Tech Spark 8 - Data Visualization

Visualizing the Invisible The power of data visualization in finance TECH SPARK | 1

Contents The Data Visualization Issue 04 In This Issue 06 Why Data Visualization for Finance?

12 Charting in Finance: Which, When?

14 Presenting Data Graphically 16 Data Viz: Human-Centered Reflection 18 Visualization Tools in Finance (L) 26 Big Numbers: Explore or Explain 20 The Three-Tier Approach of Visualizing Data 26 Playing with Patternization 32 Case Study: Visualizing Risk Data 36 A Data Scientist Can’t Be Replaced with Software (Alex)

2 | TECH SPARK

13 Milliseconds

80

Fact 01

The human brain can process entire images that the eye sees in as little as 13 milliseconds.

Fact 02

About 80% of the information we take in comes through our eyes.

Fact 03

60,000

The human brain processes images 60,000 times faster than words.

Fact 04 Data Visualization

65%

After 3 days,10% of heard information can be recalled.

With data visualization, recall is up to 65%.

TECH SPARK | 3

In This Issue...

Unlocking the potential of your data is an art and a science, but it doesn’t have to be overwhelming. With a basic understanding of how different datasets should be visualized, along with a few fundamental design tips and best practices, you can create more accurate and effective visualizations. In this issue, we provide some of our thoughts and guidance around data visualization for finance.

THE GREATEST VALUE OF A PICTURE IS WHEN IT FORCES US TO NOTICE WHAT WE NEVER EXPECTED TO SEE. JOHN W TUKEY 1977

4 | TECH SPARK

TECH SPARK | 5

Why Data Visualization?

As humans we respond to, and process, visual information better than any other type of data. In fact, the human brain processes images 60,000 times faster than text, and 90% of information transmitted to the brain is visual. We are visual by nature and can use this expertise to enhance data processing and improve the effectiveness of information organization and interpretation.

Our ability to scan, recognize, recall and interpret helps us to understand more accurately and at speed. The human brain is an amazing pattern-recognition tool, and it can detect changes in size, color, shape, movement and texture, very efficiently. We unconsciously perceive the world, constructing meaningful things and objects by combining the individual parts.

Some of the main benefits of visualizing data include: Answering a question Visualization can provide a quick, high-level summary of the main information contained in the data.

Posing new questions Quite often the initial data investigations can lead to more questions and further exploration.

6 | TECH SPARK

Exploring and discovering Sometimes the data shows some unexpected patterns and outliers — data points which are well outside the normal data range. Exploring these data points can lead to new discoveries.

Communicating information Graphical representations of data are more effective as a means of communication than long textual files. A story can be told more efficiently, and the time taken to understand a picture is a fraction of the time required to understand the textual data.

Supporting decisions Visualization can provide quick answers and improve situational awareness. In turn, this can lead to faster and more timely decisions.

Increasing efficiency A well-designed chart can save a lot of time otherwise needed to read pages of numbers and long textual reports. This time can be better spent on making sound business decisions.

Inspiring Sometimes you may come across a visualization that really appeals to you or presents the data in a different and more effective way, challenging your own practice. It can be useful to use this inspiration to enhance the way you present the data.

Using innovative data visualizations can make your audience more enthusiastic about what you are trying to communicate to them. In turn, this can inspire your audience to take action based on your more persuasive visualization.

TECH SPARK | 7

hich chart

Charting in Finance

8 | TECH SPARK

Working with lots of different financial datasets can be overwhelming and trying to decide on the right chart to display data, even more so. What data should you be tracking? What actually matters? What’s the best type of visualization to provide insights and actionable information? Couple this with time pressures and the need for an accurate visual representation, and its hard to know where to start.

We have compiled a list of common charting types which will help you make some of those decisions based on whether you want to compare, see a distribution, analyze trends or understand data relationships.

When?

TECH SPARK | 9

Ranking Ranking& &Comparing ComparingData Data

Correlated CorrelatedData Data

Deviated DeviatedData Data

When Whenyou youneed needtotoorder orderdata data based basedononitsitsvalue valueororimportance. importance.

Shows Showsthe therelationship relationshipand and correlation correlationbetween betweentwo twooror more moredifferent differentdata datapoints. points.

Shows Showsthe the(+(+oror–) –)from froma abaseline baseline reference referencepoint. point.

