engaged can be challenging, but here are a few tips to make the process easier: w w w. i n f o r m s . o r g

Colossus Real-time analytics is critical for a fast and easy online customer experience. For a company like Enova with its business operating entirely online and spread over six countries, this is of paramount importance. However, implementing models in real time can be challenging. The legacy system on which we used to run models was a homegrown system written in C. Though it was fast, it was tightly integrated into our production system, making changes very difficult. Once built, models had to be translated for software engineering to implement them â&#x20AC;&#x201C; many times resulting in multiple cycles of debugging and testing. We wanted a system that was fast and that supported easy deployment of models without restricting modelers in the use of their choice of technology. So we set out to build a new real-time platform from the ground up. We evaluated multiple software solutions, and we finally partnered with a leading scientific

Focus on the larger goal â&#x20AC;&#x201C; not how it is achieved. Keep project direction limited to achieving business requirements; to the extent that it makes sense, do not specify how they are achieved. Managers need to keep in mind the strengths and aspirations of the team members as soon as business users come up with projects.

a na l y t i c s

computing vendor to design our custom realtime platform, Colossus. Through a phased approach, we gathered the initial requirements, had two paid prototypes made by external vendors, completed the design, built and implemented the system, and trained team members. The system was up and running within one year of the ideaâ&#x20AC;&#x2122;s inception. Not only has Colossus provided sub-second performance for running models (0.07 seconds on average), the average model deployment time has been cut in half. It also helps with the decoupling of development and running the models. Now, modelers can use their specialized knowledge and develop models using their favorite tools; once finalized, the models are translated and deployed on the Colossus platform. More importantly, the translation is done by the analytics team itself, so there is no longer dependency on the software engineering team for day-to-day model deployments.

It is important to identify and capture the business needs and separate them from the implementation details. For some projects, the manager can design a highlevel solution and match the best-suited team members to the project. For other projects, the manager can relay the business requirements to the team member

s e p t e m b e r / o c t o b e r 2 015

man ag ed au to n omy

and let the team member decide on the implementation. As one of Enova’s managers says, “You hire smart people. You shouldn’t have to tell them what to do. They should tell you the right solution.” Be flexible. It is also important to provide sufficient flexibility for the team members – both in terms of hours and in terms of deadlines. Though some projects are very deadline driven, many model-building projects do not have strict external deadlines. In these cases, the manager can negotiate an additional week or more, so that a team member can investigate and implement a new methodology. Not only does this give an opportunity for the team member to improve her skills, but also, if successful, it adds another tool to the toolbox. Be tool-agnostic. Another aspect of flexibility is being able to use the tool you know best and that which is suitable for the job. This is why Enova does not mandate one language for conducting analysis. Team members can use their technology of choice to create models. The worth of a model is judged by the ROI it produces. Though there are some advantages in having a uniform set of technologies, more often than not, different tools and technologies can in fact help with creating an innovative solution. For example, one of the business problems we faced at Enova was how to parse and store XML data into 46

a n a ly t i c s - m a g a z i n e . o r g

relational databases. The problem was made harder because proper schema definition was not always available. One of our experienced team members came up with an effective solution using a combination of Mathematica, XSLT and UNIX scripts. Always have an upside. On the other hand, managers must also be very vigilant that the team as a whole is making progress. They should use sound judgment when deciding when a research project has gone too far, and when the team should switch gears to produce something more tangible. A corollary of this is that when identifying research projects, it is important to design them such that even in the worst case, there are some useful by-products (e.g., reusable code to run a specific algorithm) that can perhaps be used in a future project. Have the right system architecture. But then, you would ask, how do you actually support all these different models? This is where another aspect of the organization comes in – having a flexible, centralized analytics platform. At Enova, successful models that make the cut are translated and deployed into Colossus Engine (see sidebar story). “Isn’t that duplication of work?” you might ask. Yes, it is. But the duplication is far outweighed by the benefits we obtain by letting team members work in the technology with which they are most comfortable. w w w. i n f o r m s . o r g

In Conclusion So there you have it. Successfully balancing employee engagement and production of business value can be achieved by instituting a “managed autonomy” architecture that consists of strategic alignment and proper resource allocation. The organization must be designed so that the analytics operations – developing models, providing business value and keeping employees engaged – are streamlined. Managers need to make sure that employees’ strengths

and aspirations are taken into account while designing projects. Providing sufficient flexibility, decoupling project goals from implementation details and nurturing a tool-agnostic environment are some of the ways to improve employee engagement. With engaged, talented employees, there is no doubt your analytics team will be ready to achieve incredible results. ❙ Vinod Cheriyan is a senior data scientist at Enova International. He is a member of INFORMS.

Join us this fall for PAW events focused on business, healthcare and government

BOSTON

Sept. 27 - Oct. 1, 2015

BOSTON

Sept. 27 - Oct. 1, 2015

WASHINGTON DC Oct. 13 - 16, 2015

Predictive Analytics World is the leading cross-vendor event for predictive analytics professionals, managers and commercial practitioners. This conference delivers case studies, expertise and resources to deliver the promise of big data.

SAVE

15%

on full conference passes with discount code: INFPAW15

www.pawcon.com s e p t e m b e r / o c t o b e r 2 015

Vir t ual I n t e r n a l C lo u d

Building global business collaboration Ideal infrastructure relies on three best practice pillars: security, freedom and mobility

By Rasesh Shah and Aliasgar Rajkotwala

wift advances in mobile devices and worldwide information networks have untethered employees from their traditional office environments. Most organizations face challenges in building a culture that ensures work-life balance and increased job satisfaction. The work culture needs to ensure that deadlines are met under relatively stress-free environments while ensuring data security. This article describes challenges and lessons learned along the way in enabling 48

a n a ly t i c s - m a g a z i n e . o r g

(l-r)

a virtual global workforce with freedom, mobility and the highest standards of security. Global Business Collaboration The advantages of mobility are so attractive that IDC projects 1.3 billion professionals are expected to work remotely by 2015, representing 37.2 percent of the total workforce. Adding another layer of complexity, most organizations understand that todayâ&#x20AC;&#x2122;s fight for top talent demands they offer employees the ability w w w. i n f o r m s . o r g

To successfully adopt a people-principled culture, organizations need a reliable and scalable infrastructure.

