Agent-based technology supports service provision at BT
Our events fully explore the disciplines of operational research, analytics and data science. We also run specialist events that focus on simulation, artificial intelligence, defence and analytics, and more. Our events are open to all, with members receiving preferential rates.
The OR Society puts on a full calendar of events and workshops to help you further your career and:
⚫ Connect and network
⚫ Share knowledge and research
⚫ Expand your skillset
⚫ Find new ways of tackling real-world problems
Become part of our community of excellence: www.theorsociety.com/events
EDITORIAL
Welcome! Thank goodness this inaugural issue of a new course for Impact magazine is finally reaching you, after a year that saw the formation of a brand-new editorial team, some restructuring of what’s on offer, and some unforeseen circumstances.
In this issue we present several experiences in the practice of OR, pointing straight at the impact they made to industrial practice (planning of telecom networks; credit scoring and financial education), policy-making (antimicrobial resistance in US agriculture), and several communities (NHS patients in the South-West of England; young adult school leavers aged 16–18 years; South Africa’s population at greater risk of contracting HIV).
Each article looks at relevant aspects of method or technique (agent-based simulation applied to telecom network planning), methodology (systems methodologies in support of policy-making at the US Department of Agriculture; action research and its role in managing interventions as learning journeys), the development and roll out of novel technology (the award-winning NHSquicker app), or the impactful use of established tech (Simul8’s pro bono work in South Africa).
Our hope is that OR practitioners, existing and prospective, may pick up and use whatever ideas put forward in our articles that may apply to them and their organisations. We offer several pieces of pragmatic advice, such as the entire article on interventions as learning journeys, and the methodological reflections closing the article on antimicrobial resistance. The piece on credit risk and financial education suggests ways to foster impactful collaboration between academics and practitioners that will benefit both sides. Our column by Geoff Royston reminds us all of what unseen data can tell us. Finally, Matthew Howells’ light-hearted introduction to the Travelling Swiftie Problem is a semi-serious contribution to fostering the impact of Swiftonomics on the UK capital city. More seriously, the article provides practical tips on how to model and solve one of the most iconic problems in combinatorial optimisation (if you are new to OR, you need to read this!).
Maurizio Tomasella
OPERATIONAL RESEARCH AND DECISION ANALYTICS
The OR Society is the trading name of the Operational Research Society, which is a registered charity and a company limited by guarantee.
Seymour House, 12 Edward Street, Birmingham, B1 2RX, UK
Published by Taylor & Francis, an Informa business
All Taylor and Francis Group journals are printed on paper from renewable sources by accredited partners.
Operational Research (OR) is the discipline of applying appropriate analytical methods to help those who run organisations make better decisions. It’s a ‘real world’ discipline with a focus on improving the complex systems and processes that underpin everyone’s daily life – OR is an improvement science. For over 70 years, OR has focussed on supporting decision making in a wide range of organisations. It is a major contributor to the development of decision analytics, which has come to prominence because of the availability of big data. Work under the OR label continues, though some prefer names such as business analysis, decision analysis, analytics or management science. Whatever the name, OR analysts seek to work in partnership with managers and decision makers to achieve desirable outcomes that are informed and evidence-based. As the world has become more complex, problems tougher to solve using gut-feel alone, and computers become increasingly powerful, OR continues to develop new techniques to guide decision-making. The methods used are typically quantitative, tempered with problem structuring methods to resolve problems that have multiple stakeholders and conflicting objectives. Impact aims to encourage further use of OR by demonstrating the value of these techniques in every kind of organisation –large and small, private and public, for-profit and not-for-profit. To find out more about how decision analytics could help your organisation make more informed decisions see https://www.theorsociety.com/about-or/or-in-business/. The OR Society is the home to the science + art of problem solving.
5
BT: SIMULATING HOUSEHOLD BANDWIDTH USAGE
In presenting their Cassandra simulation model, Lucy Gullon, John Bicknell and Mathias Kern show how BT Research and Network Strategy (RaNS) aid decision making around future network planning
9
MANAGING INTERVENTIONS AS LEARNING JOURNEYS
Jim Scholes offers observations derived across several consultancy interventions dealing with strategy and change in large organisations, looking at them as learning journeys for the client 14
NHSQUICKER: IMPROVING URGENT AND EMERGENCY CARE DECISIONS IN THE SOUTH-WEST
Nav Mustafee, Alison Harper, Tom Monks and Rachael Shine present their award-winning solution to support patients’ decision making in a way to relieve pressure from busy A&E departments and other centres for urgent care
20
MAKING THE DIFFERENCE IN CREDIT RISK AND FINANCIAL EDUCATION – LESSONS FOR FUTURE REF CASE STUDIES
Jonathan Crook, Tina Harrison, Galina Andreeva and Arseniy Morgunov describe advancements made to credit risk assessment through the work conducted at the Credit Research Centre of the University of Edinburgh Business School
3 Seen Elsewhere
Analytics making an impact
35 fOR everyone
Introducing and solving the Travelling Swiftie Problem, Matthew Howells helps die-hard members of the most popular fandom of the last fifteen years to find their way through the capital city of the UK
38 Dark Matters
Geoff Royston discusses data we do not see, why it matters, and what it can tell us, based on Dark Data: Why What You Don’t Know Matters by David Hand
24
29
SIMULATION FOR SOCIAL GOOD – COMBATTING SOUTH AFRICA’S HIV EPIDEMIC
Laura Reid presents the successful collaboration between Simul8 and Shout-It-Now, a Non-Profit Organisation providing HIV prevention, sexual and reproductive health advice, and genderbased violence support services to communities in South Africa
KEEPING ANTIBIOTICS EFFECTIVE – COMBATTING ANTIMICROBIAL RESISTANCE IN U.S. AGRICULTURE
Gerald Midgley, Amber Elkins, Guy H Loneragan, Megan Babowicz, Mayukh Dass, Yrjo T Grohn, Ellen Jordan, Guillaume Lhermie, Lucas Lunt, William A McIntosh, Juan M Piñeiro, Jason Sawyer and H Morgan Scott discuss how they worked with stakeholders such as food-animal producers, veterinary services, pharmaceutical companies, public health policymakers and consumer advocates, and more, to examine the potential for improving voluntary stewardship of antimicrobials in US agriculture.
DISCLAIMER
The Operational Research Society and our publisher Informa UK Limited, trading as Taylor & Francis Group, make every effort to ensure the accuracy of all the information (the “Content”) contained in our publications. However, the Operational Research Society and our publisher Informa UK Limited, trading as Taylor & Francis Group, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by the Operational Research Society or our publisher Informa UK Limited, trading as Taylor & Francis Group. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. The Operational Research Society and our publisher Informa UK Limited, trading as Taylor & Francis Group, shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
Reusing Articles in this Magazine
All content is published under a Creative Commons Attribution-NonCommercial-NoDerivatives License which permits noncommercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way.
SEEN ELSEWHERE
SUCCESS AND FAILURE, REVISITED
In our Issue 17 (Spring 2023), we reported on the first in a series of ten articles by Douglas A. Gray in INFORMS’ Analytics magazine (https:// bit.ly/Gray2023). The article discusses successful Data Science projects and the reasons why unsuccessful ones fail. In the ensuing months, the author completed the series, delving into relevant aspects such as the misapplication of (many) models, the key need to move from ‘sandbox models’ (your own desktop or work area in the cloud) to fully-fledged ‘production systems’ (microservice, standalone, etc.), solving problems that are not a business priority, and more. Its tenth and final instalment offers a concise summary that looks at all the aspects discussed in the series, through the lens of communication skills. Quoting verbatim: ‘Human involvement in data science is substantial in every step of the process and materially significant, requiring the development and application of many “soft skills” that necessarily facilitate successful execution and completion of data science projects’. The author then argues that communication is ‘… by far the most critical and foundational [of all soft skills] to successfully execute data science projects’. He breaks communication skills further down into: listening (to understand); being heard (to be understood); ‘speaking and writing concisely, impactfully and with clarity for the target audience’; and ‘gaining a deep level of mutual understanding in all aspects of a project’. He then puts all this in the context of data science projects and the individuals involved in running them, focusing on ‘project challenges’ and the dynamics of ‘managing change’,
concluding that ‘[in data science projects] communication alone is half the battle’. Finally, the author observes: ‘… empathy is the most important quality to embody regardless of how incredibly difficult things get along the journey. And trust me, as worthwhile as data science projects are in every respect, things will get difficult at many, many points along the way, and you cannot afford to alienate any constituents, partners, teammates or stakeholders. People never forget heroes, and they never forget jerks. You may (barely) get through one project, but you will never get through another one by treating anyone who matters with anything less than The Golden Rule’. How could anyone possibly disagree?
AUTONOMOUS WEAPONS SYSTEMS
Twenty-first-century warfare has gotten very serious, in some sense more than ever before. A key new feature in this respect is the use of AI in what have come to be known as Lethal Autonomous Weapons (LAWs) or equivalently ‘Autonomous Weapon Systems’ (AWSs). Not at all surprisingly (we are talking about ‘military operations’ and ‘systems’, after all), OR has been contributing to the debate, providing recent impactful contributions to its reshaping. Dr Stephen Harwood of the Department of
Space and Climate Physics at University College London argues in a recent article (https://bit.ly/Harwood24) that AWSs are a poorly understood form of emerging technology. His reasoning develops along three main lines: the systems aspects of AWSs, including a recursive perspective that will be familiar to many of us; their autonomy as self-regulating systems; and the values underpinning their functioning, as they manifest in legal and moral considerations about their use. Dr Harwood’s work was picked up by The ‘AI in Weapons Systems Committee’ of the House of Lords, who in a recent report (https://bit.ly/HoL2023) carrying the emblematic title ‘Proceed with caution: Artificial Intelligence in Weapon Systems’ quotes verbatim part of Dr Harwood’s contribution to the discussions: ‘The “programmer” who designs the AWS is distanced by the unpredictable learning capability of the system, with the analogy of whether parents can be held responsible for the actions of their children once they mature. Likewise, it is questionable whether it is fair that the commanding officer who ordered the deployment is held accountable for a technology that can choose its own target’. Both the systematic nature of AWSs and the accountability aspect in relation to their deployment are recognised in the report as particularly central to the global and ongoing debate on this dividing technology. The committee recommends (paragraph 189) that ‘The Government must commit to integrating meaningful human control in all AI-enabled AWS[s] which it deploys so that human accountability can clearly be assigned for use of AWS[s] on the battlefield’.
Photo by Jason Goodman via Unsplash
BT: SIMULATING HOUSEHOLD BANDWIDTH USAGE
LUCY GULLON, JOHN BICKNELL AND MATHIAS KERN
BT Group are the UK’s largest provider of fixed-line broadband and mobile services. We have operations in around 180 countries across the world and are committed to connecting people. To meet this goal, we must ensure our networks have the capacity to sustain future demand. This is especially true of the highly competitive consumer broadband market; understanding where and when to invest in network expansion and development is essential to balancing cost and performance to meet the evolving needs of our customers.
The demand for high-speed broadband has grown rapidly worldwide over the last few decades, not least thanks to the invention of new technologies requiring reliable, high speed Internet access [1]. For example, consumers have invested heavily in large, high-quality TVs over the last 4 years [2] leading to fierce competition amongst VOD (video on demand) providers to offer increasingly high quality, high bandwidth streaming services. Similarly, the gaming industry has enjoyed year-on-year growth for over a decade [3] with users demanding
not only highly responsive, low latency ‘in-game’ play, but also short ‘download-and-play’ wait times. With games downloads now routinely exceeding one hundred gigabytes [4] a significant games release has the potential to cause high spikes in demand well above weekly norms. Furthermore, new and emerging technologies such as AR/VR (augmented and virtual reality) and cloud gaming are just beginning to gain traction [5] but the impact of their take-up in the UK is, at least for now, largely unknown.