Column Column & line & line

Spine Spine chart chart

Bar Bar chart chart (row (row version) version)

Lollipop Lollipop (column (column or or row row version) version)

Standard Standard barbar charts charts display display thethe ranks ranks of of values values much much more more easily easily when when sorted sorted into into order order

Bar Bar chart chart (column (column version) version)

Useful Useful forfor showing showing when when things things areare growing growing or or leading leading and and thethe style style can can emphasize emphasize thethe data data a lot a lot more more

Dot Dot plot plot

The The orientation orientation chosen chosen can can enhance enhance and and make make clear clear thethe story story youyou areare trying trying to to telltell

Vary Vary slope slope

Bubble Bubble chart chart Ideal Ideal forfor showing showing related related data data with with different different start/min start/min and and end/max end/max points points across across multiple multiple categories categories

Stacked Stacked barbar chart chart (column (column or or row row version) version) Allows Allows youyou to to add add another another dimension dimension of of data data insight insight

Paired Paired barbar chart chart (column (column or or row row version) version) Allows Allows youyou to to add add a variation a variation of of thethe same same data data next next to to itself. itself.

A good A good example example may may bebe previous previous years years

10 | TECH SPARK

Similar Similar to to a scatterplot a scatterplot butbut with with a a data data point point third third dimension dimension of of point point size size (indicated (indicated byby thethe circle circle size) size)

Scatter Scatter plot plot

Strip Strip plot plot Shows Shows ranking ranking variance variance across across data data (e.g., (e.g., over over time time or or between between different different categories) categories)

Shows Shows thethe relationship relationship between between quantity/amount quantity/amount (bar) (bar) and and a rate a rate or or benchmark benchmark

Shows Shows thethe relationship relationship between between two two or or more more datasets datasets based based onon individual individual points points onon thethe x/yx/y axis axis

Strip Strip plots plots areare likelike scatter scatter plots plots butbut areare continuous continuous and and often often contained contained within within a category a category

Ordered Ordered symbols symbols

Connected Connected scatter scatter plot plot

Used Used to to show show bigbig differences differences between between categories categories where where seeing seeing thethe finefine detail detail is not is not soso important important

Bump Bump chart chart Used Used to to show show thethe differences differences in in rank rank over over a given a given time time period period

Similar Similar to to a scatter a scatter plot plot butbut with with a time a time dimension dimension that that shows shows change change

Heatmap Heatmap Gives Gives youyou a helicopter a helicopter view view of of thethe data data which which allows allows youyou to to identify identify patterns patterns quickly quickly

Shows Shows opposing opposing deviations deviations from from a a central central point point spine spine (e.g., (e.g., buy/sell) buy/sell)

Diverging Diverging barbar Shows Shows negative negative and and positive positive values values onon a chart a chart in sequence in sequence

Diverging Diverging stacked stacked barbar Shows Shows both both negative negative and and positive positive values values plus plus anyany in-between in-between areas areas

Positive/negative Positive/negative filled filled line line Alternative Alternative visualization visualization to to a diverging a diverging barbar chart chart where where thethe toptop or or bottom bottom data data points points areare joined joined byby lines, lines, and and thethe shaded shaded area area represents represents thethe entire entire data data

Time-Based Time-BasedData Data

Flow-Based Flow-BasedData Data

Shows Showschanging changingtrends trendsover overtime time either eitherintra-day intra-daymovements movementsoror long-term long-termtime timeseries. series.

Shows Showsvolumes volumesororintensity intensityofof movement movementbetween betweentwo twooror more morestates statesororconditions. conditions.

Line Line chart chart

Fan Fan projection projection The The usual usual ‘simple’ ‘simple’ way way to to show show a a change change over over time time

A chart A chart that that joins joins a line a line graph graph forfor observed observed past past data, data, and and a range a range area area chart chart forfor future future predictions predictions

Priestley’s Priestley’s timeline timeline

Sparklines Sparklines Shows Shows change change in ainmeasurement, a measurement, such such as as temperature temperature or or stock stock market market price price over over time, time, in ainsimple a simple and and highly highly condensed condensed way way Horizon Horizon

Candlestick Candlestick Shows Shows day-to-day day-to-day activity, activity, these these charts charts show show opening/closing opening/closing and and high/low high/low points points of of each each dayday

Area Area chart chart

A chart A chart that that displays displays cumulative cumulative areas areas over over time time rather rather than than individual individual values values

Shows Shows thethe major major transfers transfers or or flows flows within within a system. a system.