to complete assignments on time and still successfully manage their personal lives. Supporting security, productivity and employee freedom is easier said than done. Most organizations struggle to balance these competing goals with the right infrastructure that enables their employees a wide degree of flexibility while retaining the companyâ&#x20AC;&#x2122;s accountability to safety and security. Having a formal expression of â&#x20AC;&#x153;People Principlesâ&#x20AC;? creates an environment of trust and freedom by encouraging employees to define their goals, a na l y t i c s

as well as self-regulate and evaluate their performance. Such policy guidelines ensure an open, innovative, collaborative and meritocratic culture by empowering people with ownership and accountability to serve and satisfy clients. To successfully adopt a peopleprincipled culture, organizations need a reliable and scalable infrastructure. However, the use of a large-scale internal cloud infrastructure that supports a diverse set of software platforms also increases security risks, especially for global organizations. s e p t e m b e r / o c t o b e r 2 015

Glo bal B u s i n e s s Co lla borat ion

A people-principled culture will allow global collaboration to thrive with greater employee satisfaction by enabling freedom and mobility while safeguarding client data.

Such organizational systems need to have: • secure data transfer and storage capabilities; • the ability to tightly manage authorization and data access controls; • the ability to process large data volumes in the internal cloud infrastructure; and • safeguards to avoid potential data leaks. Building a Global Collaboration Infrastructure The ideal infrastructure relies on three best practice pillars: security, freedom and mobility. Security: The ideal mobility solution gives each employee access to two environments: production and personal. The production environment represents the security of an internal cloud that is controlled, monitored and secured using authentication, authorization and access protocols that meet or exceed the client’s security standards. The security-mobility bridge creates physical restrictions and control measures to ensure production data cannot be shared or copied to the personal environment, and vice versa. Freedom: The personal environment, in contrast, is an open space dedicated to the employee. This space is kept completely partitioned from the production environment, and gives users the flexibility to engage in activities outside of client servicing, such as company administration, training or personal email. The personal environment allows the employee to manage their personal business and check in on families and friends while upholding client and corporate data usage protocols.

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

Mobility: Mobility allows a global services organization to deliver the best service to clients, bringing a more flexible way of working from anywhere and anytime without compromising productivity, security or committed delivery timelines. For example, analysts and data scientists can build solutions that solve complex, big data problems using a range of software and analytical tools on their laptops, with standard security appliances that protect client data. Meanwhile, their consulting counterparts can log into secure servers to view project results from tablets while traveling or take a quick late night status call while their kids are sleeping, without risking a data breach. Lessons Learned As we set up our own virtual global workforce, we faced three key challenges when we decided to enable virtualization throughout the company: 1. sizing challenges, 2. user adoption, and 3. managing cost. Sizing challenges: Virtualization is a proven global technology, most often related to cloud-based servers for data storage and access. Desktop virtualizing is less common and is still in the early stages of adoption by the big data industry. When it comes to enabling a global team of analytics professionals, especially where freedom, mobility and a na l y t i c s

security are paramount, virtualizing a desktop environment is a necessity that is fraught with unique challenges. Before designing a desktop virtualization infrastructure, you first need to define the scope, scale and requirements of your solution. This requires careful planning and consideration across a number of dimensions including: â&#x20AC;˘ accurate sizing of computational and graphical requirements; â&#x20AC;˘ storage area networks and input/output ratios; and â&#x20AC;˘ network throughput. To overcome these challenges, we captured software, computation and storage requirements from a cross-section of teams representing different functions and locations. We reviewed and prioritized the requirements and developed a plan to address the most common and critical issues in a systematic manner. User adoption: Users demand high performance from their technology systems, and most users are very comfortable with how they experience local computing. Business users are particularly demanding. Even when they recognize the benefits, they are often slow to adapt to system changes. To move people out of their comfort zone into a virtual environment, the computing experience needs to be seamless with equivalent s e p t e m b e r / o c t o b e r 2 015

Glo bal B u s i n e s s Co lla borat ion

performance, with the look and feel of a local desktop. Unfortunately, achieving adoption often requires using the lemming approach. Find a group of highly motivated users to pilot the new system before rolling it out across the organization. This approach will allow you to fix bugs and create a case study with happy promoters to share their experiences to encourage others to give it a try. Ask your marketing team to help you build a promotional program using a “slow reveal” to generate curiosity and make the adoption process fun and engaging. Take the time to set up training with multiple iterations and repeated reminder tips, then gamify with adoption statistics showing how many users are coming onboard each week or month to build momentum. Managing cost: Virtualization can be expensive to enable collaboration across a large global workforce. Providing the benefits of security, flexibility and mobility comes at a cost. The goal, of course, is to find an approach at marginally higher costs as the current infrastructure to accommodate both production and personal environments. Practices such as standardizing the software and laptop models can contribute in a significant way toward standardizing the virtual environment and controlling the overall implementation costs. 52

a n a ly t i c s - m a g a z i n e . o r g

Conclusion To fuel the explosion of analytics demand around the world, a virtual internal cloud infrastructure can ignite workforce morale, boost productivity and create a robust, global data security environment. It is possible to build an internal infrastructure to support the development of analytics based on the three best practices of global collaboration: security, freedom and mobility. Choosing the right technology and sizing the infrastructure is very important to make the project successful and feasible from a technical and commercial point of view. Set up your virtual environment to engender a happy and productive workforce on your journey toward institutionalizing analytics. ❙ Rasesh Shah is senior vice president and CIO of Fractal Analytics, a global provider of predictive analytics and decision sciences headquartered in Jersey City, N.J. He has 23 years of technology experience building and delivering IT solutions and services. His mission is to scale analytics with quality, by providing a big data environment that enables real-time analytics for clients. Aliasgar Rajkotwala is associate director and global head of IT at Fractal Analytics. He has more than 15 years of diversified experience from conceptualizing to implementation of a large, scalable and secure infrastructure. He is currently the architect of information security and governance at Fractal, responsible for driving innovation strategies that further evolve an advanced big data IT infrastructure in support of global Fortune 500 clients.

w w w. i n f o r m s . o r g

2015

Philadelphia Think the INFORMS Annual Meeting is only for Academics? THINK AGAIN.