To be able to assess and analyse the impact of these external bandwidth demand drivers, we look to the operational research methods of forecasting and simulation to aid our future planning and decision making.
WHAT QUESTION DO WE NEED TO ANSWER?
BT’s fixed-line broadband network provides Internet connectivity to household and business premises across the UK. As it exists today, each household has a router that connects to the wider network via fibre and/or copper access. The choice and availability of broadband package determines the upload and download speeds available for each household. The capacity of the network is carefully planned to ensure we can deliver the line speed offered to each customer.
we look to the operational research methods of forecasting and simulation to aid our future planning and decision making
This planning process is complex as we not only need to plan for typical usage but also for one-off special events
that cause peaks in demand. As well as the aforementioned large games downloads, this includes large audiences streaming exceptional live events such as a royal wedding or (perhaps again one day) England in the football World Cup final. Dramatic changes to weekly norms can also take place. For example, during the Covid-19 pandemic a significant increase in home working caused prolonged changes to weekly consumption patterns that have, to an extent, persisted beyond the end of the pandemic.
This planning process is complex as we not only need to plan for typical usage but also for one-off special events that cause peaks in demand
The fundamental question for BT is whether our fixed line network is robust enough to provide a good level
of service to our customers based on how consumer demand may change in the future. To try and answer this question we use simulation as a tool to test evidence-based future scenarios that incorporate likely and more extreme changes from current broadband usage patterns. Simulation is therefore an exceptionally powerful tool for network, products and services planning within BT.
THE CASSANDRA MODEL
The Cassandra model is an agent-based simulator which, at a high level, represents the utilisation of BT’s fixed-line broadband network. We build up our model from individual people and device agents to household agents using information gathered on typical household demographics. Each person agent can undertake up to two different activities (watching, gaming, communicating etc.) during any minute of the week. This means inferring the probability that a member of a household
FIGURE 1 THE AGENT-BASED ARCHITECTURE
starts an activity in each minute. Each person or device agent sends bandwidth requests to the household agent, which may be fully or partially satisfied depending on the broadband package of the household agent. A simple representation of the agent-based architecture is illustrated in Figure 1.
The inputs are estimated from several sources including the Office for National Statistics (ONS) for trends on households, the Office of Communications (Ofcom) for trends in Internet usage, Barb Audiences for TV viewing data, as well as internal and external views from domain experts.
We collect activity and bandwidth data for each person, device and household agent for each one-minute window across the 7-day week (i.e. 10,080 minutes in total) and compute overall traffic, peak traffic and whether an agent’s traffic is capped. This is combined to provide a minute-by-minute weekly profile of bandwidth usage for each agent. Using this information, a picture of typical weekly bandwidth usage across the UK can be formed by combining the output of households consisting of many different household types (such as a single adult, married couple, couple with children etc.). Clearly, bandwidth consumption is subject to strong daily patterns, which is reflected in higher consumption during leisure hours (∼6-10pm) and lower consumption during the night when most people are asleep. Observed bandwidth consumption profiles from actual households are used to verify our model.
Each year our model goes through a formal validation process and, if required, it is finely tuned through re-calibration to ensure the baseline output is an accurate representation of the current pattern of weekly bandwidth consumption. Our aim is to have close agreement between the
minute-by-minute average household behaviour we see as a model output and true average core network behaviour. This validation procedure ensures that unlike the Greek goddess Cassandra (whose prophecies were true but never believed) our model is starting from a credible baseline that is shared and agreed with stakeholders before we consider running any experiments.
FORECASTING THE FUTURE
Unfortunately, unlike the goddess Cassandra we do not possess the ability to perfectly predict the future all of the time, but using simulation we have the power to investigate many possible futures. This is commonly referred to as “What-If?” scenario modelling. There is often significant uncertainty when a new product comes to market. Unknowns such as year-on-year product take-up, and how and when people will incorporate new technology and services in their daily routines can all be used as inputs to the simulation model and thus adjusted and analysed
to investigate their individual and collective impact.
If we consider, for example, a scenario in which AR/VR achieves mass-market success, in the Cassandra model this would be treated as a new activity which any person in any household has some probability of undertaking at any given time. As is usually the case, we cannot say with certainty how popular these products will become when forecasting for the next 10 years, or even 5 years. Forecasts from product manufacturers, network equipment vendors and an enthusiastic technical press are often optimistic and focused on pre-conceived growth
targets. We may therefore consider a range of plausible take-up scenarios within our experiments using related empirical data as a starting point. We must also go further and try to predict likely data rates of any new activity, whether they will be additive or substitutional with respect to existing activities (such as TV viewing and gaming) and when during the day and week we expect them to take place. A new and popular activity that requires a high-speed Internet connection, takes place in peak hours (∼8-10pm) and is carried out in addition to existing activities, has the maximal potential to drive up overall network peaks and it is these peaks which are so important to network capacity planning.
Within our modelling we often construct extreme scenarios such as these to consider the very limit of what we’d expect to happen. This is useful when it comes to activities such as planning the future of the broadband network as we aim for our services to be as resilient as possible to all eventualities. Worst-case scenario modelling can also alleviate some of the uncertainties in our modelling. It allows us to consider how well our network reacts to extreme circumstances. If we build our network to be resilient to extreme bandwidth consumption behaviour, then we can lower the probability of problems arising under normal conditions.
SUMMARY
Within BT RaNS (Research and Network Strategy) we use simulation modelling to aid decision making around future network planning. The Cassandra model is key to our understanding of how bandwidth consumption will evolve over the years to come. It allows us to take forecasted behaviours for different bandwidth consuming activities and consider the impact on overall household bandwidth demand.
In future, we plan to evolve the Cassandra model to take into account strategies for reducing energy consumption. This involves an important trade-off between controlling network capacity and quality of service. Whilst endlessly increasing capacity may be the simplest (and most costly) way of meeting demand for extreme and rare events, on a typical day this results in the continuous operation of network equipment that consumes vast amounts of energy. The Cassandra model could be
REFERENCES
key to investigating these conflicting objectives, helping BT to deliver sustainable and commercially viable network solutions for the future; as such, the importance of Operational Research in this context cannot be overstated.
Dr. Lucy Gullon is an AI and Modelling Research Specialist in BT’s Research and Networks Strategy team and a Visiting Researcher at Lancaster University. She received her PhD in Statistics and Operational Research from Lancaster University in 2018 for which she was awarded the OR Society’s Doctoral Prize. Her expertise lies in simulation uncertainty quantification.
John Bicknell is a chartered engineer specialising in the quality of experience of networks and applications. He joined BT in 2004 and is a Research Manager in the real-time content and communication team. As a SME in UK interoperability standards, he is a strong believer in the power of cross-industry collaboration for improving network resilience and reliability.
Dr. Mathias Kern received his MSc and PhD in Computer Science from the University of Essex in 1998 and 2006, respectively. He is Senior Research Manager for sustainable operations and optimisation in BT’s Research and Networks Strategy team, and is an experienced industrial researcher and strong advocate for both Artificial Intelligence and Operational Research technologies.
This article offers reflections and lessons derived across several consultancy interventions dealing with strategy and change in large organisations. Though the context and examples are based on strategy work I hope and believe that the observations should be more generally useful to practitioners working with client organisations on many aspects of change whether they be operational or strategic in nature.
STRATEGY THAT IS BEYOND AN ORGANISATION’S CAPABILITY TO CONCEIVE WILL BE BEYOND ITS CAPABILITY TO IMPLEMENT I once interviewed 12 European CEOs about their personal experiences of successfully leading their organisations through strategic change. One, the Dutch CEO of a manufacturing company, summed up a recurring theme
Photo by Dariusz Sankowski on Unsplash
thus: ‘Failure in execution is the main source of failure in strategy. It is the most difficult and the most important. Nine out of ten strategic plans that fail, fail in execution’. Unsurprisingly this a commonly held view amongst senior executives. More surprising is that managers just a couple of layers down in the same organisation often tell a consultant that their problem is absence of strategy: ‘We just don’t have a clear direction’ is a typical comment. This isn’t simply lack of internal communications: such apparently contradictory perspectives are commonplace in organisations where the development of strategy is entirely separate from its implementation. When strategy effectively comes from outside of the mainstream organisation – often from consultants and sometimes from inside the heads of one or two smart senior people in HQ – implementation is almost bound to be problematic. If a strategy is beyond an organisation’s capability to conceive, why should anyone believe it is within its capability to implement?
It can be enlightening to compare espoused strategy contained in the CEO’s presentation deck with the observable day-to-day decisions and actions being taken by managers lower down the organisation. These day-today actions often provide the clearest indication of where a company is actually heading for: effectively these represent de facto strategy. When espoused strategy doesn’t translate to real world action it can’t really count as strategy. Sometimes the gap between espoused and de facto indicates a deeper dysfunctionality in culture and ways of working but more often we find that both the senior group and people lower down the organisation really do want their actions to be purposeful and make a difference.
The disconnect frequently arises in companies that are run simply on the basis of numeric targets as a substitute for a priori strategic thinking. For such organisations, which are sometimes reduced to a ‘core’ set of activities, there may be few options available when their environment changes: it becomes unclear what to do next. In a world of constant flux ‘strategy’ can never be truly complete. In practice organisations are faced with the challenge of managing their ongoing development and implementation of strategy as a continuous activity. This doesn’t sit well with the traditional annual rituals of strategy, planning and budgeting. The world doesn’t stand still waiting patiently whilst companies develop strategy and allocate resources at the same time of year, every year. This is glaringly obvious in high-growth, technology-based industries, but similar dilemmas now face executives in all sectors as they deal with increasing external turbulence and uncertainty. In today’s world there’s need to shift attention from ‘strategy’ (the noun) towards ‘strategising’ (the verb).
When espoused strategy doesn’t translate to real world action it can’t really count as strategy
Strategising can be described as a process of dialogue dealing with both the ongoing development of strategy and building of organisational capabilities. This is at the core of managing an intervention as a learning journey, see Figure 1. In strategy work, there is a risk of becoming trapped in history – the future is seen as an extrapolation of our past experience. The challenge of transformation is one of strategising for a different future –stretching ambition and building new capabilities
LEARNING JOURNEYS ARE DESIGNED AND MANAGED AROUND THE QUESTIONS TO BE ADDRESSED AND THE CLIENT’S STARTING CAPABILITIES
Implicit in the model of simultaneously working on strategy and organisational capabilities is that such work must be done within the organisation rather than outsourced to consultants who then deliver recommendations or write a strategy document. In a learning journey the consultant’s role is to guide and help the client do the work of strategising rather than do it to them or for them. In practice this means that the consultancy team needs to deploy both hard and soft skills.
The design of each intervention varies based on the particular findings of an initial diagnostic exercise. Choices such as intervention modalities, tools, data requirements, expert inputs, project phases, deliverables, resource and time constraints must be weighed against the practicalities of making progress towards desired outcomes when faced with the realities of the client’s own starting capabilities, resources and timeframe. Some of the interventions I have been part of involved hundreds of people in the client organisation, working globally or regionally in teams on specified tasks to gather data and insights relating to questions that needed to be addressed, and feeding their findings into a direct dialogue with their top management. At the other end of the spectrum were single country-based projects with small internal teams providing new insights and perspectives for a series of top management ‘strategic dialogues’.