They They areare helpful helpful in in locating locating dominant dominant contributions contributions to to anan overall overall flow flow Waterfall Waterfall

Similar Similar to to a straight a straight timeline timeline thisthis chart chart hashas a ‘jump’ a ‘jump’ dimension dimension to to show show significant significant events events

Calendar Calendar heatmap heatmap

Helps Helps in understanding in understanding thethe cumulative cumulative effect effect of of sequentially sequentially introduced introduced positive positive or or negative negative values values

Chord Chord

A visual A visual way way of of showing showing temporal temporal patterns patterns (daily, (daily, weekly, weekly, monthly) monthly)

Renko Renko charts charts

Displays Displays time time onon thethe Y axis Y axis rather rather than than thethe X axis X axis and and works works well well forfor portrait-based portrait-based screens screens such such as as mobile mobile Streamgraph Streamgraph

Used Used to to visually visually represent represent quantities quantities that that change change over over time time

ArcArc timeline timeline

Usually Usually displays displays discrete discrete values values of of varying varying size size across across multiple, multiple, different different categories categories

Vertical Vertical timeline timeline

Sankey Sankey

Milestone/date-based Milestone/date-based timeline timeline associated associated with with displaying displaying historic historic events events

Commonly Commonly known known as as a Gantt a Gantt chart, chart, it isit used is used to to cover cover a vast a vast period period of of time, time, highlighting highlighting keykey date date points points

Circle Circle timeline timeline Shows Shows metric metric behaviour behaviour over over time time in relation in relation to to a baseline a baseline or or horizon horizon

Straight Straight timeline timeline

Used Used in financial in financial technical technical analysis analysis to to measure/plot measure/plot price price changes changes

A graphical A graphical method method of of displaying displaying thethe inter-relationships inter-relationships between between data data in ainmatrix a matrix

Network Network This This type type of of visualization visualization displays displays relationships relationships between between entities entities

Step Step chart chart A stepped A stepped lineline chart chart can can bebe useful useful when when youyou want want to to show show thethe changes changes that that occur occur at at irregular irregular intervals intervals

TECH SPARK | 11

Distributed Data

Data Magnitude

Whole v Part

Show values (highlighting non-uniformity) in a dataset and how often they occur.

Shows relative size comparisons (‘counted’ numbers).

Displays how a single entity can be broken down into its component elements (parts).

Mekko chart

Stacked column

Barcode plot

Cumulative curve Ideal for displaying all data in a table, they work best when highlighting individual values

Box plot

Shows how unequal a distribution is: y axis is always cumulative frequency, x axis is always a measure

Frequency polygons Shows multiple distributions by displaying the median (center) and range of data

Violin plot Used with complex distributions (data that cannot be summarized with simple averages)

Shows data points that start to overlap and plots density, rather than points

12 | TECH SPARK

Uses dots to show the presence of a feature or phenomenon

A popular way of showing part-to-whole data but displaying no more than five segments

Donut

Shows multiple values and benefits from highlighting important values

Bullet

Like a pie chart, the center can be used to include more information about the data (e.g., total)

Voronoi Shows a measurement against the context of a target or range

Group symbol

Dot density The usual way to show statistical distribution – gaps between columns are tight to highlight the ‘shape’

Pie chart

Parallel co-ordinates

Hexagonal binning

A simple way of showing part-to-whole relationships. Works best with a few components

A very visual way of representing data and best used with whole numbers or values

Emphasizes individual points in a distribution. Points can be sized to an additional variable

A way for showing distributed and comparative data such as age and sex population distribution, effectively.

Back-to-back histograms Histogram

Pictograms

Displays multiple distributions of data. Use a maximum of three or four datasets

Beeswarm

Population pyramid

Shows the size and proportion of non-complicated data at the same time

Ideal to show count data or highlight individual elements within a dataset

DIsplays points in areas—any point within each area is closer to the central point than any other region

Grid plot Ideal for showing percentage information. Works best when used on whole numbers

Ranking & Comparison Man y ca teg orie s

Conservative

1,233,032

2018

43%

Heatmap

Paired bar chart

1,655,770

Grid plot

30% UK Stock US Stock

Bar chart

Bonds S/T Invests

10%

2019 Growth

52%

1,944,322

2020

25%

Aggressive

2,322,311

Stacked column

18%

Magnitude

Diverging bar

Pictograms

HSBA IAG IHG 0.00% +10.92% +15.45% +10.35%

JMAT -7.93%

KGF -5.30%

LAND LGEN +0.76% +0.71%

NMC 0.00%

NXT -1.23%

OCDO -6.22%

LSE +4.77%

PHNX POLY PRU -0.48% +12.00% +8.77% RTO +6.96%

SBRY -2.09%

SMIN Diverging SMT SN SPX SSE stacked +6.29% +8.77% +7.56% +1.92% +1.33%

STAN -1.65%

UU -6.29% Straight timeline

Mekko

LLOY -0.85%

RSA -2.15%

+/- Filled line

Chart Suggestions Group symbol

HLMA +0.94%

RIO RMV RR +1.25% -11.29% -1.44% Donut

AUTO

HL -6.62%

Few c ateg orie s

Ordered symbols

62%

ANTO

VOD -3.77%

WPP WTB -1.03% +11.22%

GBP/USD H1.1674 L1.1559

Basic Material...