Join Us at the Analytics Practice

“Meeting Within a Meeting” at the 2015 INFORMS ANNUAL MEETING Industry Sessions Span Entire Meeting

November 1–4, 2015 Daniel H. Wagner Prize: Excellence in O.R. Practice Franz Edelman Award Reprise INFORMS Prize: Effective Integration in Organizations Innovative Application in Analytics Award Reprise Industry Job Search Panel Discussion

Attend the Analytics Section Meet and Greet Monday, November 2, 6:15–7:15 pm Meet the people that care about what you care about.

Receive Individual Consulting

Analytics Magazine readers are invited to meet one-on-one or in small groups with INFORMS Analytics Section officers in Philadelphia. We’ll help you arrange. For details, email tom.fink@informs.org.

November 1–4 Philadelphia, PA

Ne w r even u e st r e a m s

Predictive analytics in the publishing industry The transformation of traditional print media and the growing intelligence behind monetization strategies

By Arvid Tchivzhel eclining circulation, bankruptcy, short attention spans of millennials, technology disruption, new competition from content aggregators â&#x20AC;Ś the list goes on and on. These are all common threads in most articles that address print and digital media since I have been working with newspapers and media companies over the last seven years. The stereotype of the slowly dying newspaper is easy to accept at face value. And itâ&#x20AC;&#x2122;s true that since the financial crisis, dozens of newspapers shuttered the printing press and hundreds drastically cut staff to remain solvent. However, two closely related narratives are rarely discussed and go against the stereotype:

a n a ly t i c s - m a g a z i n e . o r g

1. Traditionally profitable and bulky newspapers finally had to come to terms with a competitive and rapidly changing technological environment, and thus had to face the exact same challenges as airlines, retailers, hotels, car rental companies and cable television. 2. Newspapers have begun and are continuing to expand their use of analytics and data to improve profitability, much like companies in the competitive industries mentioned above have used it to improve yield and customer retention. This article will focus on the second narrative, which is indeed an outcome from the first narrative. w w w. i n f o r m s . o r g

The newspaper of the new millennium is embracing big data technologies and learning how to become a data-driven decision-maker.

While the term “creative destruction” may not elicit sympathy from those who have suffered due to layoffs and budget cuts, the positive externality is it has led to leaner, more adaptive organizations. The newspaper of the new millennium is embracing big data technologies and learning how to become a data-driven decision-maker. No longer is advertising, circulation and pricing led by the phrase “this is what we have always done,” but rather the business wing of the newspaper (the other being journalism) is looking to data scientists and predictive modeling to drive important decisions. Among the things being driven by data: 1. products/prices to offer new and existing customers; 2. setting advertising rate card and inventory premiums; 3. targeted marketing messages (who gets a na l y t i c s

a sports-themed creative vs. a business newsletter, etc.); 4. customized customer service experience (which complaining customer should get a 50 percent discount vs. who should be kept at full price); 5. where to set the paywall (how much content should be paid vs. free); and 6. what content to share on social media (and when to share on Facebook vs. other platforms). Countless decisions can be made by looking at data. However, data is only as valuable as the person looking at it. Derrick Harris from Gigaom.com summarized it nicely: “Data is the new oil – it’s very valuable to the companies that have it, but only after it has been mined and processed.” The analogy makes some sense, but it ignores the fact that many people and companies don’t have the means to collect the data they need or the ability to process it once they have it. A lot of us just need gasoline. Mixing Gasoline and Newspapers The most amazing impact of the Internet is that publishing content online makes it immediately accessible to every part of the globe. This allows a local newspaper like the Peninsula Clarion in s e p t e m b e r / o c t o b e r 2 015

ANALYTICS AN D PUBLISHING

Kenai, Alaska, to be readily available to a curious individual in Durban, South Africa, who can read about salmon dip-netting on the North Beach. Seemingly, the potential audience of any digital content could be infinite if the content and marketing is right. Detailed information of user engagement, both paid and anonymous, is revealed using big data tools by capturing and crunching billions of rows of data. Before these tools, publishers had only surveys, focus groups and angry customers to tell them what resonated and what did not. By using “revealed preference” instead of “stated preference,” publishers can learn exactly which users and what content drives engagement. While this is a simple exercise in data clustering to a data scientist, user segmentation has informed how content producers sell and market their products, helped to set advertising rates on premium inventory and created more effective targeted marketing campaigns. At last, publishers can understand in real time what content attracts the loyal users and what content attracts the flyby user. I will refrain from naming specific newspapers for the sake of privacy, but below are some quick examples of how newspapers have used detailed engagement data: 56

a n a ly t i c s - m a g a z i n e . o r g

1. The newspaper: a large metro daily with a rigid paywall at five free articles per month. The analytics: Detailed log-level website traffic was analyzed and segmented using simplistic clustering and categorization. The output: Major “content anchors” were identified that attracted loyal and returning users. A significant subset of users anchored themselves exclusively in sports content with minimal overlap in other content. This was compared to the other significant user group anchored in news content who overlapped with politics, community news and some business content. The outcome: It was determined that the users anchored in news content would be more amenable to the “all access” offer, which already existed. However, the users anchored in sports content would respond to a targeted offer so a digital-only sports product at a lesser price point was proposed to monetize this highly segmented audience. 2. The newspaper: a digital-only national sports publication with premium content. The analytics: Detailed loglevel website traffic was analyzed and matched at the user level (via account ID and login ID) to a customer data warehouse. Detailed customer records, transactions and payments were reconciled with online behavior to create a complete profile of individual customers using both w w w. i n f o r m s . o r g

online and offline data. Predictive analytics, using survival analysis, were applied to measure expected retention based on historical data. The output: A Customer Lifetime Value (CLV) score was assigned to each customer using the mix of online and offline data as well as predictive modeling. Customers were scored 1-100 in terms of expected lifetime value. The outcome: It was proved statistically that recent new starts had a significantly lower CLV compared to new starts in earlier years, exposing the need to redesign and reinforce the value of the product to new subscribers. New initiatives were taken to redesign the website, adjust promotional pricing and offers, create a mobile-friendly experience and improve new start retention and value. The outcomes above were achieved by doing fairly straightforward analysis, data matching and segmentation. The next example uses a more advanced algorithm to find an optimal outcome. Subscribers are the New Black Advertising has and will continue to be an important revenue stream for media companies, but many content producers have embraced paywalls. The Wall Street Journal, New York Times and the Financial Times, for example, understand that valuable content should not be given away whether it is printed on a na l y t i c s