“We’ve
created something very important – a process through which we can renew ourselves”
Two quite different examples are offered to illustrate how these variations can play out in practice:
1. Faced with increased margin pressure and faltering volume growth in a well-known European
branded business, a large FMCG company set up an initiative which brought together over 60 managers from HQ functions and country management across Europe in a process of strategy ‘co-creation’. For just over four months the managers worked together through a series of facilitated workshops to develop shared understanding of the challenges facing the business, identify growth options, agree
strategic direction and then implement country-specific plans. They identified significant growth opportunities that helped them re-shape the business and in doing so created a blueprint for a new global business model that was then shared across all regions: this blueprint was effectively a bonus that emerged from what they had learned and a realisation that their findings would be transferable.
2. By way of contrast, a large European headquartered global technology company was reshaping itself to find new growth following a difficult but successful turnaround. Constraints were such that the intervention started initially with just 12 fulltime, mostly junior, managers from across the company. From small beginnings a ‘Strategy Project’ grew involving hundreds of people, ranging from top management to newcomers who had only been at the company a few weeks. Collectively they built a company-wide perspective on a wide range of strategic options and ideas on how they could shape industry and technology changes to their advantage – starting as an underdog. Neither the options nor the required actions were clear at the outset. These emerged through the process itself. This unconventional approach and their unique perspective helped the company overtake incumbents, becoming a leader in its industry over the next decade. Later interviewed by a leading business journal, the CEO said: ‘We’ve created something very important – a process through which we can renew ourselves’. He credited the company’s progress to the journey they had been on and the lessons learned.
FIGURE 1 FROM ‘STRATEGY’ (A) TO ‘STRATEGISING’ (B).
IT’S EASIER FOR TOP MANAGEMENT TO PREVENT CHANGE THAN TO MAKE IT HAPPEN
During a particularly challenging intervention a colleague remarked that ‘The stopper is always at the top of the bottle!’ It was an astute observation. Top management usually commissions work on change but can, and sometimes does, block the possibility of change happening. This may be inadvertent, for example senior people often underestimate the extent to which their participation, questions and comments are reinterpreted and second guessed by more junior people of a nervous disposition. But it is also true that this kind of difficulty can arise from top management’s own nervousness about the consequences of major change for themselves. It is easier for senior executives to be comfortable with the rational logic of what needs to be done, but problematic when faced with the consequent hows of implementation
– especially when hows necessitate changes in prevailing leadership behaviours. Transformational change is always accompanied by uncertainty that goes beyond the current routine and this creates adaptive challenges for all - especially top management. The practitioner’s role is to design and orchestrate this process in a way that supports learning across the whole set of participants – challenging (senior) and protecting (junior) individuals from time to time.
The learning journey provides a somewhat de-risked opportunity for senior and junior people to work in new ways during and beyond the project intervention. The fluidity needed to work on strategic challenges in cross-organisational teams cutting across both structure and hierarchy provides a test bed for developing new ways of working and experiencing values and behaviours likely to be needed in future. This can also help the client transition from a project managed by consultants to the organisation taking ownership and embedding change beyond the intervention. For example, some of my clients created ‘temporary structures’ as a bridge between the project and the internalisation of an ongoing process. Temporary structures were not bound by existing structures and hierarchies and, like the intervention, they cut across boundaries. They were typically sponsored by a senior executive who was actively involved, drew resources from across the company and reported progress directly to the Executive Team. Sometimes they focused on building new capabilities, sometimes they were outward looking and focused on new growth initiatives. Occasionally the
temporary became permanent, but often they folded when they had run their course. The most important point was that transition arrangements were in place to ensure progress was sustained beyond a one-off project.
“People hold dear the truths they discover for themselves”
CONCLUSION
At the core of designing and managing a learning journey is a recognition that people don’t always learn simply based on what others tell them. The Socratic notion of letting the questions be the curriculum (or in this case defining the learning journey) is a useful shorthand description. Socrates is also credited with having said that ‘I cannot teach anybody anything. I can only make them think’. Fast forward to the present day, nearing completion of a company-wide change initiative in a large US based global computer services company, a senior member of their Leadership Team commented that ‘People hold dear the truths they discover for themselves’. Whilst his reflection was rooted in his own personal experience of working on the project, he was also acknowledging the importance of the learning opportunity provided for all 150 participants closely involved in doing the work and a much wider group of around 2000 employees around the world who had contributed.
This appreciation that people hold dear the truths they discover for themselves also applies to reflective practitioners ! The lessons shared here are meaningful for me and I hope they are at least thought provoking for
Photo by Julentto
Photography on Unsplash
the reader. They derive from an Action Research framework developed to support practice development in a consulting firm (Figure 2): an outline for project design and delivery in three main phases (1); accounts from each project add to practice development input (2); reflections on project experience and lessons add to an accumulating body of knowledge (3); as practice develops so the design and delivery of projects evolve (4).
Dr Jim Scholes has been a Visiting and Honorary Professor at Lancaster University Management School since 1998. Since 1987, he has been an experienced practitioner of strategy and systems thinking, having worked with leaders in public and private sector organisations in Europe, the US and Asia. Prior to that, Jim worked in a variety of management roles in UK Government and the private sector. He co-authored (with Professor Peter Checkland) the book ‘Soft Systems Methodology in Action’.
FIGURE 2 PRACTICE DEVELOPMENT IS ROOTED IN ACTION RESEARCH.
IMPROVING URGENT AND EMERGENCY CARE DECISIONS IN THE SOUTH
WEST
NAV MUSTAFEE , ALISON HARPER , TOM MONKS AND RACHAEL SHINE
Accident and Emergency (A&E) departments provide urgent and emergency care services for serious and life-threatening conditions, while other services such as Urgent Treatment Centres, Minor Injury Units (MIUs), Walk-in Centres deal with many of the more common minor injuries and minor illnesses for which people may attend A&E.
A “four-hour standard”, set in 2010, states that at least 95% of attendances are admitted, transferred, or discharged within four hours of their arrival at any type of urgent and emergency department. By 2016, after several years of stability, performance against NHS A&E waiting times had been showing a sustained decline. In December 2022, an intermediary
threshold target of 76% was introduced, to be met by March 2024, with further improvements expected in 2024/25. The Keogh Review [1] of Urgent and Emergency Care stated that patients with urgent but non-lifethreatening needs should be treated outside of hospitals, with care delivered in or as close to peoples’ homes as possible. It was important to spread patient demand amongst regional facilities to reduce waiting times.
High volumes of patients attending A&E can lead to over-crowding, rising pressure on A&E services and poorer experience for patients. The NHSquicker platform was developed in response to these issues, to help patients in need of non-life-threatening urgent care to make an informed decision about suitable alternatives available close by, with shorter wait times than their local A&E.
High volumes of patients attending A&E can lead to over-crowding, rising pressure on A&E services and poorer experience for patients
ADDRESSING THE PROBLEM
A team of researchers from the University of Exeter Business School collaborated with hospital colleagues at South Devon and Torbay NHS Foundation Trust to model the performance at the A&E department at Torbay Hospital. The modelling also considered MIUs and Urgent Care Centres (UCCs in South Devon. They found that whilst all MIUs/UCCs met the four-hour standard, A&E was underperforming by 20% [2].
The University team investigated how existing data, already captured by the NHS, could be used to relieve pressure
on A&E departments. In 2016, they brought together a group of experts that included databases and business intelligence professionals from the NHS, operational researchers, clinicians and managers from NHS Trusts in the South West of England – and founded the Health and Care IMPACT Network to work on a solution.
Together they agreed on a format for waiting time data from A&E departments, MIUs/UCCs and other centres for urgent care. A common data standard was necessary as the objective was to develop a digital platform at a regional level, rather than a Trust-specific solution, ensuring that the platform could receive data feeds from multiple urgent care patient-flow systems.
The research team developed a digital solution to reduce pressure on busy A&E departments by giving patients the ability to make real-time informed choices on the best place to go for urgent care. Using system integration, the team developed a digital platform and the NHSquicker [3] app to provide live waiting times for A&E departments and other centres of urgent care. By transforming real-time data into actionable insights and nudges, NHSquicker is designed to encourage patients to choose the most appropriate treatment facility for their condition, alongside the factor of waiting times, so that only those with the most serious conditions present at A&E and those patients can be seen quicker. This reduces demand and waiting times at A&E, and shapes demand across urgent-care facilities by encouraging patients to choose a destination with a lower waiting time. The free app, and the website, has been operational since 2017 and empowers patients to make more informed decisions about suitable alternatives to
A&E for non-life-threatening urgent care needs. This not only reduces the wait experienced by patients but also supports the national NHS four-hour standard agenda.
IMPACT OF NHSQUICKER
Launched in 2017 across Devon and Cornwall, we have since expanded the reach of the real-time platform. Version 3.0 is the latest release of NHSquicker. As of April 2024, NHSquicker receives live data from 37 centres of urgent care, including seven A&E departments across the South West of England. The app is available to 1.9 million people in Cornwall, Devon and Somerset.
IMAGE 1. NHSQUICKER COMBINES REAL-TIME WAITING TIME AND JOURNEY TIME TO ENABLE PATIENTS TO MAKE INFORMED CHOICES ON THE BEST PLACE FOR URGENT CARE.
The platform supports easy integration of new real-time data feeds from Trusts that capture patient flow data using systems like EPIC and Symphony . As well as the user-facing app and website/desktop-based application (accessed using a browser), it includes a business intelligence dashboard designed for use in urgent-care centres. These analytics provide evidence of how the public choose to interact with the app, for example by navigating to travel
directions for an MIU/UCC, what time of the day the public are looking for urgent care services, and more. We are initiating analysis on two questionnaires integrated into the app, which have been collecting responses since the launch of Version 2.0. This feedback will help us to understand how we might improve the platform for supporting patient choice and improving the waiting times in urgent care facilities across the region.
In 2022, the team was awarded the Lyn Thomas Impact Medal [4] by the Operational Research (OR) Society. The medal is awarded in recognition of academic OR research which best demonstrates both novelty and real-world impact backed up by evidence. The work towards NHSquicker was also developed as a REF2021 [5] Impact Case Study for the Business and Management assessment panel (UOA17). The REF impact study “Designing and implementing a digital platform to reduce A & E peak time demand across the South West through the provision of real time information to empower patient decision-making” used impactrelated data from early adopter Trusts to evidence the efficacy of the solution. The project highlights how OR can be used for solving complex health and care issues today, that can support both patients and the NHS.
RESPONSE TO RECENT CHALLENGES FOR URGENT CARE
At a national level, A&E waiting times have continued to increase steeply over recent years, particularly in the wake of COVID-19. In January 2024, NHS figures show the number of people who waited more than 12-hours in A&E departments in England increased by nearly 25% compared with the month before [6]. This is a long way off the NHS four-hour standard, which also applies to MIUs and UCCs as part of the NHS Urgent Care Network. By the end of 2023/24, only 55% of A&E patients met the 4-hour standard (indeed, the intermediary threshold target of 76% which was set to be met
IMAGE 2. NHSQUICKER HELPS PATIENTS WITH URGENT CARE PROBLEMS TO DECIDE WHETHER TO GO, WHERE TO GO AND WHEN TO GO? NHSQUICKER DISSEMINATION LEAFLET FOR DEVON.
by March 2024 also remained unattained); not far off the worst A&E performance in over a decade 1 Despite declining A&E performance, the figures show that MIUs typically continue to meet the standard [7].
NHS England principles and standards for urgent care emphasise the need to provide a consistent urgent treatment offering to the public to reduce attendance at A&E departments and improve patient experience and access to local services. This includes the requirement for all Urgent Treatment Centres to have an up-to-date NHS Directory of Service (DoS) profile to enable effective referrals from NHS 111 and 999 services [8]. This means there is a robust list of services and symptom groups that are available to referring services to help navigators understand the services available at different urgent care facilities. It clearly outlines provision of services and highlights exceptions to provisions to maintain DoS accuracy and maintain clear provision and care pathways. The DoS is also nationally and locally updated with current open and closed facilities making it a real time hub of currently available services.