1.1950

Real Estate

Line chart

Technology

1.1900 1.1850

Industrial G...

1.1800

Cumulative curve

Candlestick

Financials

Services

1.1750

Network

1.1700

Services

1.1643

Circle timeline

Dot density

Analyze Volitility

1.1600

Energy

1.1550 Healthcare

Consumer Goods

r st Di

Utilities

ib ut ion

Sankey

Barcode plot

1.1500

m

e

20%

Connected scatter

Voronoi

13%

AHT

CRDA CRH DCC DGE EVR EXPN -3.07% +11.29% +6.54% +1.12% +6.96% +4.36%

Bubble

21%

ADM

BDEV BHP BKG BLND BNZL +11.93% +3.18% +3.35% +0.47% +4.75% +4.55%

Balanced

21%

ABF

Co +0.93% +5.30% -5.22% +0.48% -1.00% -11.77% rre l

Deviation

2017

21%

16%

AAL

t ar P v

ion at

11%

21%

W ho le

Strip plot

Ti

1.1450 1.1400

18:00

20:00

06:00

12:00

18:00

Chord

Flow

TECH SPARK | 13

Presenting Data Graphically Data visualization is one of the best ways for our brains to analyze and comprehend data. By placing data in a graphic form, the decision-maker can more easily understand the significance of the data, compare multiple pieces of data, or even set goals and objectives based on the data.

Sourced d 01

DATA Accuracy

What does the audience care about?

What are

Empathize 04

The importance of data visualization in interpreting data — key facts: • The human brain processes entire images that the eye sees for as little as 13 milliseconds • About 80% of the information we take in comes through our eyes • The human brain processes images 60,000 times faster than words • A fter 3 days, only 10% of information heard can be recalled.

With data visualization, recall is up to 65% As organizations deal with more complex datasets that have scores of variables and subsets, data visualization is even more instrumental in our making sense of it. The world is becoming more complex, and organizations that adapt accordingly will be the ones most likely to prosper. That said, the process of making sense of complex data and presenting it graphically is pretty straightforward. With the right subject matter experts it’s possible to end up with groundbreaking insights based on the right analysis, spotting the right story to tell and visually delivering that story in a meaningful way. The diagram opposite illustrates the process.

14 | TECH SPARK

If there is no story you have the wrong dta

Fin

VISUALIZE

Form a mental image

data could be raw or triaged

Look for correlations and meaning 02

CHECK ANALYZE Relevance

Quality

Patterns

Trends

Aid understanding

03 EXPERT

nd a story in the data

SOFTWARE

THE STORY Analysis method

we trying to communicate?

Ensure source quality

Determine the best way to visualize/communicate

Charting type, visual analogy or information graphic

Clear data integrity

SUCCESSFUL DATA VISUALIZATION TRAITS

Captivating & meaningful

Functional & usable

Aesthetically pleasing

TECH SPARK | 15

Designing a data visualization goes beyond just th the purpose, who, why, where and when they will

DATA VISU HUMAN-CENTERED RE PURPOSE What are you starting from – a user need, a dataset, a request from a client or internal department?

What problem are you trying to solve and how will this visualization help?

What are the key goals you want to address with the visualization?

What is the nature of your intention?

To make a point, support a story or provide exploration?

16 | TECH SPARK

he aesthetic you need to keep it human-centered and think about use the visualization. Below are a few things to consider...

UALIZATION EFLECTION, KEY QUESTIONS AUDIENCE Who is the target user for this visualization?

DATA

CONTEXT

How complex is the data and has it been triaged/cleaned?

Where will the visualization live (online, in reports, printed, an article or blog post)?

Is the data real time or static and how will it be updated?

How will the user utilize the visualization and is it meant to be explored?

What cultural-, domain-, or industry-speciďŹ c needs does your user need to be able to interpret the visualization (if any)?

How will the data be accessible for different user needs?

Is the visualization static or dynamic, passively consumed or interactive?

How will you measure if the visualization has worked and the user is satisďŹ ed?

Is there an obvious series of patterns or stories being told through the data?

What will the user be looking for and what will they do with their insights?