a piece of paper or if it is published via content management system to an iPad app. Valuable content and compelling journalism can indeed attract a paying audience large enough to forego potential ad impressions and revenue. Given the generally low cost per thousand impressions (CPMs) sell-through rates and click-through rates, it is not surprising that publishers are monetizing through a digital subscription or print/digital â&#x20AC;&#x153;all accessâ&#x20AC;? subscription. Recent trends even suggest there might be a larger shift in the industry toward paid journalism rather than ad-supported journalism. That said, there is still a risk of losing valuable advertising revenue if the paywall is too aggressive. The challenge is determining what should be paid vs. free. The challenge posed above is common for publishers and represents the constant pendulum between advertising dollars and paid digital subscribers. On the one hand, leaving a website completely free guarantees the maximum amount of ad revenue and impressions, but it creates a conundrum for publishers asking subscribers to pay a recurring monthly price to a print product while leaving the same content online for free, as well as foregoing potential new digital subscriptions. On the other hand, a hard paywall will maximize the amount of paid digital subscriptions and reinforce s e p t e m b e r / o c t o b e r 2 015

ANALYTICS AN D PUBLISHING

the product but put at risk all the ad revenue that could be possible. It is not an “either-or” question; the optimal point for a paywall is where both are maximized simultaneously. Using detailed traffic patterns and advertising data, this problem can be solved. Measuring user engagement by geography (local vs. non-local), device, day of week, time of day and what content drives engagement can be used to model expected conversion probability per user segment. Measuring key advertising metrics such as click-through, sellthrough and CPMs by the exact same segments (geography, device, seasonality, day-part, content and segment) shows the expected ad value in an “apples-to-apples” context. An algorithm that performs an optimization formula can be deployed to dynamically adjust and reset the paywall based on continuously optimizing the balance between advertising and subscribers. The outcome from this type of analysis might be to set the paywall for entertainment content (let’s assume high CPMs, high sell-through and low subscriber engagement) to be light, perhaps even letting this content be completely free. Conversely, for sports content (let’s assume low CPMs, low sell-through and high subscriber engagement), the paywall could be set much more aggressively. In 58

a n a ly t i c s - m a g a z i n e . o r g

this way, advertising revenue is realized for content with high ad value, but subscriber revenue can also be realized for content that is higher value for subscribers. Local vs. non-local traffic, mobile web vs. desktop vs. native apps, etc., also could be split in similar ways to take advantage of very different ROIs for the two major revenue streams. Data and analytics can bring a newfound longevity to a traditional print product and allow efficient use of new digital content. Publishers are learning to take advantage of analytical tools and analysts who can help find insights and create real action plans to drive actual dollars to their bottom line. They are learning to become efficient and competitive, as have other industries that found themselves in the same position. ❙ Arvid Tchivzhel, a director with Mather Economics (www.mathereconomics.com), oversees the delivery and operations for all Mather Economics consulting engagements, along with internal processes, analytics, staffing and new product and services development. He has led numerous consulting engagements across various industries, working with econometric modeling, forecasting, economic analysis, statistics, financial analysis and other rigorous quantitative methods.

Subscribe to Analytics It’s fast, it’s easy and it’s FREE! Just visit: http://analytics.informs.org/

w w w. i n f o r m s . o r g

Freedom of location. Freedom from tuition. Freedom to take your career everywhere. The Master’s Degree in Quantitative Finance from WorldQuant University: the world’s first online, tuition-free master’s degree, designed to give you the skills you need to stand out in the financial services industry. Learn more and apply today at www.worldquantu.org.

cor porate p rof i le

Analytics at IDA Analytics has always played a vital role at the Institute for Defense Analyses. IDA assists government agencies in addressing important national security issues, focusing particularly on those requiring scientific and technical expertise.

By David E. Hunter

Founded in 1956, the Institute for Defense Analyses (IDA) is a not-for-profit corporation that currently operates three federally funded research and development centers (FFRDCs): the Systems and Analyses Center, the Science and Technology Policy Institute and the Center for Communications and Computing. FFRDCs are unique, independent entities sponsored and funded by the U.S. government to meet long-term technical needs that cannot be met as effectively by existing governmental or contractor resources. These entities were initially established after World War II as the U.S. government and Department of Defense (DoD) tried to find a way to maintain continued access to the technical and scientific expertise that had proved so valuable during the war effort. IDA’s sole business is operating its three FFRDCs. Collocated with IDA headquarters in Alexandria, Va., IDA’s Systems and Analyses Center assists the Office of the Secretary of Defense as well as other government agencies – such as the Department of Homeland Security, the Director of National Intelligence and the Department of Veterans Affairs – in addressing important national security issues, focusing particularly on those requiring scientific and technical expertise. IDA exists to promote national security, preserve the public welfare and advance scientific learning by analyzing,

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

evaluating and reporting on matters of interest to the U.S. government. IDA’s goal is to empower the best scientific and strategic minds to research and analyze the most important issues of national security. To achieve this goal, IDA maintains a highly educated and diverse research staff. In fact, more than 90 percent of IDA’s researchers have advanced degrees, with the majority having earned doctoral degrees in a technical field. Each year, IDA researchers execute hundreds of projects for government sponsors. For each project, research teams comprising the precisely necessary scientific, technical and analytical skills – and with disparate life experiences and backgrounds – are assembled a na l y t i c s

from across IDA’s eight research divisions. IDA’s flat organization and culture of internal collaboration allow researchers to easily and collegially interact with each other and the Institute’s leaders. Analytics has always played a vital role at IDA. IDA researchers do not use any one specific analytical technique or tool to solve all problems, but rather seek to employ the most appropriate techniques to address each individual research question. Following are some examples of analytical techniques used by IDA researchers to address specific research questions: IDA text analytics (ITA). ITA is a customized software capability, built on proven open source components, for s e p t e m b e r / o c t o b e r 2 015

cor porate p rof i le

ITA uses a variety of different techniques based on machine learning and natural language processing to facilitate rapid insight discovery.

exploratory analysis of highly heterogeneous collections of documents (i.e., exploratory search). It is employed on a wide range of problems at IDA from cybersecurity applications to program evaluation. ITA uses a variety of different techniques based on machine learning and natural language processing to facilitate rapid insight discovery. It supports both search (e.g., looking for specific information) and discovery (e.g., interactive browsing to reveal information for which one may not have even known to look). ITA goes beyond simple keyword search tools through its implementation of analytics-powered facets (or filters), which allow an analyst to view a document set along different dimensions (or through various lenses). These facets, in addition to other visualizations and auto-generated reports, provide rich overviews of the entire information space and can help answer various researchable questions of interest. ITA utilizes numerous techniques to implement such facets including, but not limited to: • key phrase and concept discovery, • topic clusters, • s upervised machine learning facets – technology area and document type, • customizable entity extractions, and • file metadata facets – location, time and format. ITA is actively developed with new functionality made available regularly such as graphbased visualizations of text corpora, duplicate detection and various other reports to help answer researchable questions.