NHSquicker provides users with real-time, up to date information about wait times, number of patients waiting to be seen, number of patients in departments as well as if their local centre is experiencing an uplanned closure
Version 3.0 of NHSquicker was launched in March 2024 in collaboration with Rachael Shine, Head of Urgent Care Transformation
Programmes NHS Devon Integrated Care Board (ICB), and Tom Monks, Associate Professor of Health Data Science from the University of Exeter Medical School. The key in this version is that, in addition to wait time and travel time information, it uses the DoS Urgent and Emergency Care API to report non-routine closure of urgent care facilities. This is a huge improvement as there are often unplanned closures of local care provision due to workforce shortages and increased demand. Therefore, now, the population is able to have realtime, up to date information about wait times, number of patients in department as well as if their local centre is experiencing an unplanned closure and their next most suitable alternative for care. The integration also provides users with on-demand information for services such as dentists, opticians, pharmacies, and sexual health clinics.
CURRENT BENEFITS OF NHSQUICKER
The benefits have been three-fold. Firstly, NHS Trusts in the South West have interfaced their IT systems with NHSquicker. This app is a way for Trusts to use existing data in an innovative way to improve the patient experience of their urgent care services.
Secondly, patients are finding the app useful. An in-app survey in 2020 found that 78% of users agreed that NHSquicker helped them decide where to go. The new version has advanced analytics features that will further help with evidence.
Finally, data analysis from early adopting Trusts found a significant shift in the pattern of attendance, with a reduction in A&E attendance and an increase in MIU/UCC visits [9]. The
Trusts had a well-planned publicity campaign that increased awareness of the solution. We intend the launch of Version 3.0 will boost awareness to both patients and NHS Trusts to support ICBs to take account of the wider provision of emergency and urgent care services for their populations.
FUTURE RESEARCH USING NHSQUICKER
Alonside the launch of Version 3.0, a number of initiatives are underway by the Exeter research team to maximise the potential value that can be gained from NHSquicker’s real-time data feeds. One ongoing stream of research is investigating the potential for realtime decision-support across urgent care networks through innovative digital twin technology. This work is investigating integration of partial real-time data feeds into a discreteevent simulation model to increase the accuracy of short simulation runs and support short-term system recovery in overcrowding situations. Future work,
in collaboration with Somerset ICB, will pilot additional live data feeds to initialise the simulation model, with a view to expanding the back-end platform of NHSquicker for Trustlevel decision making support.
TO CONCLUDE
The researchers behind NHSquicker have shown that local-level digital initiatives using OR can be scaled up regionally, and there is significant potential to scale from regional to national scale. By unlocking data from multiple sources, NHSquicker empowers patient decision-making and transforms their experience of urgent care services, giving them control over the services they use and the choice of lower wait times.
Simultaneously, the platform is helping NHS Trusts to manage their day-to-day operations. By being responsive to NHS principles and standards, and improving communication and coordination between urgent care services, NHSquicker can improve patient flow
and resource utilisation across the system as a whole.
The researchers behind NHSquicker have shown that local-level digital initiatives using OR can be scaled up regionally, and there is significant potential to scale from regional to national scale
Future research into novel methods such as digital twin technology has the potential to offer enhanced decision support to both hospital and urgent care networks. Real-time simulation can offer live decision-support to reduce or prevent overcrowding in overburdened A&E departments.
ACKNOWLEDGEMENT
Work of this magnitude could only be realised through the shared vision of ‘making data work at both an individual (patient) level and at a more system (Multi-Trust, regional) level’. We would like to specially acknowledge Emeritus Professor John Powell, Susan Martin and Dr Andrew Fordyce who helped
initiate the project and were associated with it for a number of years. We thank Professors Todd R. Kaplan and Surajeet Chakravarty who contributed to the REF2021 Impact Case Study from Exeter. We acknowledge funding received from the initiation of the project and which has included: ESRC Impact Acceleration Awards, Torbay Medical Research Fund, South West Academic Health Sciences Network, University of Exeter Open Innovation awards, and the Business School’s research and impact awards. We thank NHS Devon ICB for commissioning the NHSquicker service (Q2, 2024), and the significant support from the working group team including Marc Jeffery, DoS Lead at NHS Devon, Sebastian Lawrence, Deputy National Directory of Services (DoS) Lead at NHS England, Victoria Hartland, Service Owner – Find the Right Service, Digital Urgent & Emergency Care at NHS England for their continued support, co-development and testing of the app. Additionally, our colleagues at YelloStudio, the software developers, Andy Lake, Geoff Atkins and James Evans for their inputs, collaboration and co-development.
We could not have achieved this without you all.
ORCID
Nav Mustafee https://orcid. org/0000-0002-2204-8924
Nav Mustafee is Professor of Analytics and Operations Management at Exeter Business School. He has led the NHSquicker project since 2017. His research focuses on M&S methodologies and real-time simulation,
and their application in healthcare, circular economy and resilience to climate change. He is Editor of Journal of Simulation and leads the Africa Focus initiative.
Alison Harper is a Lecturer in Operations and Analytics at the Centre for Simulation, Analytics and Modelling, University of Exeter Business School. Her research interests include applied health and social care modelling and simulation, real-time simulation, and reusable open models in healthcare.
Tom Monks is an Associate Professor of Health Data Science at University of Exeter Medical School. His research interests include open science for computer simulation, urgent and emergency care, and real-time discrete-event simulation.
Rachael Shine is the Head of Urgent Care Transformation for NHS Devon. Her portfolio of programmes includes attendance and admissions avoidance in A&E, reducing ambulance dispatch and conveyance where there are suitable alternatives,
FOR FURTHER READING
minimising low acuity patients at A&E, effective navigation of patients to the right place, first time, and optimising utilisation of non-life-threatening urgent care services for appropriate patients.
[1] NHS (2013). England’s Sir Bruce Keogh sets out plan to drive sevenday services across the NHS (NHS England; 15th December 2013). https://www.england.nhs.uk/2013/12/sir-bruce-keogh-7ds/.
[2] Mustafee, N., & Powell, J. (2020). Providing real-time information for urgent care. Impact 2021(1): 25–29. https://doi.org/10.1080/205880 2X.2020.1857601
[4] Exeter researchers receive impact award for digital platform that can reduce A&E waiting times (5th January 2023). https://news.exeter.ac. uk/faculty-of-environment-science-and-economy/university-ofexeter-business-school/exeter-researchers-receive-impact-award-fordigital-platform-that-can-reduce-ae-waiting-times/.
[5] REF. (2021). website. https://2021.ref.ac.uk/
[6] Sharp rise in people waiting over 12 hours in A&E, NHS England figures show (Sky News; 8th February 2024). https://news.sky.com/story/ number-of-people-who-waited-more-than-12-hours-in-a-e-up-bynearly-25-in-january-13066641
[7] Analysis of A&E waiting time (Nuffield Trust). https://www.nuffieldtrust. org.uk/resource/a-e-waiting-times.
[9] Mustafee et al. (2021). Designing and implementing a digital platform to reduce A&E peak time demand across the South West through the provision of real time information to empower patient decision-making. REF2021 Impact Case Study (UOA17). https://results2021.ref.ac.uk/ impact/5b7e4de5-9ccf-452e-8231-34ee3c4a7269.
MAKING THE DIFFERENCE
IN CREDIT RISK AND FINANCIAL EDUCATION –LESSONS FOR FUTURE REF CASE STUDIES
JONATHAN CROOK, TINA HARRISON, GALINA ANDREEVA AND ARSENIY MORGUNOV
CREDIT IS ESSENTIAL IN MODERN SOCIETY due to its pivotal role in promoting economic growth. It allows individuals, businesses, and governments to access funds beyond their current means, enabling them to make purchases, invest in assets, and manage cash flow effectively. Furthermore, credit provides flexibility
in managing financial resources by smoothing consumption over time. However, there is an inherent risk of non-payments or defaults, and throughout the history of credit, lenders have attempted to grant loans to those who are likely to pay back. The modern technology to do so is known as credit scoring, and the vast majority
of people who have a credit card, a mortgage or other type of loan have had their creditworthiness assessed with a credit scoring model.
Credit scoring is a traditional field of Operational Research. Prof Lyn Thomas, IFORS Distinguished Lecturer, in his 2014 INFORMS conference keynote noted: ‘Credit scoring … has been one of the most successful if unsung applications of mathematics in business for the last fifty years.’
Part of this success is due to the work of Credit Research Centre (CRC) at the University of Edinburgh Business School (Figure 1) that has been advancing credit risk assessment for almost 30 years. This article describes the impact made by CRC members as evidenced by REF impact case studies. The Research Excellence Framework (REF) is the UK’s system for periodically assessing (every 6 years) the excellence and impact of research produced by the 157 UK-based Higher Education Institutions (HEIs). The outcomes of each assessment informs the allocation of about £2 billion per year of public money to fund research in universities. In this article we highlight the challenges that the authors encountered while putting these case studies together. This is relevant to both academics and the practitioners sponsoring or simply participating in their investigations. We hope that our experience will help in shaping future rules and procedures for impact evaluation of academic research.
‘Credit scoring … has been one of the most successful if unsung applications of mathematics in business for the last fifty years.’ [Lyn Thomas]
‘Designing the next generation of credit risk models’ [1] by Crook et al. from REF 2014 explained the impact of three research programmes at the CRC that were partly funded by an EPSRC grant for a total of £504k. Survival analysis: Prior to CRC research, lenders used cross sectional models to predict the probability that a borrower would default within a fixed time window of gaining a loan. The new methodology parameterised the first multivariate survival model that predicts the probability a borrower will default for the first time, for any future month. It was also shown that these predictions vary with the state of the macroeconomy. This allowed a bank to predict the expected profit for any macroeconomic scenario. Industry respondents have confirmed that CRC research gave practitioners confidence to follow this approach. Reject inference: Lenders estimate a new credit scoring model using data concerning the performance of previously accepted applicants. Commonly, using such a sample may yield biased parameters for a model intended to relate to the population of all applicants. We showed the effects of using a technique that corrected for this bias and showed the true effects of the methods that were being used in practice. The respondents confirmed that this research changed the way they approach the reject inference problem. Loss given default (LGD): Computing the amount of capital a bank must hold to protect depositors in the event of unexpected severe events necessitates the prediction of the proportion of an outstanding loan that is never recovered in the event of default. This research showed the effects of including indicators of the macroeconomy into these models so that a model-based approach could be adopted to compute
‘stressed’ values of LGD, as are required by the Basel Accords and for internal managerial purposes.
‘Enhanced Credit Risk Assessment for SMEs’ [2] by Andreeva and Ansell from REF2014 showed how CRC research on risk assessment of SMEs helped lenders in understanding SME behaviour to support their credit decisions, increasing credit availability to this vital economic sector. Problems that SMEs encounter in getting credit are well-known and are a constant concern for governments worldwide. There is evidence clearly showing that more accurate credit risk models and better information enhances financial inclusion. The case study presented the evidence of impact from three organisations. The impact consisted in new models being implemented, better understanding of variables that should go into models, and better understanding of SMEs vulnerability during the credit crunch.
There is evidence clearly showing that more accurate credit risk models and better information enhances financial inclusion
‘Helping banks comply with a new regulation on provisions for default risk’ [3] by Crook et al. from REF 2021 described the changes in methodology as prompted by a new
FIGURE 1. THE CRC AND ITS MAIN RESEARCH BENEFICIARIES
regulation (IFRS9) covering most banks outside the US, and another (CECL) covering banks in the US. According to these regulations, each bank must predict the expected future losses on each loan, if the risk associated with the repayments on that loan changes after the loan was granted. The vast majority of banks had not modelled this concept before. Research on survival models by members of the CRC provided a methodology to make these predictions. The researchers also investigated whether the parameters of such models remained stable over the 2008 crisis and demonstrated methodology to make the models more accurate by allowing their estimated parameters to change over time. The research very strongly influenced the methods adopted by lenders and consultancies in the US and UK to comply with the regulations. This was evidenced by written submissions from leading consultancies and banks.