TECH SPARK | 17

A I /M L

I B E

O O T

IS R P

R E T

EN

DATA VISUALIZATION TOOLS IN FINANCE NO

N-

OLS

R TO LOPE

DEVE 18 | TECH SPARK

DE

VE

LO

PE

RT

OO

LS

Historically, banks have relied on IT and third-party consultants for data management, data aggregation and decision support. But, increasingly, financial institutions are turning to data visualization tools to help them aggregate, analyze and gain insight from their data.

According to Qlik, a business intelligence and visualization software provider, data visualization tools are being used by thousands of financial institutions across the globe, including 47 out of the top 50 financial institutions.

In the Big Data @ Work survey conducted by IBM, 71% of the 124 respondents from the financial sector reported that use of big data and analytics (including data visualization) creates a competitive advantage for their organization.

It’s easy to get lost with the unprecedented amounts of data being created, stored and shared. As organizations seek to derive greater insights and present their findings with clarity, the premium placed on the right visualization framework will only continue to increase.

Data visualization is constantly changing and as new technologies and techniques push the boundaries of data, your organization should be flexible enough not only to adapt to these new requirements, but also remain pragmatic in your selection of visualization techniques.

Financial institutions, ranging from commercial banks to asset managers and insurers, use data visualization to address a variety of market needs, such as regulatory compliance, portfolio analysis, benchmarking and model development.

We have identified four groups of tools across the industry:

ENTERPRISE BI TOOLS

DEVELOPER TOOLS

NON-DEVELOPER TOOLS

AI/ML-DRIVEN TOOLS

TECH SPARK | 19

ENTERPRISE BI TOOLS

ENTERPRISE DEVELOPER TOOLS BI TOOLS

Cloud-based BI platform designed to explore and analyze data from multiple sources in real time.

Provides a suite of business analytics tools that allows you to connect multiple data sources, simplify data prep and run sophisticated visualizations.

Hosts a comprehensive portfolio of solutions that provide advanced analytics across the spectrum of BI needs.

Allows users to simply and quickly connect, visualize and share data from a PC to their iPad across a suite of visualization products and services.

Enterprise BI and big data analytics platform that is used to explore, analyze and share real-time business insights.

Turn your data into customized dashboards and reports without coding. It provides prebuilt connectors to several database sources.

20 | TECH SPARK

NON-DEVELOPER DEVELOPER TOOLS TOOLS

NON-DEVELOPER AI/ML-DRIVEN TOOLS TOOLS

D3.js is a JavaScript library based on data manipulation documentation. D3 combines powerful visualization components with datadriven DOM manipulation methods.

HighCharts is a chart library written in pure JavaScript that makes it easy and convenient for users to add interactive charts to web applications. This is the most widely used chart tool on the web, and business users require a commercial license.

Echarts is an enterprise-level chart tool from the data visualization team of Baidu. It’s a pure JavaScript chart library that runs smoothly on PCs and mobile devices, and is compatible with most current browsers.

Leaflet is a JavaScript library of interactive maps for mobile devices. It has all the mapping features that most developers need.

AI/ML-DRIVEN TOOLS

Vega is a set of interactive graphical grammars that define the mapping rules from data to graphic, common interaction grammars and common graphical elements.

Users can freely combine Vega grammars to build a variety of charts.

Deck.gl is a visual-class library based on WebGL for big data analytics. It’s developed by the Uber visualization team.

Chart.js is a tiny open-source library that supports just six chart types: line, bar, radar, polar, pie and doughnut. It’s one of the most popular open-source charting libraries to emerge recently.

ENTERPRISE BI TOOLS

DEVELOPER TOOLS

NON-DEVELOPER TOOLS

AI/ML-DRIVEN ENTERPRISE TOOLS BI TOOLS

Online chart builder that allows you to build basic charts very quickly and pull in data from multiple external sources.

DEVELOPER TOOLS

NON-DEVELOPER TOOLS

Next-generation platform for building data stories.

Simple, online tool for making interactive charts.

Creates leading open-source tools for composing, editing and sharing interactive data visualizations via the web.

Helps simplify the creation of engaging charts, infographics, maps and reports.

Open-source, data-visualization framework that simplifies the visualization of complex datasets.

Combined gallery and infographic generation tool that offers a simple toolset for building data representations.

AI/ML-DRIVEN TOOLS

Smart data analysis and visualization service in the cloud that helps quickly discover patterns and meanings in datasets.

Uses artificial intelligence to fill the gap between dashboards and stories for immediate insights.

Platform for data science teams that unites data prep, ML and predictive-model deployment.

Data science company that allows business scientists of all skill levels to build and deploy accurate ML models quickly.

First BI service with pay-per-session pricing and ML insights for everyone.