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

SAVE THE DATE APRIL 10-12, 2016

HYATT REGENCY GRAND CYPRESS, ORLANDO, FLORIDA Learn how analytics and O.R. can maximize the value of your data to drive better business decisions.

http://meetings2.informs.org/analytics2016

cor porate p rof i le

FFRDCs: Unique capabilities “FFRDCs were established to provide the Department of Defense with unique analytical, engineering and research capabilities in many areas where the government cannot attract and retain personnel in sufficient depth and numbers. They also operate in the public interest free from organizational conflicts of interest and can therefore assist us in ways that our industry contractors cannot.” Hon. Ashton B. Carter Under Secretary of Defense Acquisition, Technology and Logistics

Statistical analyses and data mining. Statistical analyses and data mining are some of the more common analytical techniques employed by IDA researchers. These tools were particularly valuable a few years ago when the Department of Veterans Affairs (VA) asked IDA to investigate the causes of perceived inequities in the VA’s disability compensation program. This multi-billion dollar program provides monthly payments to military veterans with injuries or disabilities incurred or aggravated during military service. 64

a n a ly t i c s - m a g a z i n e . o r g

The IDA research team met with VA leadership, traveled across the country to interview hundreds of claims adjudicators, and – perhaps most importantly – collected and analyzed data on millions of disability compensation awards. During this project, IDA researchers formulated hypotheses based on their gained understanding of the VA adjudication process. Further, they employed advanced data mining and exploratory data analysis techniques to find additional factors and interactions implicit in the data. From the hypotheses and data, IDA employed statistical analyses to test each hypothesis and to quantify the amount of the observed variations that is accounted for by each factor. The IDA analysis identified the main factors contributing to the observed variation, dispelled some common misperceptions, and made policy recommendations to further improve the equity and consistency of disability compensation awards. Econometrics and optimization. Of the roughly $500 billion dollar annual defense budget, about 20 percent is used for procurement – buying systems for our armed forces to use, as opposed to developing new systems or maintaining the systems we already have. What we buy ranges from bullets to ballistic missiles, half-ton trucks to M1 tanks, and inflatable w w w. i n f o r m s . o r g

Is the largest association for analytics in the center of your professional network? It should be. • INFORMS allows you to network with your professional peers and others who share your interests • INFORMS Connect, the New Member-only, Online Community Lets You Network with your Collagues • INFORMS Provides Unsurpassed Networking Opportunities Available in INFORMS Communities and at Meetings • INFORMS Offers Certification for Analytics Professionals • INFORMS Helps You Take Leadership Roles to Help Build your Professional Profile • INFORMS Career Center Provides You with the Industry's Leading Job Board

Join Online Today! http://join.informs.org

cor porate p rof i le

rafts to aircraft carriers. About $50 billion per year is spent on major defense acquisition programs (MDAPs), the most sophisticated and most expensive military systems. These are the nation’s investment portfolio against future military operational needs. Understanding the cost, schedule and risks of any one major program is complicated. Understanding their interactions and behavior as a portfolio is even more daunting. IDA has been working with DoD to develop and improve a decision-support tool that models the cost and schedule of all MDAPs simultaneously. This tool, called “PortOpt,” allows DoD analysts to predict the likely cost and schedule impact of proposed procurement schedule changes, and to find practical schedules that minimize total procurement cost across all programs, given a fixed budget and fielding requirements. It also provides a means to estimate the overall cost and schedule impact on existing programs of adding a new program or cancelling a program. These capabilities have direct applications to affordability analysis, portfolio analysis, and reprogramming in response to unexpected budget reductions. PortOpt gives DoD the ability to identify – in days or weeks, rather than weeks or months – opportunities for savings, feasible responses to disruptions or impending budget crunches. 66

a n a ly t i c s - m a g a z i n e . o r g

At the heart of PortOpt are two key analytical tools. The first is an econometric model of how future procurement costs for each program would vary as a function of production schedule. Because this is a causal model, sophisticated statistical techniques are required to distinguish the effect of schedule on cost from the equally common effect of cost on schedule, or the effect of technical challenges on both. The second key tool is a large mixed integer linear program (MILP) that approximately describes the problem of finding the minimum-cost set of simultaneous schedules subject to constraints on annual budget, latest permitted fielding dates, minimum and maximum production rates, plant capacity constraints and practical limits on which production schedules could be implemented in real life. The MILP uses piecewise-linear approximations to the econometric cost functions, resulting in a formulation with thousands of binary variables, tens of thousands of continuous variables, and tens of thousands of constraints. Discrete event simulation modeling. IDA uses discrete event simulation modeling to assess defense weapon systems. A suite of similar models called IMEASURE, built with ExtendSim, examines various aircraft types; IDA has used these simulation models to examine w w w. i n f o r m s . o r g

fighters, helicopters, cargo aircraft and unmanned aerial systems. Given a particular system’s reliability and maintainability (RAM) performance and a target operational capability metric, the model can be used to independently estimate maintenance manpower requirements by job specialty and/or appropriate spare stock levels. Alternatively, given a particular set of available maintenance manpower and spares stock, the model can assess the system’s operational capability (mission capable rate, sortie generation rate, operational availability). We often use the model at IDA to make assessments and predictions of operational test performance or to estimate program unknowns (e.g., manning or sparing resource requirements) to support independent cost estimates. Aside from the RAM inputs, there are many additional data and modeling assumptions required to run the model (aircraft turn durations, abort rates, mission schedule, etc.), contributing to the intractability of solving such analytical problems without simulation. Cost/benefit analyses. For the U.S. Department of Homeland Security, IDA modeled the cost and benefits of early warning and detection technologies employed to defeat biological weapons attacks on major U.S. cities. This work involved modeling the dispersion of a na l y t i c s