The above case studies described the impact on lenders, yet credit is a bilateral process, and borrowers’
responsible participation is equally important. On one hand, better risk models ensure that financial institutions can offer credit more responsibly and with greater precision. On the other hand, improved financial education equips borrowers, and in particular young people, with the knowledge to manage that credit wisely, reducing the risk of defaults and fostering a culture of responsible borrowing and lending, as ‘Impact on financial wellbeing of Young Adults’ [4] by Harrison et al. from REF2021 demonstrated.
credit is a bilateral process, and borrowers’ responsible participation is equally important
Young adulthood is a key stage of life when young people are transitioning to adulthood and financial independence. Being financially capable is crucial. Yet, over half (52%) of UK 18–24 year-olds show signs of potential financial vulnerability with 11% already in difficultly (Financial Lives Survey, 2017).
Despite the introduction of financial education into the school curriculum in England in 2014, the mandatory requirement to teach financial education currently stops at age 16, just at the point when young people need it most. Moreover, teachers report a general lack of confidence and skill to teach financial education and a lack of specialist training and support.
This research has helped to understand what works in developing effective financial education for young adult school leavers aged 16–18 years. The impacts included shaping the direction of the UK-wide Financial Wellbeing Strategy for young adults, influencing a leading financial education charity’s training for teachers that led to an improvement in post-16 school leavers’ ability to manage money, and influencing the development of a pathfinder project with the Welsh Government to establish a sustainable approach to teacher professional development for financial education.
Yet putting together a REF case study is not an easy task. To gain impact, the benefits of new research
have to be disseminated to potential users. The findings from projects above were presented at the practitioner attended biennial Credit Scoring and Credit Control conference (typically attracting over 400 delegates) that the Centre organises and at other major conferences which also attract practitioners, like the OR Society conference, INFORMS and the SAS conferences. Invited presentations were given directly to model builders in banks and to regulatory agencies like the FCA. There was also a wide-spread media coverage of the research. This all requires a lot of effort and an excellent research support team at the University.
REF impact cases were assessed according to ‘reach’ and ‘significance’.
Credit scoring methodologies can be a source of competitive advantage and commercially very sensitive. So, it can be challenging to identify the lenders whose methodologies were changed by a particular research project and to gain indications of how profound the changes were. Lenders are reluctant to share the specifics of what they use. Even when they find the research useful, they may not be willing to talk about it. The contributors often choose to remain anonymous.
A key challenge for the Young Adults project was demonstrating and quantifying the downstream impacts on young people in a project where the impact was channelled through teachers who participated in the professional development. Access to young people was via teachers and with consent of parents. Teachers are very busy and research is not their priority. The researchers needed teachers on board to help with data collection. Incentives are
helpful, but cannot overcome pressures on teachers’ time. There were also difficulties due to COVID and school closures.
The preparation of future impact cases may be enhanced if the funding councils could award universities some of the cost incurred to discover which organisations have been influenced by a new piece of research. It might also help if funding councils could give credible commitments to financial and other institutions that the content of their evidence will be retained in a highly secure environment. More meaningful incentives and novel ways to recognise the contributions should be considered. Perhaps, the OR society can lead a discussion around it.
Jonathan Crook is Emeritus Professor of Business Economics, CRC founder and former Director. He concentrates on: modelling of credit risk and operational risk; the Economics of consumer credit including the demand, the supply of credit and credit constraints using household level data; household over-indebtedness.
REFERENCES
Tina Harrison is Personal Chair of Financial Services Marketing and Consumption, Co-Head of the Marketing Group, CRC member. Tina's research interests are in the area of financial services marketing, specifically financial wellbeing, analyses of consumer use and understanding of financial services and the use of technology in enabling and empowering effective financial management and decision-making.
Galina Andreeva is Personal Chair of Societal Aspects of Credit, Co-Head of the Management Science & Business Economics Group, CRC Director. She researches credit risk using advanced statistical and machine-learning techniques, applied to Big Data, in particular Open Banking. An area of specific focus is financial vulnerability, fairness and social impact of credit.
Arseniy Morgunov is CRC administrator and startup advisor for entrepreneurs in developing countries. His research focuses on price-setting characteristics in developing art markets and legal aspects therein.
[1] Crook, J. (2014). Designing the next generation of retail credit risk models. REF Impact Case Study, University of Edinburgh. https:// impact.ref.ac.uk/casestudies/CaseStudy.aspx?Id=23966
[2] Andreeva, G. and Ansell, J. (2014). Increasing insights into credit risk of Small and Medium sized Enterprises (SMEs). REF Impact Case Study, University of Edinburgh. https://impact.ref.ac.uk/casestudies/CaseStudy. aspx?Id=23968
[3] Crook, J., Bellotti, T., Leow, M. and Djeundje, V. (2021). Helping banks comply with a new regulation on provisions for default risk. REF Impact Case Study, University of Edinburgh. https://results2021.ref.ac.uk/ impact/29c75d9f-5348-4bbc-be08-0cc1343c0b8e?page=1
[4] Harrison, T., Ansell, J. and Marchant, C. (2021). Enhancing the financial capability and financial wellbeing of young adults. REF Impact Case Study, University of Edinburgh. https://results2021.ref.ac.uk/impact/ 05e2d0ae-c2c3-4a71-95f4-f30821566e6e?page=1
SIMULATION FOR SOCIAL GOOD – COMBATTING SOUTH AFRICA’S HIV EPIDEMIC
LAURA REID
South Africa has the largest population of people living with HIV in the world, with one in five people living with the virus. Close to 200 young women become HIV positive every day. Meanwhile, one in three women has experienced physical and/or sexual violence, and eight women report gender-based violence every hour.
Shout-It-Now (Shout) is a NonProfit Organisation (NPO) providing
HIV prevention, sexual and reproductive health advice, and gender-based violence support services to communities in South Africa.
To prevent HIV infections and support those living with HIV, Shout brings a range of free health services to hard-to-reach communities. These include HIV testing, PrEP (the HIV prevention pill), STI screening, contraception, gender-based violence
prevention and counselling, and educational life skills programs.
The service is delivered via a fleet of mobile healthcare vehicles that are staffed by nurses, social workers, specialist health advisors and a support team that travels into communities daily to deliver the vital services where they are most needed.
It’s about making someone feel important and building a connection.
Buhel Sithole is a quality assurance officer at Shout-It-Now. She explained how the clinics work to deliver a vibrant and welcoming experience to all of their clients to encourage more positive conversations about sexual health and relationships: ‘We want to run our clinics more like a premium shopping experience. When you walk in, staff will start the conversation with a smile and give that person their full attention. It’s about making someone feel important and building a connection. This focus on improving the client experience is essential in encouraging more young women to use the service, share with their peers, and make repeat visits’.
ENHANCING SERVICES WITH TECHNOLOGY
With the HIV epidemic so rife in South Africa, and with the typical limits on resources faced by NPOs, Shout is always on the lookout for new ways to help them reach more vulnerable young women. The use of innovative technologies to improve services is something that has often served Shout well. It has multiple integrated systems, apps and chatbots, each specifically designed to deliver value for the client and/or the service
provider. It was this type of thinking that led the team to approach the process simulation experts at Simul8 to see if they would be interested in working together to help improve process efficiency.
At the heart of Simul8’s corporate values is a Tech for Good program that offers free software licences and consulting on a pro bono basis to charitable and social good organisations around the globe. When they discovered more about the work that Shout carries out, it was an easy decision, as Laura Reid, CEO at Simul8, explained: ‘We’re fortunate in the West where we have relatively generous resources to diagnose and treat people with HIV. When we heard about Shout-It-Now’s efforts to tackle this epidemic in South Africa in less fortunate communities, we felt a real responsibility to offer our support. We’re so happy our technology is making a difference to young women accessing these essential services’.
Simulation offered an excellent approach for Shout to improve its evidence-based decision making to enhance the efficiency and effectiveness of its service delivery, a concept that closely aligned with its technologydriven approach.
REDEFINING PROCESSES THROUGH SIMULATION
Shout outlined three core objectives for the project: increase access to HIV testing; increase the use of the HIV prevention pill; and provide more support to those experiencing genderbased violence. A target was set for a 15 percent uplift in the number of people that staff could see through uncovering process improvements with the use of simulation.
In the first three months of the project, however, Shout managed to reach double this target thanks to the insights discovered via process simulation:
• A 44% increase in clients being tested for HIV
• A 27% increase in at-risk clients being initiated onto PrEP
• A 30% increase in the number of gender-based violence victims they were able to support (Figure 1)
CREATING THE SIMULATION
The operations team at Shout worked closely with a Simulation Excellence Team assigned by Simul8 to conduct a detailed analysis of the fleets’ typical visitors and the types of services they required.
Andrew Wylie was one of the Simul8 consultants that worked on the project. He outlined the approach that was used to build the simulation: ‘First, we needed to understand the profiles and numbers of the different people using Shout’s services in terms of their age group and gender, as well as whether they were in a rural or city setting. We also incorporated whether they’d previously tested positive for HIV into our analysis, all to build up typical customer profiles that we could map against the varying demands for Shout’s services’.
‘We didn’t want to take a blanket approach to the model, not least because we were dealing with a very emotive subject. It was really important to acknowledge that the experience of each visitor would differ depending on their needs. For example, a teenage girl seeking birth control advice in a rural area would have a different pathway compared to a woman in her 40s that had contracted HIV. They may require
more than one nurse for safeguarding issues, or longer appointments to discuss complex problems. They might need to see a nurse, a counsellor, or both … how long would a typical engagement take?’
We didn’t want to take a blanket approach to the model, not least because we were dealing with a very emotive subject. It was really important to acknowledge that the experience of each visitor would differ depending on their needs.
‘The beauty of simulation over any other technique is that it’s able to capture all of these nuances. We’re able
to simulate the journeys of multiple individuals and give them all unique pathways so that the simulation paints an extremely accurate picture of how things perform in the real world. This is what gives simulation its superpower in helping to analyse all kinds of different processes, including those with lots of moving parts’.
‘We created specific attributes for each touchpoint within the service, as this allowed us to identify the patients’ makeup and how they travelled through the system. Using simple steps and limited inputs, we confirmed these fundamental drivers behind the model: age, gender, services required and location’.
‘Once we had broken down the current processes into suitable data, we
created a clearly defined scope for the model, which helped us to understand what data was essential to the outcomes. We also ensured our data analysis could confirm or challenge these assumptions, and that any data that fell outside the agreed framework didn’t feature in the simulation’.
Incorporating short sprints of development with regular feedback from Shout-It-Now led to the creation of a simulation that enabled the team to run hundreds of simulations representing different configurations. Here they were able to find the optimal performing configuration and then plan the effective distribution of resources, including people, time, skills and equipment, and this could apply to both now and in the future.
FIGURE 1 FIELD TEAMS AND SUPPORT STAFF ARE AT THE CENTRE OF THE SHOUT-IT-NOW APPROACH.
FUTURE-PROOFING THE MODEL
Moving forward, the simulation will be used routinely to plan staff configurations, levels, shift patterns etc. By using Simul8’s dashboard functionality, simple interfaces were created so that operational managers with no prior simulation skills could easily pick it up and run their own simulations. This is a successful approach to democratise the use of simulation in all kinds of organisations, and Shout was left with a really useful tool that would help them continue to optimise operations even as variables changed over time. That could mean extending its fleet, changes in staffing or making decisions on how to incorporate more services.