TECH SPARK | 21

A Visualization Hub Example LUXTRADE SENDARY ACCOUNT

BALANCE

NET EQUITY

RND TRIPS

CENTURY ACC

£728,066,342.91

£1,034,579.31

£32,979,342.00

£0

£34578.01

22.1%

P&L

NEON ACC

43.62%

£228,894.93

£34578.01 JAN

FEB

MAR

APR

MAY

JUN

JUL

AUG

SEP

OCT

NOV

DEC

£3,542 109

MARGIN

£10,342.950.01

Y

T

LIFFE

30%

40%

ICE

28%

20%

CME

13%

15%

LME

19%

10%

CBOT

10%

10%

PERFORMANCE

ZUIS ACC

69.6%

£34578.01

0.221 EXCHANGES

63.2%

0.219

0.189

0.099

0.121

0.201

0.009

32.1% OTHELLO ACC

CME 15%

£34578.01

61%

62%

41%

26%

34%

52%

14%

JULY 20

AUG 20

SEPT 20

OCT 20

NOV 20

DEC 20

JAN 15

24.3% NEPTUNE ACC £34578.01

ALUMINUM

USD/CAN

COPPER

BONDS

OIL

SUGAR

£3457.02

£5248.55

£1273.02

£7397.01

£8882.02

£9263.01

-104

32

-49

31

101

98

GBP/USD

GBP/EUR

GOLD

STEEL

EQUITIES

CHF/JPY

£2882.02

£2248.55

£1112.02

£7397.01

£8882.02

£3456.01

-67

22

123

-11

21

198

41.2% PLUTO ACC £34578.01

69.7% HARMONY ACC

Business transformations driven by data visualization could include real-time, data quality dashboards, a comprehensive view of risk across the organization, and self-service analytics that cater for business users and decision-makers. Data dashboards allow practitioners to easily pinpoint portfolio outliers and identify potential data-quality issues. Data from multiple systems and sources can be linked via a data-visualization tool which serves as a hub for customer intelligence.

The hub facilitates reporting at the enterprise-level rather than across disconnected lines of business, breaking down silos and producing new business insights.

IT dependency and long service-request queues will be a distant memory, as business users will have direct access to empowering data and analytics for decision-making, root-cause analysis and reporting.

Last but not least, self-service data visualization will free resources to work on more important business goals.

22 | TECH SPARK

£34578.01

20.2%

Data visualization tools can be used for both data discovery and for explaining data to others – an important distinction. Data visualization for exploring can be imprecise. It’s useful when you’re not exactly sure what the data has to tell you, and you’re trying to get a sense of the relationships and patterns contained within it.

TECH SPARK | 23

Data Visualization for Exploration

Managing Audience Expectations

Data visualization for exploring is best approached in such a way that allows it to be iterated quickly and experimented upon, so that you can find the valuable information and disregard the noise.

An important part of this process is determining what is most appealing and enlightening to your audience, by considering the following elements:

Data Visualization for Explaining Data visualization for explaining is best when it’s most transparent.

The ability to pare down the information to its most basic form makes it easier for the decision-maker to understand. This is the approach to take once you understand what the data is telling you, and you want to communicate that to others.

Technology Advances in technology over the past 20-plus years have made this task much easier, more effective and more efficient. We can now incorporate terabytes and petabytes of data, both internal and external to our organizations, at the touch of a button. We can pull in our actuals, in real time, to help make better, smarter, faster decisions to help our business partners and organizations thrive.

24 | TECH SPARK

What information is important to your audience?

How much detail does your audience need?

What actions can be taken with your discoveries?

How do learned insights and foresights affect current actions?

What additional information can be produced to amplify reach?

TECH SPARK | 25

The Three Tiers of Visualizing Data Different users need different views of the same data. We have found that the data hierarchy and visualization style are the main indicators for differentiating these views. When designing financial applications we tend to see three tiers of data and information aimed at these different user groups.

Tier 1 (the big org picture – wide and broad) Personalized high-level executive summary dashboards. Provide a monitoring view often with a RAG status (often aimed at c-level).

Tier 2 (the departmental view) A deeper dive into business streams or organizational areas such as payments, customers, incidents, breaches and KPIs. Can also be temporal/time-based or scenario.

Tier 3 (the individual view – narrow and deep) Drills down to some of the specific data often displayed in data grids and provides insights into how some of the patterns/insights in tiers one and two have been generated.

A classic example of where these tiers are valuable is as follows: A relationship manager might want to know which of the portfolios she covers are at risk of redemption.

Senior management would like to know the firm-wide view of accounts at risk, trends and coverage for those accounts.