aerosolized pathogens in various venues such as an outdoor park in Chicago, O’Hare International Airport and Grand Central Terminal in New York City. The lifecycle costs and benefits (i.e., reduced mortality and morbidity) of these technologies were simulated over a range of pathogens, venues and operation cycles. The results suggest that net present value of all the technologies was positive. The Future Over the past 60 years, IDA researchers have been asked to provide independent analytic assessments and analyses on a wide range of public policy questions. In fact, the variety of interesting work is one of the oft-mentioned reasons why so many talented people enjoy working at IDA. The future looks to be no different. While it is not possible to predict with certainty the specific research questions that IDA will be asked, they will most assuredly continue to involve some of the more critical aspects of national security. And IDA analysts and researchers will continue to leverage the latest analytic techniques to provide government decision-makers with high-quality independent assessments. ❙ David E. Hunter (dhunter@ida.org) is an assistant director in the Cost Analysis and Research Division (CARD) at the Institute for Defense Analyses (IDA), as well as IDA’s representative to the INFORMS Roundtable. He is a member of INFORMS.

s e p t e m b e r / o c t o b e r 2 015

co n fer e n c e p r e v i e w

Philadelphia meeting offers analytics, education & networking The 2015 INFORMS Annual Meeting will be held Nov. 1-4 at the Pennsylvania Convention Center and Philadelphia Marriott Downtown hotel. The conference will include an “Analytics Practice Meeting Within a Meeting,” featuring industry-oriented sessions, many of them organized by the Analytics Section of INFORMS. Other highlights include presentations by finalists for the Daniel H. Wagner Prize for Excellence in O.R. Practice and the INFORMS Prize for Effective Integration in Organizations, as well as a panel discussion on the industry job search. In addition, the Analytics Section of INFORMS will hold a “Meet and Greet” reception on Monday, Nov. 2, from 6:15-7:15 p.m., and Analytics magazine readers are invited to meet with Analytics Section officers at the conference to receive one-on-one or small-group consulting. For details, email tom.fink@informs.org. Throughout the conference, an impressive list of plenary and keynote speakers will deliver talks from a wide variety of application areas, including data sciences, healthcare, energy, security, service systems, logistics, the environment and more. Margaret 68

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

Philadelphia is known for its arts, culture and history.

Brandeau of Stanford University, Michael Jordan of U.C. Berkeley, Bill Rouse of Stevens Institute of Technology and Alfred Spector of Google will deliver plenary presentations. The meeting will offer not only educational opportunities but also several networking opportunities. The Welcome Reception will be held on Sunday evening (Nov. 1), subdivision meetings will be held predominantly on Monday evening (Nov. 2) and the General Reception will be held Tuesday evening (Nov. 3). The Career Fair will provide an opportunity to meet and collect resumes from many job seekers and the ability to set up private interviews. The city of Philadelphia is known for its arts, culture and history. It has more outdoor sculptures and murals than any other American city and is home to one of the largest art museums in the United States, the Philadelphia Museum a na l y t i c s

of Art. Philadelphia is also the home to 67 National Historic Landmarks, including Independence Hall and the Liberty Bell. The Reading Terminal Market, a historic public market that offers everything from locally grown produce to table linens, occupies the ground floor of the center. There are a wide variety of restaurants in the area surrounding the Convention Center and conference hotels. A tour of Boathouse Sports will highlight the diversity, creativity and productivity that this leader in customer performance apparel is known for. The convention center and conference hotel are located just 10 miles from Philadelphia International Airport and easily accessible by the Southeastern Pennsylvania Transportation Authority. For more information regarding the 2015 INFORMS Annual Meeting, visit http:// meetings2.informs.org/philadelphia/. â?&#x2122; s e p t e m b e r / o c t o b e r 2 015

co n f er e n c e p r e v i e w

WSC 2015: Social and Behavioral Simulation WSC is the central meeting place for simulation researchers, practitioners and vendors spanning all disciplines and working in industry, government, military, service and academic sectors.

The Winter Simulation Conference (WSC) has been the premier international forum for disseminating recent advances in the field of dynamic systems modeling and simulation for more than 40 years. In addition to a technical program of unsurpassed scope and quality, WSC is the central meeting place for simulation researchers, practitioners and vendors spanning all disciplines and working in industry, government, military, service and academic sectors. WSC 2015 will be held Dec. 6-9 in Huntington Beach, Calif., at the Hyatt Regency Huntington Beach Resort and Spa. The appeal of simulation is its relevance to a diverse range of interests. WSC has always reflected this diversity, and WSC 2015 aligns with and expands upon this tradition. For those more inclined to the academic aspects of simulation, the conference offers tracks in modeling methodology, analysis methodology, simulation-based optimization, hybrid simulation and agent-based simulation. For those more inclined to the application of simulation, tracks include healthcare, manufacturing, logistics and supply chain management, military applications, business process modeling, project management

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

About the Competition The purpose of the competition is to bring forward, recognize, and reward outstanding examples of operations research and the management sciences in practice. The client organization that uses the winning work receives a prize citation; the authors of the winning work receive a cash award.

Call for Entries

A $15,000 Competition with a $10,000 First Prize Application Deadline: October 14, 2015

Key Dates for the Competition Wednesday, October 14, 2015

Deadline to provide a single pdf document containing a three page summary of your achievement, and a cover page with a 60-word abstract, and the name, address, phone number, and affiliation of each author.

Monday, December 14, 2015

Finalists will be selected based on the summaries and the INFORMS/CPMS verification process.

Friday, February 12, 2016

Deadline for finalists to provide a full written paper.

Monday, April 11, 2016

Each finalist group will give an oral presentation of their work in a special session at the INFORMS Conference on O.R. Practice in Orlando, Florida, April 10â&#x20AC;&#x201D;12, 2016.

Entry Requirements Visit the website www.informs.org/edelmanaward for detailed information. Entries should report on a completed practical application and must describe results that had a significant, verifiable, and preferably quantifiable impact on the performance of the client organization. Finalist work will be published in the January-February 2017 issue of Interfaces. Any work you have done in recent years is eligible, unless it has previously been described by a Franz Edelman Award finalist. Previous publication of the work does not disqualify it. Anyone is eligible for the competition except a member of the judging panel.