Andrew said: ‘We aimed to lead the user to the required results through simple step-by-step instructions. For example, the demographics of those using Shout’s services would change depending on the location of the mobile health vehicles, which would impact the number of staff needed. Shout asked us to incorporate a function to also simulate this so they could see as many clients as possible, regardless of the location’ (Figure 2).
‘How did we do it? We looked at the vehicle layout, which has five consulting cubicles, three for Health Care Advisors (HCA) and two for nurses. Then, we looked at how tweaking each scenario to include varying levels of staff would impact the data running through the model. The team then built a function that
enabled users to choose between City or Rural and manually adjust the number of HCAs and nurses available and the time each cubicle would be in operation. This easy-to-use functionality enables all staff to model multiple scenarios with minimal input, so it can be used over and over again’.
MEANINGFUL RESULTS
As well as increasing the number of people it can see each day, this simulation project means that Shout is also able to reduce waiting times, maximise the time its team can spend one-on-one with clients and improve working conditions for staff.
Buhle Sithole described the difference that it has made: ‘The
FIGURE 2 THE SIMUL8 DASHBOARD FOR SHOUT-IT-NOW
simulation project has given us the benefit of time, and when you give people more time, you give them the opportunity to make the right choices. If waiting times are too long, for example, the embarrassment or stigma around sexual health for young people in South Africa often meant clients would leave rather than wait to be seen. This could mean that they would not receive emergency contraception or vital support’.
This project has also positively impacted the wellbeing of the volunteers and healthcare professionals that work for the NPO, who often face emotionally challenging situations. By optimising operations, the workforce is now able to manage the workload better and even take a break for lunch, which was not always possible before.
The simulation project has given us the benefit of time, and when you give people more time, you give them the opportunity to make the right choices.
Laura Reid started as a Simulation Analyst in the NHS, where she was involved in using simulation software to explore scenarios for emergency and elective patients. She then joined SIMUL8 Corporation in 2002 and has been CEO
since 2012. Laura has an MSc in OR from the University of Strathclyde and a BSc in Mathematics and Statistics from the University of Glasgow.
To find out more about Simul8’s Tech for Good programme, visit www.simul8. com/about/tech-for-good. To find out more about the work of Shout it Now, or offer support, visit www. shoutitnow.org.
INTRODUCING SIMULATION INTO YOUR ORGANISATION: Q&A
What is simulation?
Discrete event simulation is a powerful technique for optimising processes and making confident, impactful decisions.
How does simulation work?
A simulation is an animated model that mimics the operation of an existing or proposed system, like the day-to-day operation of a bank, running an assembly line, or assigning staff in a hospital or call centre.
What processes can be simulated?
As a general rule, any system that involves a process flow with events can be simulated – so any process you can draw a flowchart of, you can simulate.
What are the benefits of using simulation?
• Less costly than real life experimentation
• Testing a variety of different ideas under the same circumstances
• Determining the long-term impact of process changes
• Gaining impartial insights to facilitate process improvement
• Determining the potential impact of random events
• Utilising non-standard distributions
• Enhanced process management
• Its visual and animated features help to communicate your ideas
KEEPING ANTIBIOTICS
EFFECTIVE – COMBATTING ANTIMICROBIAL RESISTANCE IN U.S. AGRICULTURE
GERALD MIDGLEY, AMBER ELKINS, GUY H. LONERAGAN, MEGAN BABOWICZ, MAYUKH DASS, YRJO T. GROHN, ELLEN JORDAN, GUILLAUME LHERMIE, LUCAS LUNT, WILLIAM A. MCINTOSH, JUAN M. PIÑEIRO, JASON SAWYER AND H. MORGAN SCOTT
ANTIMICROBIAL RESISTANCE
Antimicrobials are medicines that are designed to kill a variety of organisms, most notably diseasecausing bacteria. They include antibiotics and other anti-bacterial substances.
Unfortunately, bacteria may develop resistance to antimicrobials, and this poses a major threat to human health in the 21st Century: the discovery of
new antimicrobials has slowed down, but resistance to existing ones is continuing to increase. If antimicrobials lose their effectiveness, millions of people will die from illnesses that can currently be treated.
Any use of antimicrobials can stimulate resistance if the context enables resistant bacteria to multiply at the expense of non-resistant ones. The over-prescription and misuse of antimicrobials increases the likelihood
of resistance spreading, so this needs to be tackled by public health systems.
Unfortunately, this is not just a problem for human health: antimicrobials are also prescribed to food animals (like cows, pigs, sheep and chickens), and some antimicrobial resistant bacteria can move between animals and people. Antimicrobial resistance (AMR) in agriculture therefore poses a double risk to humans: reduction in the effectiveness of antimicrobials and disruption to food security.
The over-prescription and misuse of antimicrobials increases the likelihood of resistance spreading, so this needs to be tackled by public health systems.
A WICKED PROBLEM
This is a truly ‘wicked’ policy problem because there are so many interacting factors to account for: not just between the human and animal health systems, but also between uses of antimicrobials and the profitability of livestock industries and pharmaceutical companies.
There are multiple stakeholders with an interest in the problem, and they often find themselves disagreeing on what it means to use antimicrobials judiciously (or wisely). They also commonly disagree on governance, with government bodies often looking at how industry could be better regulated, and industry resisting top-down regulation and instead advocating for self-regulation, or the voluntary stewardship of antimicrobials. This disagreement is particularly acute in the USA, where there is a strong culture in many agricultural communities of resisting ‘big government’. It is quite unclear what an effective model of voluntary stewardship, that addresses disagreement, might look like.
It is quite unclear what an effective model of voluntary stewardship, that addresses disagreement, might look like.
IMPROVING VOLUNTARY STEWARDSHIP
In this context, our team has been working with stakeholders to look at the potential for improving voluntary stewardship of antimicrobials in US agriculture. The funder of our work was the US Department of Agriculture (USDA) [7].
Voluntary stewardship is an approach that relies on the willingness of foodanimal producers and supportive industries (e.g., veterinary services and pharmaceutical companies), as well as broader stakeholders (e.g., public health policymakers and consumer advocates), to collectively ensure the judicious use of antimicrobials without the need for regulation, legislation, mandatory compliance or statutory enforcement.
We ran design workshops with four separate stakeholder groups: beef producers, dairy industry representatives, public health policymakers and consumer advocates. We used an approach called
Critical Back-Casting for this. It integrates methods from two different systems methodologies: Idealized Design [1], which liberates the creativity of participants and moves toward agreement on far-reaching plans for change, and Critical Systems Heuristics [6], which offers twelve questions on what a system ought to be doing, who should have decision-making authority, what expertise is important, and what could give the system legitimacy. These questions are particularly useful for exploring governance issues, so deploying them in the context of Idealized Design enabled our stakeholders to think creatively about what would be required for the governance of the voluntary stewardship of antimicrobials [8].
The participants were asked to imagine that all current projects and programs to address AMR had been stopped, and their task was to propose new, creative designs that would be unconstrained by what currently exists. However, to prevent the proposal of unattainable utopias, all the designs had to be technologically feasible (either using current technology or technology that could be developed in a timely manner), viable (affordable and
socio-environmentally sustainable), and adaptable (capable of flexing, or being revised, in response to future, emerging challenges). Because all four stakeholder groups answered the same questions, their views were comparable.
See the Methodological Reflections near the end of this article for practical pointers on how to use Critical BackCasting in your own projects.
SHARED VIEWS ACROSS STAKEHOLDER GROUPS
There were some common views across the beef, dairy, public health and consumer advocate stakeholder groups on what a voluntary stewardship program should look like. There were also differences of perspective that would have to be addressed to make voluntary stewardship fully operational.
The common views, or emerging consensus, can be summarized in a list of the major characteristics of a potential voluntary stewardship scheme:
• The aim of a program should be the judicious use of antimicrobials, not zero use (less is better, but for animal-welfare reasons, zero is not an option).
• Multi-stakeholder governance should be put in place, with industry in a lead role, and other diverse stakeholders included (either as full decision-makers, external partners or experts to be consulted).
• The governing body should oversee training and information provision for producers and other key decision-makers, and should promote education in wider society about AMR and stewardship.
• The program should be strongly science-informed, with research driving development and change.
• The governing body should also oversee the certification of
demonstrating good stewardship practice.
• The results of certifications (and monitoring information more generally) should be publicly accessible, to enable peer pressure, benchmarking, producer self-reflection and informed action.
• There are existing governance and certification programs for other purposes that could readily be built upon.
• Certification should be linked to marketing, so financial benefits for producers flow from engagement in the program.
• Over-uses or misuses of antimicrobials mostly stem from operational and communication issues in wider agricultural production systems, so judicious use means changing those systems, thereby reducing disease and antimicrobial use. This should be the focus of certification, and participants said it would involve taking a systems approach.
• In line with the last point, governance should be focused on improving overall system perfor-
mance, rather than control over clinical judgments made by veterinarians who decide whether to prescribe antimicrobials to individual animals. If action for systems change successfully reduces disease, then decreases in the use of antibiotics will follow, together with lowering the risk of AMR.
• Funding should come from ‘checkoff dollars’ – money collected by industry organizations and professional associations that work for the collective benefit of their members.
GOVERNANCE OF THE COMMONS
It is striking that this model conforms in almost every respect to the principles in Elinor Ostrom’s Nobel-prizewinning approach to governing common-pool, natural resources [4]. Antimicrobial-susceptible bacteria need to be viewed as a common-pool resource, as they are integral to the ecosystems used by human beings when they raise food animals, and Ostrom’s model could provide a useful template when
designing an actual, effective voluntary stewardship program.
Ostrom’s research [9] shows that, when dealing with a common-pool resource, government-enforced regulation and laissez-faire policies can both meet strong stakeholder resistance, resulting in conflict that damages businesses and undermines sustainability [5].
A more effective approach is often the collective governance of the resource by relevant stakeholders, who need to make decisions in relation to a broad set of economic, social and environmental values. All these values must be managed simultaneously, and it is unacceptable to put off considering one while another is exclusively focused upon – prioritizing values and focusing on them one at a time generally results in the continual discounting of longer-term needs, such as combatting AMR, in favor of addressing shorter-term concerns, such as profitability.
Governance is enabled by a strong focus on the collection of data on activities and impacts, and the provision of information in a way that makes it immediately transparent when important values are being compromised, so peer pressure can be applied and remedial action taken.
A more effective approach is often the collective governance of the resource by relevant stakeholders, who need to make decisions in relation to a broad set of economic, social and environmental values.
DIFFERENCES OF PERSPECTIVE
The above consensus, and its alignment with Ostrom’s model, are encouraging for the potential utility of voluntary stewardship. Nevertheless, some
difficult issues and important differences between the views of stakeholder groups were identified, and these need to be addressed in the development of an actual voluntary stewardship system:
There are structural differences between some food-animal industries. For instance, there is a fair amount of vertical integration in the dairy industry: cooperatives buy milk from the producers, and they have the power to set standards to mitigate AMR. Producers must conform to these standards if they want to sell their milk to that cooperative. In contrast, there is little vertical integration in the beef industry: many small producers sell calves to the feedlots, often through intermediaries such as auction marts and cattle buyers, and then the feedlots compete to sell animals to wholesalers and retailers via ‘meat packers’ (slaughterhouses or abattoirs). The beef producers are particularly concerned that husbandry standards among the small producers are variable (and these standards are typically unknown prior to the purchase of calves), which accounts for most of the perceived need for antibiotics. However, there is no single body (e.g., a cooperative) able to strongly influence husbandry standardsetting. Different kinds of programs, each with different emphases, will therefore be needed for different food-animal industries.