A client may want to drill deeper into their portfolio to see which funds are performing better than others to discuss longer-term objectives with their RM. From an individual presentation to a firm-wide risk perspective, the narrative needs to adjust and tilt to meet the needs of the different users.

26 | TECH SPARK

High-level dashboards monitoring often with a RAG status

A deeper dive into business streams or organization areas such as payments, incidents, KPIs or breaches, for example.

Often time-based supported with scenarios

Low-level drill-downs into the raw data, often displayed in pivoting data grids. Provides data insights, source information and details the patterns to back up any decision-making

TECH SPARK | 27

Playing with Patternization for Visual Recall Region | Sector Europe

Asia

AME

Cyclical

Sensitive

Americas

Defensive

Developed Emerging

Low

Value

Core

Growth

Investment Style

Small Medium Large Market Capitalisation

Performance Historical Maximum (100%) Current return (% of HM)

28 | TECH SPARK

High

Patterns are everywhere and even when they’re not obivous, we try and make sense of them. It’s part of our evolution to use pattern recognition to help us survive. Our skill at pattern recognition enables us to capitalize on opportunities and cope with challenges.

Data visualizations rely on pattern recognition.

A chart interprets measurable values as a visual image. In turn, this allows us to recognize patterns for the purposes of making comparisons, identifying outliers, etc. Using recognized patterns can help us identify and remember information that otherwise would be too complex to interpret in one go. A way to do this is to connect datasets to everyday objects we recognize. And once we understand the visual framework, we can decode the information quickly. An interesting example of this was devised by Herman Chernoff in 1973 called Chernoff Faces.

Chernoff Faces display multivariate data in the shape of a human face. The individual parts of the face (eyes, mouth, nose) represent values of the variables by their shape, size, placement and orientation. The idea behind using human faces is that humans easily recognize faces and notice small changes. This helps people to see patterns, groupings and correlations. The concept of the Chernoff Face applied to finance was explored by Julie Rodriguez and Piotr Kaczmarek in their book Visualizing Financial Data. They devised a Chernoff Fish to represent multivariate financial data. A variation on this idea, the Chernoff Eye, is illustrated opposite and shows how changing the data values changes the visual representation of the eye.

The point of Chernoff Faces is to display multiple variables at once by positioning parts of the human face, such as ears, hair, eyes and nose, based on numbers in a dataset. The assumption is that we can read people’s faces easily in real life, so we should be able to recognize small differences when they represent data. It’s debatable how well they work but, debate aside, they’re interesting for exploring data visually.

Long Strategy

Long | Growth | Mid Cap

Long | Value | Small Cap

Long | Core | Large Cap

Long | Value | Large Cap

Short Strategy

Short | Value | Small Cap

Short | Growth | Mid Cap

Short | Core | Mid Cap

Short | Growth | Small Cap

Investment Size

Playing around with the concept of Chernoff Faces using an eye pattern provides some interesting results, but they need further exploration to determine how useful this type of visualization could be.

January 2010

January 2020

Historic Comparison

Exploring Chernoff visualizations using investment data

Long View

Short View

Investment Strategy Comparison

TECH SPARK | 29

Bank’s Data Visualization The Business Problem The monthly management information (MI) reporting format was problematic on several levels. Basic issues included: • Time-consuming to prepare • Difficult to read and compare data • Trend analysis was not easy • Multiple visualization styles used in reports

01.

Evaluate the Need to Change Study current situation: • Understand reports and usage • Understand how results are reported monthly • Learn how reports are currently generated – from which groups and products Identify key issues: • Hard to compare monthly data • Information overload • Data does not relate to other data • Difficult to spot red flags • No temporal visibility across the business • High-level views require a lot of manual work • No clear way to interrogate data • No way to tie incidents like breaches to overall data Determine scope of PoC: • Establish reporting categories across products Format: • Determine required views

DXC Luxoft MI Solution • Clarity and transparency Eliminated manual • Better decision-making analysis and provided tailored • Optimized reporting workflow reporting • Predictive trend analysis • Tailored reporting • Visibility across the business • Consistent visualization style and usage

Apply Correct Solution The solution provided: • Clarity and transparency • Better decision-making • Optimized reporting workflow • Predictive and trend analysis • Tailored reporting • Visibility across the business

02.

30 | TECH SPARK

03.

Our Uniqu

• A digital a which ena through a reports fo built red-fl for ease o potential

• A showca with the b data visua stakehold organizati

ue Customer Focus Created:

and fully clickable solution ables the bank to navigate an interface to build custom or their monthly meetings. We flag highlighting into the system of use, and the rapid spotting of problems and risk areas

ase video of the data journey bank, allowing them to share alization internally with ders and key members of their ion

Achievement • Eliminated cumbersome manual analysis and provided tailored MI reporting that met the client’s needs

04.