E-mail Submissions

Please e-mail your submission to: trick@cmu.edu ................................................................. Michael Trick Chair, 2016 Edelman Award Competition Committee

CPMS

The Prac ce Sec on of INFORMS

Deadline for Applications is Friday, October 30, 2015.

The UPS George D. Smith Prize is created in the spirit of strengthening ties between industry and the schools of higher education that graduate young practitioners of operations research. INFORMS, with the help of CPMS, will award the prize to an academic department or program for effective and innovative preparation of students to be good practitioners of operations research, management science, or analytics.

2015 UPS Smith Prize Winners Sauder School of Business, University of British Columbia - Center for Operations Excellence

The prize will be awarded to an academic program or department, and will include a trophy and a $10,000 award. The UPS George D. Smith Prize will be announced at the 2016 Edelman Gala at the INFORMS Conference on Business Analytics and Operations Research.

For more information, questions can be sent to Robin Lougee, 2016 Smith Prize chair ing ties at rlougee@us.ibm.com. hen

call for applications 2016

You are the best in OR/MS/Analytics education?

t eng

academia str

www.informs.org/smithprize

industry

ups george d smith prize

co n fer e n c e p r e v i e w

and construction, homeland security and emergency response, environmental and sustainability applications, and networks and communications. The theme for WSC 2015, “Social and Behavioral Simulation,” is timely and relevant. The explosion in the number of simulations that include descriptive models of behavioral and social decisionmaking is creating both opportunities and challenges to the business, operations and scientific communities. Presenters will discuss how simulation can help. In addition to special tracks on social and behavioral simulation and agentbased simulation applications, conference keynote speaker Joshua Epstein, professor of Emergency Medicine and director, Center for Advanced Modeling in the Social, Behavioral and Health Sciences at Johns Hopkins University, will speak about the rise of artificial social simulation. The military keynote speaker is Timothy H. Chung, deputy director of the Consortium for Robotics and Unmanned Systems Education and Research at the Naval Postgraduate School, who will speak about multi-agent coordination for information gathering applications. A distinguished speaker lunchtime program features the “Titans of Simulation,” Pierre L’Ecuyer and Averill M. Law. 72

a n a ly t i c s - m a g a z i n e . o r g

The Modeling and Analysis of Semiconductor Manufacturing (MASM) is a conference-within-a-conference featuring a series of sessions focused on the semiconductor field. The Industrial Case Studies track affords industrial practitioners the opportunity to present their best practices to the simulation community. The Simulation Education track presents approaches to teaching simulation at education levels ranging from K-12 to graduate and professional workforce levels. Finally, WSC provides a comprehensive suite of introductory and advanced tutorials presented by prominent individuals in the field, along with a lively poster session, Ph.D. colloquium and a new attendee orientation. The WSC is designed for professionals at all levels of experience across broad ranges of interest. The extensive cadre of exhibitors and vendor presentations, the meetings of various professional societies and user groups, along with the various social gatherings give all attendees the opportunity to become involved in the ever-expanding activities of the international simulation community. For more information, visit http:// wintersim.org/2015/. ❙ — Charles M. Macal, general chair of WSC 2015. He is a member of INFORMS. w w w. i n f o r m s . o r g

ADVERTISE IN ANNUAL MEETING ENEWS DAILY

2015

http://meetings.informs.org/philadelphia2015

Philadelphia

Reserve your advertising space now in the Annual Meeting eNews Daily, the ofﬁcial news source of the INFORMS Annual Meeting. Each morning attendees will receive the Annual Meeting eNews Daily in their in-boxes. eNews Daily provides a preview of the day’s key events and a recap of the previous day’s happenings. Over 60% of the 5,000+ attendees open and read eNews Daily.

DELIVERY SCHEDULE:

eNews Daily will be delivered by 5:00 a.m. on the followings days: Wed., 10/28 - preview edition to all meeting pre-registrants Sun., 11/1 - all attendees

Mon., 11/2 - all attendees

ADVERTISING RATES: any 1 day - $250

any 2 days - $450 any 3 days - $600 any 4 days - $700 all 5 days - $750

Tues., 11/3 - all attendees

Wed., 11/4 - all attendees

QUESTIONS: Please direct questions to Tom Fink, Senior Marketing Manager INFORMS, 5521 Research Park Drive, Suite 200, Catonsville, MD 21228 Phone: 443-757-3561, e-mail: tom.fink@informs.org

CAREER CENTER

Job Seekers: Find Your Next Career Move INFORMS Career Center contains the largest source of O.R. and analytics jobs in the world. It’s where

careercenter.informs.org JOB SEEKER BENEFITS • POST employers to you. • SEARCH • PERSONALIZED job alerts notify you of relevant job opportunities right to your in-box. • ASK the experts advice, resume critique and writing, career assessment test services and more!

www.informs.org | 800.446.3676

Five- M in u t e A n a lyst

Battle of Hoth

The battle is good for analysis – the pace is slow enough to track, but there is enough action to be interesting.

By Harrison Schramm, CAP 74

This time, we take a military historical [1] look at the Battle of Hoth from “Star Wars Episode V: The Empire Strikes Back.” There are several reasons to choose this among the Star Wars battles to analyze. First, Empire is considered by many to be the best Star Wars movie, and the battle of Hoth is the only major “force on force [2]” engagement in it. Most importantly, it is good for analysis – the pace is slow enough to track, but there is enough action to be interesting. Surprisingly, I was unable to find a time trace of the battle on the Internet. We analyzed the battle and the relative value of the units by watching the film repeatedly and pausing to write down major events. We mark the beginning of the battle of Hoth as Han Solo’s destruction of the Probe Droid 19:11 in the film, and its ending when Luke departs for Dagobah at 36:18. We ignore the space battle and focus on the ground battle. In order to get a good “apples to apples” comparison of forces, we weight them as shown in Table 1.

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

Unit

Value

Starting Strength

Rebel Soldier

200

Imperial Soldier

200

Rebel Ground Gun

Snow Speeder

AT-AT

100

Table 1: Comparative values (Expert Judgment) for the battle of Hoth. Relative strengths are based how each unit compares to a single soldier.