The beef industry participants mostly focused on local-scale governance, while the dairy participants primarily looked at the national scale. There is a useful model of an adaptive, multi-scale organization, the Viable System Model (VSM) [10], that can reconcile these foci. It is possible to identify three or more ‘levels’ of governance: national-level (establishing general parameters for stewardship in a science-informed manner), program-level (ensuring each industry certification program meets the specific needs in that industry, as mentioned earlier), and
business-level (looking in each company at how to implement stewardship and secure certification). While the OR project discussed in this article was only focused on the US context, it would be possible to have a global level of governance too, if this could be agreed through international negotiations.
The participants in the four workshops identified different stakeholder groups when it came to inclusion in governance. There is an opportunity for synergy here, because the industry participants mainly looked at the key professions who would need to be involved in implementing voluntary stewardship in agriculture, while the public health policymakers and consumer advocates mainly discussed beneficiaries in wider society. Referring again to the VSM and the three levels of governance mentioned above (national-, program-, and business-level), different stakeholders might be involved at each of the three different levels. Also, the researchers offered a new method for stakeholder analysis that could be useful in the design of an actual voluntary stewardship scheme, as it counters two biases that are common in stakeholder analysis: bias to the status quo, and bias to those who already have a voice in the system [3].
There was a discussion of shaming and stigmatization, with some public health policymakers saying that these are good things to encourage when producers opt out of voluntary stewardship. Peer pressure is certainly necessary, but if the stigmatization is perceived by industry as coming from regulatory authorities, it could undermine voluntary stewardship. The right kind of peer pressure comes about when all industry players can see their own performance in relation to the performance of others, so those who are more successful in addressing AMR then encourage others to make improvements.
CONCLUSIONS
The full report of this study can be found in [7]. Ultimately, the findings from this research (and indeed the Ostrom governance model and the VSM) should be considered a useful resource, not a blueprint for implementation [8]. This is important because multi-stakeholder trust and collaboration can be undermined by attempts to impose top-down ‘solutions’.
For the design of an actual voluntary stewardship system, it may be useful to replicate the workshops process used in this research, as it was highly participative – except that more time would be needed, after initial workshops with separate stakeholder groups, to bring stakeholders together and develop a fully collaborative vision. Also, there will no doubt be technical questions about antimicrobials and their use (put beyond the boundaries of this research) that will need to be addressed once voluntary stewardship is established.
The research team would be very interested in supporting any industry or policy organization that wants to take forward this vision of voluntary stewardship. Please feel free to make contact.
METHODOLOGICAL REFLECTIONS
The twelve Critical Back-Casting questions, presented in a generic, plain-English form (not specific to the governance of antimicrobials), can be found in the shaded box. The word ‘system’ has been used, but it could be replaced by ‘service’, ‘organisation’ or any other term that is relevant to the context. These questions are used in workshops with stakeholders, where the participants imagine that current systems have ceased to exist, and they have come together in a team to design new ones. However, as mentioned
THE TWELVE QUESTIONS
(1) Who or what should benefit from the system, and how?
(2) What should be the purposes of the system; i.e. what goals should it aim for in order to deliver to the beneficiaries?
(3) What should be the system’s key measures of success?
(4) Who should be seen as the key decision makers; i.e., have the authority to change who should benefit, what the purposes should be and how success should be measured.
(5) What components (resources, people, policies, etc.) should be under the authority of the decision makers?
(6) What is essential for delivery of the benefits and purposes, but should not be under the authority of the decision makers?
(7) Who, either in addition to or instead of the decision makers, should be involved in delivering the benefits and goals?
(8) What should count as expertise; i.e. who should be considered an expert and what should be their roles?
(9) What are the key factors that will guarantee (or increase the likelihood of) success?
(10) Who or what could be affected by the activities of the system; should the affected be represented in decision making, and (if so) how?
(11) To what extent should the affected be able to retain independence; i.e., opt out or neutralise the effects on them, and/or take actions of their own choosing?
(12) Upon what core values and assumptions should the system be based?
earlier, the participants need to make sure that their designs are technologically feasible, viable and adaptable in the face of future changes.
Critical Back-Casting has been used in approximately 20 projects besides the one presented here, with various participants: homeless children, older people, children in residential care, people with mental health problems and many service providing stakeholders [2]. Several general reflections based on this experience can be provided to support readers in thinking about how they might apply Critical Back-Casting in their own projects:
Facilitation:
• A facilitator is needed to make this work.
• Once a facilitator has used the questions in several projects, they
become internalized sufficiently to inform more free-form facilitation exercises, without the need to go through them systematically.
The questions:
• For every one of the questions above, 6-10 follow-up questions need to be asked to tease out details specific to the context.
• The questions work equally well with professionals, ordinary citizens and people with marginalized identities who have had no previous experience of planning and management. Indeed, more frequently than not, ‘ordinary’ citizens and marginalized stakeholders find it easier to generate far-sighted designs than professionals,
because the latter tend to be more disempowered by limitations built into their current organizations.
The process:
• Expectations need to be managed. Ideally, the method is used in a real planning initiative where stakeholders can be confident that their ideas will inform action. If this is not the case (e.g., if the exercise is only going to inform recommendations for action that may or may not be implemented), then participants need to know this.
• Power relations matter. If the participants don’t feel they can talk freely and openly in front of one another, the process will fail. When free and open communication is not possible, an antidote is to run separate groups with different categories of stakeholder.
• There are usually moments in the flow of the discussion when it feels natural and necessary to deviate from the questions to look at what the specific structures for governance should be. This often happens once people have realized that the meaningful engagement of stakeholders is necessary, and they want to look at how this can be accomplished.
Implementation:
• Supporting people with action planning after Critical Back-Casting is essential.
Finally, as long as people can talk freely, workshops using this approach are tremendously exciting (sometimes euphoric) because they almost always generate far-reaching insights. This is therefore a useful approach for providing
a foundation upon which to build further collaboration into the future.
Gerald Midgley has had research leadership roles in government, academia and consultancy, and has extensive experience in public policy, community development, technology foresight and resource management projects. He is an Emeritus Professor at the University of Hull (UK) and holds Visiting Professorships at the University of Birmingham (UK) and Linnaeus University (Sweden).
Dr. Amber D. Elkins is a Senior Data Scientist in Public Health for projects under the United States Department of Defense, Booz Allen, and MAT, Inc. She focuses on fostering improved understanding of complex social problems, given practical and theoretical constraints, within or emergent from different socio-ecological,
FOR FURTHER READING
cultural and political systems to increase intervention efficacy.
Guy Loneragan was raised in rural Australia. He is the founding Dean of the Texas Tech University School of Veterinary Medicine, the sole purpose of which is to serve the needs of rural and regional communities. Guy is a veterinarian and an epidemiologist. His scholarship focuses on the connection between animal and human health.
H. Morgan Scott is a veterinarian with a PhD and post-doctoral training in public health. A professor of epidemiology in the Department of Veterinary Pathobiology at Texas A&M University, he applies both epidemiological and ecological approaches to better understand the propagation, dissemination and persistence of enteric bacterial pathogens – including those resistant to multiple antimicrobials.
[1] Ackoff, R. L., J. Magidson and H. J. Addison (2006). Idealized Design: Creating an Organization’s Future. Upper Saddle River: Wharton School Publishing.
[2] Midgley, G. (2000). Systemic Intervention: Philosophy, Methodology, and Practice. New York: Springer.
[3] Midgley, G. et al (2023). Toward System Change to Tackle Antimicrobial Resistance: Improving the Voluntary Stewardship of Antimicrobials in US Agriculture. Hull: Centre for Systems Studies.
[4] Ostrom, E. (1990). Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge: Cambridge University Press.
[5] Ostrom, E., C. Chang, M. Pennington and V. Tarko (2012). The Future of the Commons: Beyond Market Failure and Government Regu lation London: Institute of Economic Affairs.
[6] W. Ulrich (1994). Critical Heuristics of Social Planning: A New Approach to Practical Philosophy. Chichester: Wiley.
[7] Final report to the US Department of Agriculture available here: bit.ly/ USDA-AMR-Report-Hull.
[8] Blog providing more methodological guidance and additional reflections: bit.ly/Critical_Back-Casting_blog
[9] A summary of Ostrom’s principles for managing common-pool natural resources: https://www.onthecommons.org/magazine/elinorostroms-8-principles-managing-commmons/index.html
[10] For a brief, up-to-date description of the viable system model, see (pp. 749–751): https://www.sciencedirect.com/science/article/pii/S0377 22172300512X
THE TRAVELLING SWIFTIE PROBLEM
MATTHEW HOWELLS With the world renowned The Eras Tour finally reaching the United Kingdom this Summer, I present how a classic OR problem can be applied to one of Taylor Swift’s hits.
Specifically, I’m looking to her 2019 album, Lover, and its eleventh track, London Boy. In this catchy tune, Swift tells the story of a Tennessean woman’s
love of a London boy, and the trips and memories they make together in London, name-dropping several locations in the city.
On release, the song caused quite a stir on social media, with many assuming that the couple visited all locations in one day, pointing to the impracticality of such a day out. Swift
Photo by Stephen Mease via Unsplash
has since clarified that the events of the song actually unfold over 3 years, but for the sake of fun (and OR!), I raise the following questions—(1) What if the couple did visit all these locations in one day? (2) How could their day out be made as efficient as possible? Sounds like a Travelling Salesman Problem (TSP) to me!
In the unlikely event you landed on this page and have never heard of the TSP, do not worry! The TSP is a classic problem in OR. It poses: “Given a list of locations, and the distances between each pair of locations, what is the shortest possible route that visits each location exactly once, and returns to the origin location?” The open variant of this problem removes the constraint of returning to the origin location, which is what we will consider in this scenario.
Thus, the Travelling Swiftie Problem is born. Given a list of the nine locations the Tennessee girl recalled visiting in London Boy, what is the shortest possible route that visits each location exactly once, and ends the journey for a night out in Brixton? When setting up the problem, we have taken the following assumptions:
• In the song, some places (e.g. Highgate) are repeated in choruses. Since they are choruses, we choose to ignore these, so each place is visited once.
• Our Tennessee girl mentions enjoying nights in Brixton in the song. We assume Brixton will end the tour so the two can enjoy a night out.
• The song begins with a snippet of the London boy saying the couple can go riding on his scooter, so we assume this is the mode of transport.
• Traffic is ignored, so our scooter is not slowed down and we are
concerned only with distance travelled.
• They park the scooter at feasible parking facilities near the locations, the precise location of which is not relevant.
Leaving mathematical details aside, in formulating the problem we aim to minimise the total distance travelled by the pair across the tour of London, ensuring that: each of the first eight locations is both entered at most once and left exactly once; the final stop, Brixton, is never left, but is entered exactly once, we have no sub-tours in our solution but instead one complete tour. All mutual distances between any one location in the list of nine and the remaining eight need to be gathered before attempting to devise a solution. These may be collected from widely available systems such as Google Maps, and is likely expressed in (statute) miles.
Taking Taylor Swift’s route to be the order of locations as they are first mentioned in the song, this suboptimal route is: Camden Market, Highgate, West End, Shoreditch, Hackney, Bond Street, Hampstead Heath, SoHo, Brixton. On this route, the couple would travel for 32.5 miles—quite a long journey around the City!
Typically, TSPs are very computationally expensive to solve, and we may need to use a heuristic (a rule of thumb) to achieve a credibly strong estimate of the answer, as opposed to finding the optimal answer. However, since this problem features only nine locations, we can solve it to optimally with computational ease. In true Reputation style, I used the coding language Python to cycle through all permutations and return the best route to us.