Leading Retail Bank

Success Story – Enterprise Visualization

Visualizing Risk Data The monthly management information (MI) reporting format was problematic on a number of levels: • Time-consuming to prepare • Difficult to read and compare data 12 Week Discovery • Takes time to do any trend analysis Segment: Investment Professionals and Reporting • Multiple styles and types of visualizations Managers

Challenge To improve a leading retail bank’s monthly MI reporting format. The client challenged DXC Luxoft to cut preparation time, improve readability and data comparison, speed up trend analysis and reduce the number of styles and types of visualizations.

Solution Our clickable prototype enabled the client to navigate through the interface easily, building custom reports for their monthly meetings. We built-in red-flag highlighting for ease of use and the rapid spotting of potential problems and high-risk areas. The accompanying showcase video of all the key journeys allowed our client to share progress with their internal stakeholders and other key members of the business. Core technologies: User-Centered Design (UCD), Agile development, Sketch and Zeplin (design), and Justinmind (prototyping).

Results Enabled the visual representation of available data, creating tailored reporting that provides: • Clarity and transparency • Better decision-making • Optimized reporting workflow • Predictive trend analysis • Visibility across the business • Consistent visualization style and usage

TECH SPARK | 31

A Data Scientist Can’t Be Replaced with Software A common joke is, “Data scientists spend 80% of their time collecting and cleaning data. They spend the remaining 20% complaining about collecting and cleaning data”. But what do they actually do and why there are no tools to replace their skills?

32 | TECH SPARK

Derive a model with examples that optim parameters to the ov error is minimized.

Variables need to be in a similar cohort (think oranges and apples). Mixing variables can confuse and can lead to people reaching inaccurate conclusions.

Raw variables are often not good enough to be thrown at an ML algorithm on their own.

They need some intepretation and variable engineering requires domain expertise.

No need to be an expert at coding, a lot of data-prep packages exist to make the experience as smooth as possible. With intelligent features, they can even suggest ideas.

solid related mise the models verall prediction

Identifying missing or corrupted data and resolving where possible. Looking at how to handle outliers and anomolies.

Decide on which variables to use or discard. Identify other datasets where variables could be utilized.

The part where data is visualized and analyzed. Tools used are code-based (R/Python), or can be a data visualization tool such as Tableau, Qlik or PowerBI. During this step the intuition and business sense of the data scientist is particularly important.

Contributors

Nathan Snyder Head of BCM Consulting, EMEA Nathan Snyder is head of Banking & Capital Markets consulting for Luxoft EMEA. He focuses on helping Financial Services organizations capitalize on business opportunities through technology and operational change. Nathan has worked with tier one-three banks in London and New York.

34 | TECH SPARK

Luis Cameroon Principal Consultant

Phil Cranfield Director

Andy Hall Creative Director

Paul Hewitt Senior Director

Hiren Patel CX/UX Director

Luis has been designing and developing high-performing, scalable and innovative webbased applications for clients in the Financial, Energy and Media industries for more than 15 years. He’s fascinated by human-computer interaction and is a keen proponent of userexperience design.

Phil Cranfield has over 20 years’ experience of leading IT projects for clients in the Banking & Capital Markets, Insurance, Utilities and Retail industries.

Phil has worked at Luxoft for 6 years, structuring a wide range of client engagements to deliver complex IT programs in line with strategic goals and initiatives.

Andy Hall is Luxoft Excelian’s creative director. A large proportion of his 25 years’ usercentered design experience has involved helping organizations apply user experience, information design and other creative techniques to drive design innovation. Andy specializes in solving complex Financial Services problems.

Paul Hewitt leads the Data, Analytics and AI/ML practice for Financial Services. His global responsibilities include handling full-scale AI and ML projects in conjunction with the various worldwide delivery teams, as well as managing R&D on behalf of clients, and providing proofs of concept.

Hiren Patel has worked closely with clients in the Banking & Capital Markets (BCM), Telecommunications, Insurance and Technology industries for over 14 years. In his current role as Customer & User Experience Director, he leads the digital experience design services for our BCM consulting practice.

A long-term advocate of user-centered design, Hiren developed our LeanUX framework, enabling clients to visualize and deliver tangible solutions, faster.

The goal of data visualization is to turn data into information and information into insight.

Carly Fiorina, CEO Hewlett Packard, 2004

Digital.

www.luxoft.com

Experience.

Design.