We use Lanchester aimed fire equations to fit the Hoth battle data, which assume that the attrition to each side is a function of the elapsed time, amount of force on their side remaining and a fixed effectiveness multiplier. Lanchester’s equations have been a focus of practical and academic study since their introduction in 1916 [3] by a large number of operations researchers [4]. The equations are: . We’ll follow the convention of describing forces as “Red” and “Blue,” which still works because we can think of it as lightsaber color: Red for Empire, Blue/ Green for Rebels. When describing a set of ODE’s like this to non-technical audiences, the best approach is to say something like: “The Empire forces are drawn down at a rate proportional to the number of active Rebel forces.” Having the equations, we are left with determining the “best value” for a na l y t i c s

, which is our main task. One method is to find parameters to make the model agree with the losses at the end of the battle. This is somewhat unsatisfying because it throws away the time-dependent battle trace, which we took time to record. It is also trivial (why?). We choose instead to minimize the standard deviance. Herein, we turn a data analysis problem into a nonlinear optimization. In Excel, we measure the deviance of each trace (Rebel and Empire) by using the =SUMXMY2() function. Next, we minimize this quantity, but we need to do it in a way that does not unfairly favor either side. The resulting “formulette” is MIN(MAX(R_error, B_error)), encouraging solver to make both of the errors – which happen to be correlated – equal by “pushing down” the greater error. Technique is important in solving a problem like this. Before invoking any solver, we manipulate the parameters by hand until a reasonable incumbent solution appears. Human beings are very good at this. We then place the problem “in a box” by providing solver with upper bounds on effectiveness. It appeared that the Evolutionary Solver outperformed GRG non-linear. We determine that the effectiveness coefficients for the Rebels and Empire are 1.2 and .8, respectively. This means that the Rebels are approximately 1.5 times more effective than the Empire. s e p t e m b e r / o c t o b e r 2 015

Imperial Troops Enter Base

Five- M in u t e A n a lyst

Are the Rebels a good fighting force? Not really. Historically (all the way back to Sun Tzu), defenders should have a 3:1 advantage because they are dug in. If the two forces were equal, we would expect the Rebels to have a 3:1 apparent advantage. A few “loose ends”: 1. We do not count Admiral Ozzel as a casualty, even though he had “failed for the last time” and was Force-choked during the battle. 2. The rebels lose their base, and it is unclear whether the Empire lost a star destroyer hit by Ion Cannon, with a complement of 46,785 on board [5]. Counting either or both losses would have hopelessly skewed the results. 76

a n a ly t i c s - m a g a z i n e . o r g

Figure 1: Time-trace of the Battle of Hoth. Solid lines indicate observed data, dashed lines are the Lanchester “best fit.”

3. There was action in the Battle of Hoth that we did not see in the movie. This is a problem that military historians have as well! ❙ Harrison Schramm (harrison.schramm@gmail.com), CAP, is an operations research professional in the Washington, D.C., area. He is a member of INFORMS and a Certified Analytics Professional. notes & REFERENCES 1. After all, it did happen long, long ago in a galaxy far, far away! 2. By “force” we mean military force involving combat arms, not “The Force” involving midiclorians. 3. F. W. Lanchester, 1916, “Aircraft in Warfare: Dawn of the Fourth Arm.” 4. Including me! Schramm, H. and Gaver, D., 2014, “Lanchester for Cyber – the MixedEpidemic Combat Model.” 5. Numbers from Wookiepedia.

w w w. i n f o r m s . o r g

CERTIFIED ANALYTICS PROFESSIONAL Analyze What CAP Can Do For You ®

www.certiﬁedanalytics.org

thin k in g a na ly t i ca lly

Moon rover Figure 1: Searching for “science points.”

A new moon on a nearby planet has recently been discovered, and a rover has been sent to the surface to explore. Many interesting features on the new moon have been identified, but the rover has a limited travel distance ability so not all sites can be visited. In order to prioritize the scientific value of a site, “science points” have been assigned to each of the sites of interest (as indicated by the numbers inside the circles). The rover starts at location D6. Due to the battery limitations of the rover, it can only travel a maximum distance of 25 kilometers. Use a direct line between sites to calculate travel distance (for example, the distance between G3 and H4 is 1.41km).

By John Toczek John Toczek is the senior director of Decision Support and Analytics for ARAMARK Corporation in the Global Operational Excellence group. He earned a bachelor of science degree in chemical engineering at Drexel University (1996) and a master’s degree in operations research from Virginia Commonwealth University (2005). He is a member of INFORMS.

Question: What is the maximum number of “science points” you can achieve before the rover’s batteries run out? Send your answer to puzzlor@gmail.com by Nov. 15. The winner, chosen randomly from correct answers, will receive a $25 Amazon Gift Card. Past questions and answers can be found at puzzlor.com.

a n a ly t i c s - m a g a z i n e . o r g

w w w. i n f o r m s . o r g

OPTIMIZATION GENERAL ALGEBRAIC MODELING SYSTEM High-Level Modeling The General Algebraic Modeling System (GAMS) is a high-level modeling system for mathematical programming problems. GAMS is tailored for complex, large-scale modeling applications, and allows you to build large maintainable models that can be adapted quickly to new situations. Models are fully portable from one computer platform to another.

State-of-the-Art Solvers GAMS incorporates all major commercial and academic state-of-the-art solution technologies for a broad range of problem types.

GAMS Integrated Developer Environment for editing, debugging, solving models, and viewing data.

Efficiency Benchmarking for the Finnish Energy Authority Companies operating electricity networks are typically situated in a local monopoly market. In order to encourage reasonable electricity prices for the end user, this industry is therefore state-regulated in many countries. The Finnish Energy Authority (energiavirasto) provides incentives for companies to improve their efficiency while at the same time promoting investment in modern and reliable infrastructure. It establishes general efficiency targets and relies on benchmarks to adequately compare network operatorsâ&#x20AC;&#x2122; cost efficiency. A primary benchmarking challenge is to capture the vast heterogeneity of this sector. For more than 15 years the Finnish Energy Authority has been concerned with developing and improving efficiency benchmarkEnvelopment of Data) modeled with GAMS has been applied as a benchmarking tool with great success. In GAMS, the Finnish Energy Authority has found a precise, flexible and practical efficiency benchmarking tool that is capable of capturing the specific complexity of the sector. For further information please contact Matti Ilonen - Matti.Ilonen@energiavirasto.fi

www.gams.com

Kotkoa/Shutterstock.com

ing. Since 2012 the StoNED method (Stochastic Nonsmooth