As a result, we find the actual optimal route: Hackney, Shoreditch, Highgate, Hampstead Heath, Camden Market, SoHo, Bond Street, West End, Brixton. Taking this route, they travel 19 miles, a saving of 13.5 miles, and plenty more time for their Love Story (Figure 1).
BEYOND THE PLANNING OF LEISURE TRIPS
Having seen how useful using OR can be to our Tennessee girl and her London boy, the potential routing problems we can formulate and solve as a TSP extends far beyond planning an efficient day out.
A company manager may wish to utilise the TSP to find the optimal route for a delivery vehicle to traverse a geographic area. This would yield several benefits, including saving fuel and hours travelled by the delivery team, reducing the company’s carbon footprint, and raising customer satisfaction for on-time delivery.
For a long time, the TSP has also been extended to become a Vehicle Routing Problem (VRP). This is similar to the TSP, but normally involves a team of salesmen or a fleet of delivery vehicles. VRPs are solved to find the optimal tours for each salesman or vehicle to take to visit all locations around the area. This is particularly useful for logistics companies that manage multiple delivery vehicles, as it helps to ensure that each vehicle’s route is optimized for efficiency and cost-effectiveness.
Healthcare can benefit from the TSP too. For instance, TSP helps to efficiently plan the routes of nurses involved in the delivery of home care services to multiple patients over a given timespan. This not only saves travel time and fuel costs but also,
crucially, allows more time for patient care.
Modellers can build further on with increased constraints to more accurately reflect ever more realistic challenges. We ignored it in the Travelling Swiftie Problem, but traffic congestion can be a real pain to delivery vehicles and their drivers, with potential to greatly alter the distance travelled and the costs involved. We may also need to consider sudden changes to the route, from unanticipated road closures or emergencies, or from last-minute delivery alterations requested by customers. Of course, adding more constraints increases the complexity of the problem, but is often a strict necessity.
TRY IT OUT YOURSELF!
If you got to this point and feel like some of your problems (whether personal or business) would do with a TSP (or VRP) formulation and
solution, you may then ask yourself: “How can I or my team go about doing this, at no extra cost?.” The answer may depend on the level of mathematical and programming skills already available in the group.
A useful starting point may be to look at Google OR-Tools (https:// developers.google.com/optimization/ routing), a suite of optimisation tools provided by Google. The site has a guide on various routing optimisation problems, including the TSP and VRP, detailing the problems and explaining step-by-step how to solve them, with example code written in Python, C++, Java, and C# that solves these problems using Google OR-Tools.
If you are not prepared to move away from spreadsheet tools such as Microsoft Excel (arguably still the standard approach for many of us), an alternative might be to try out the VRP Spreadsheet Solver developed by Professor Gunes Erdogan of the University of Bath, on which we
extensively reported in the past (https://www.theorsociety.com/ publications/magazines/impactmagazine/) A myriad more possibilities exist. In all, by utilising available tools and resources, anyone can explore and implement solutions to optimise their travel routes, ultimately saving time, reducing costs, and enhancing operational performance. The minimum viable working solution that works for many (most?) of us will not require too high a level of mathematical or programming sophistication. Why not try, then?
Matthew Howells is a PhD student in Operational Research at Cardiff University, previously achieving an MMORS degree, also at Cardiff University. Currently, his research applies simulation and pattern mining techniques to model the Trauma & Orthopaedic clinical pathway at Cardiff & Vale University Health Board. Feel free to contact at HowellsMA@cardiff.ac.uk.
FIGURE 1 TAYLOR SWIFT’S SUPPOSED SUB-OPTIMAL TOUR (A) AND THE ACTUAL OPTIMAL TOUR (B) AROUND LONDON FOR THE PAIR (MAP DATA FROM OPENSTREETMAP).
DARK MATTERS
Geoff Royston
Scotland Yard detective: Is there any other point to which you would wish to draw my attention?
Holmes: To the curious incident of the dog in the night-time.
Scotland Yard detective: The dog did nothing in the night-time.
Holmes: That was the curious incident. from “The Adventure of Silver Blaze”, by Sir Arthur Conan Doyle.
If you are hoping to read here something about cosmology, my apologies. This article is not about unobservable matter in the universe. It is about data we do not see, why unseen data matter, and what they can tell us.
MISSING IN ACTION
Holmesian detection stories apart, one of my favourite examples of the importance of considering what is absent comes from the 2nd World War. US planes were returning shot up, with holes scattered over various places – fuselage, wings, tails…. Protective armour would help but was too heavy to add everywhere. Which of the damaged areas should be reinforced? Abraham Wald, a Hungarian mathematician working in a wartime research group at Columbia University, came up with a counterintuitive response. He argued that the planes with the most incapacitating damage were the ones that were most often not coming back at all, so that sort of damage, e.g. to the engine housing, was rarely observed. So armour would be best used on the vulnerable parts of the planes which were showing least damage.
That example illustrates what is known as survivor bias, one of several ways that missing data can mislead. A wide
variety of issues arising from data that are not available or are not recognised are presented in a recent book which has stimulated and underpinned my piece in this issue of Impact. The book is Dark Data: Why What You Don’t Know Matters, authored by David Hand, Emeritus Professor of mathematics at Imperial College London. In coining the titular term he notes that “we don’t see such data, ……and yet they can have a major effect on our conclusions, decisions, and actions …. unless we are aware of the possibility that there’s something unknown lurking out there, the consequences can be disastrous, even fatal.”
VARIETIES OF THE UNKNOWN
In his book Hand discusses and classifies fifteen types of “dark data”. That’s too many to cover here, so I will focus on just two or three, using the categorisation brought to public attention in the often (unfairly) ridiculed words of Donald Rumsfeld, a former US Secretary of State for Defense:
“Reports that say that something hasn’t happened are always interesting to me, because as we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns—the ones we don’t know we don’t know.”
Donald Rumsfeld noted that it is the “unknown unknowns”, the things we not only don’t know but haven’t even thought of, that tend to be the most problematic.
Logically there is also a fourth category, “unknown knowns”. This has, somewhat confusingly, been defined in more than one way but my favourite is “things we think we know but in fact don’t.” Such “unknown knowns” can certainly lead to major problems - recall the saga of the Iraq war “weapons of mass destruction”.
These varieties of seen and unseen data can be visualised as occupying four quadrants of a matrix as below:
KNOWN UNKNOWNS
Known unknowns are a ubiquitous problem in surveys, for instance when some respondents cannot be contacted. Other examples include predictions of people’s likely behaviour based on observed behaviour of a selected group, e.g. predicting whether applicants for credit are likely to default, based on the default record of past customers. But past customers will have been selected as being of acceptably low risk, so we will be in the dark about what default levels of those not given credit would have been. Modelling can be useful in trying to fill such gaps (the 2000 Nobel Prize for Economics was awarded for work on modelling missing data).
Sometimes the darkness of data arises not from the observed but the observer. For example Hands notes that if you ride on London’s red buses you will know that more often than not they are packed with passengers. And yet data show that the average bus occupancy is just 17 people. How come? Simply that, by definition, more people are on, and so see, a full bus. An empty bus will have no passenger to report that it was empty. In the above cases the darkness of data was happenstance, but sometimes it can be deliberate. There are plenty of examples of open darkening (concealed darkening will be discussed later). Cryptography provides a prime area of intentionally darkened data. Paradoxically perhaps, intentionally darkening data can be useful in the search for knowledge – in what Hands calls “the strategic application of ignorance”. For example, to avoid bias, clinical trials often have an experimental and a control arm in which the subjects (and often also the researchers) are “blinded” as to which arm is which.
There are also data we may want to keep dark, e.g. personal information that we would not wish to be stored on databases, but may find it difficult to do so. Hands refers to this as our data shadow, “consisting of the traces we leave from sending emails or texts, tweeting, posting a comment on YouTube, swiping credit cards, using travel cards, making phone calls, updating a social media app, logging onto a computer or iPad, taking cash from an ATM, driving past a car license plate recognition camera, and so on endlessly, in often unsuspected ways.” Of course this can be useful to us but it can also assist others to impact on us in ways we would not wish; in the extreme this failure to keep data dark can lead to fraud such as identity theft.
UNKNOWN UNKNOWNS
One of the prominent examples of unknown unknowns discussed in David Hands’ book - the 1986 Challenger Space Shuttle disaster – happens also to have featured in an earlier article of mine in Impact (Spring 2019) on data visualisation. I will briefly reprise it here.
The Challenger launch was due on a particularly cold day and the engineers had voiced concerns about the vulnerability of the rocket booster sealing rings in cold weather. Data were presented to launch managers on the air temperatures when there had been damage to booster seals at launch. There was no apparent correlation - there had been a few seals damaged in cold weather launches but also a few in warm weather ones. A decision to launch was made, with as we now know, fatal results due to seal failure and leakage of burning gases.
What the launch managers had not been properly shown, and had not realised was an important omission, were the data on all launches, including those – the majority – where there was no appreciable seal damage, nearly all of which had been in warm weather. Including these “missing” data would have indicated that the seal failure rate rose dramatically at low temperatures – and so that a launch in colder weather than had ever been attempted before could well be catastrophic.
Another situation is where data are not only being deliberately hidden (or falsified) but the occurrence of darkening is itself concealed. Hands gives the examples of insider trading and, more generally, false accounting, which hinge on knowing something others don’t (asymmetric information) and/or hiding the true state of a company by concealing or distorting data.
FIGURE 1 VARIETIES OF SEEN AND UNSEEN DATA
DARK DEEDS?
Problems with dark data can crop up in all sorts of places. Not least in police investigations and courts of law. Miscarriages of justice have occurred, obviously, where relevant data have been deliberately withheld but also where such data have been overlooked. For example in investigating clusters of adverse clinical events that have raised suspicions about people, sometimes only events that occurred when a suspect was present have been considered – but getting a full picture to see if the adverse event rate associated with a suspect is higher than the average requires also looking at adverse events that occurred when the suspect was absent, and at occasions when the suspect was present but no adverse events occurred (a criticism of the handling of evidence in the trial of nurse Letby has been that a full picture of roster data was not presented, reminiscent of the problem with the Challenger seal failure data).
Of course, an association with a higher rate of adverse events is not in itself proof of a causal link. Other unknown or unconsidered background factors may have had an influence (this was a problem in the trial of the Dutch nurse Lucia de Berk, who was initially convicted for murdering patients on her watch but later exonerated not least because other relevant factors, such as missing roster data and errors in autopsies, were uncovered).
The Royal Statistical Society have drawn attention (in a report entitled Healthcare: serial killer or coincidence? –https://bit.ly/RSS-Data-Evidence-Decisions) to these and other difficulties of distinguishing, from administrative data
alone, when clusters of adverse events are observed, between coincidence and foul play.
DARKNESS VISIBLE?
Cosmologists may argue about the importance of dark matter, but for analysts and managers, dark data certainly do matter. They occur widely and can have serious consequences. So, what to do? Hands gives some tips on mitigation (apart obviously from trying to minimise incomplete data), which I would summarise as:
–Vigilance – being aware that data may be incomplete or inaccurate
–Understanding - what sort of data may be missing, and why?
–Analysis – what might be the effect of missing data? Is there evidence of gaps or distortions? Can this be adjusted for by modelling?
Like Sherlock Holmes, use what you do know about what you don’t know.
Dr Geoff Royston is a former President of the OR Society and a former Chair of the UK Government Operational Research Service. He was Head of Strategic Analysis and Operational Research in the Department of Health for England, where for almost two decades he was the professional lead for a large group of health analysts.
“I have been on a few of The OR Society’s courses, they have really enhanced my career prospects. I was able to obtain a promotion through demonstrating and implementing what I had learnt, these courses are highly recommended.” ~ John, Senior Operations Researcher, GORS
Access our full course calendar at: www.theorsociety.com/training
