Impact Magazine Autumn 2021 by Impact Magazine from The Operational Research Society

D R I V I N G I M P R O V E M E N T W I T H O P E R AT I O N A L R E S E A R C H A N D D E C I S I O N A N A LY T I C S

AUTUMN 2021

COVID-19 SUPPORT FOR THE UK’S OVERSEAS TERRITORIES Dstl analysis informed government decision making on medical provision

HIGHLIGHTING CHANGING TRAVEL PATTERNS DURING THE PANDEMIC

Supporting Government decision making with daily reports on patterns of national travel

ESTIMATING DELIVERY WINDOWS FOR ROYAL MAIL

Determining narrow time windows when delivering to 31 million households

What does ‘build back better’ look like for your business? The OR Society runs courses for operational researchers, analysts and data scientists to keep them up to date on the latest techniques, m ethods a nd ttechnologies. echnologies. methods and Send your staff on our year-round training events, optimised for contemporary work needs: • online and interactive • full of analytics insights • apply data skills right away • typically single-day courses

Invest in OR and analytics skills in 2022. ;NJ\ YMJ HTZWXJ QNXY MJWJ \\\ YMJTWXTHNJY^ HTR YWFNSNSL್್

E D I TO R I A L Operational Research (O.R.) makes an impact in all sectors of society, including Government. Indeed, the largest employer of O.R. people in the UK is the Government Operational Research Service (GORS) with over 1000 analysts, from senior Civil Servants to sandwich students. GORS supports policy-making, strategy and operations in many different Government departments. Articles in this issue of Impact show ways in which GORS analysts addressed two of the key problems that we face: the response to the Covid pandemic and climate change. The Department of Transport were concerned about the effect of the coronavirus travel restrictions in the national lockdown. As every mode of travel was affected, they wished to know what was happening in near real time: an novel issue. We can learn how daily reports on trip levels supported monitoring and decisions as the pandemic crisis unfolded. Another article tells how the UK Government’s Department for Business, Energy & Industrial Strategy (BEIS) developed two online energy and emissions calculators for creating pathways to net zero. This timely development allows you to explore options for how the UK reaches net zero emissions. Timely? The UK hosts COP26 in Glasgow in November (COP26) where there will be a push for greater international commitment in the fight against climate change. The Defence Science and Technology Laboratory (Dstl) is an executive agency of the Ministry of Defence (MOD) providing analytical expertise for the benefit of the nation and allies. We can read how Dstl supported the UK’s overseas territories during the covid-19 pandemic. Of course, there are many areas in the world that need support. Another article tells us how World Food Programme analysts used optimisation methods to help the Ministry of Health in Côte d’Ivoire improve medical supply chains. I hope you enjoy reading these, and the other articles, which show how O.R. and analytics have made an impact. Electronic copies of all issues are available at https:// issuu.com/orsimpact. For future issues of this free magazine, please subscribe at https:// www.theorsociety.com/impact/.

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 Tel: + 44 (0)121 233 9300, Fax: + 44 (0)121 233 0321 Email: email@theorsociety.com Secretary and General Manager: Gavin Blackett President: Edmund Burke Editor: Graham Rand g.rand@lancaster.ac.uk Associate Editor: James Bleach Print ISSN: 2058-802X Online ISSN: 2058-8038 www.tandfonline.com/timp Published by Taylor & Francis, an Informa business All Taylor and Francis Group journals are printed on paper from renewable sources by accredited partners.

Graham Rand

OPERATIONAL RESEARCH AND DECISION ANALYTICS Operational Research (O.R.) 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 – O.R. is an improvement science. For over 70 years, O.R. 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 O.R. label continues, though some prefer names such as business analysis, decision analysis, analytics or management science. Whatever the name, O.R. 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, O.R. 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 O.R. 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/. O.R. is the home to the science + art of problem solving.

CO N T E N T S 7

SUPPORTING THE UK’S OVERSEAS TERRITORIES DURING THE COVID-19 PANDEMIC Pete Bailey and Struan Millar describe how the Defence Science and Technology Laboratory (Dstl) conducted modelling and analysis to assess the likely scale of medical capacity required by several UK Overseas Territories and overseas military bases

4 Seen Elsewhere

Analytics making an impact 13 Supporting your efforts on

diversity Nicola Morrill shares with us several ways that O.R. can help organisations with their EDI work

ROYAL MAIL’S ESTIMATED DELIVERY WINDOW

RENAULT’S SUPPLY CHAIN AND MANUFACTURING

32 Universities making an impact

IMPROVING MEDICAL SUPPLY CHAINS IN CHALLENGING ENVIRONMENTS

43 Making an impact with Soft OR

Betty Schirrmeister describes how Royal Mail’s data science team delivered a model to provide customers with 2-hour delivery windows

Alain Nguyen tells us how the work of a centralised O.R. team has improved Renault’s supply chain and manufacturing operations

Sara Valiño, Wouter Smeenk and Koen Peters of the World Food Programme report the use of optimisation to improve the medical supply chains with the Ministry of Health in Côte d’Ivoire

NET ZERO EMISSIONS – FROM WHY TO HOW Bevan Freake describes two new online energy and emissions Calculators letting you explore options for how the UK reaches net zero emissions

SECRETS OF SUCCESS WITH LOAN PRICING OPTIMISATION Barry Honeycombe and Marc Drobe of FICO explain how loan pricing optimisation works and how it prompted a 29% increase in new sales for Home Credit Russia

Brief reports of two postgraduate student projects

Martin Parr describes the advantages of using Soft OR, illustrating his article with an example from UK military recruitment 47 Hotter

Geoff Royston interacts with Bill Gates’ book How to Avoid a Climate Disaster and offers a few comments from an analytical perspective ahead of COP26

CREATING NOVEL MOBILITY INSIGHTS IN A NATIONAL LOCKDOWN Jordan Low and Sam Rose tell how Government analysts produced daily reports on trip levels to support monitoring and decisions as the pandemic crisis unfolded

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 FIGHTING MULTI-DRUG RESISTANT TB

TDR, the Special Programme for Research and Training in Tropical Diseases, is a global programme of scientific collaboration that helps facilitate, support and influence efforts to combat diseases of poverty. In collaboration with the WHO Global Tuberculosis Programme and technical partners, TDR has developed an operational research package (dubbed ShORRT for Short all-Oral Regimens for Rifampicin-resistant Tuberculosis) to support the implementation of such drug regimens. The ShORRT initiative now involves and supports 25 countries worldwide. In the latest World Health Organization (WHO) guidelines on drug-resistant tuberculosis (DRTB) treatment, modifications to the recommended all-oral treatments for multidrug- and rifampicin-resistant (MDR/RR) TB and novel regimens for patients with extensively drug-resistant TB are encouraged under operational research conditions. Dr. Babawale Adekunle Victor, Programmatic Management of Drug Resistant Tuberculosis Unit, National Tuberculosis and Leprosy Control Programme, Nigeria has said “Operational Research is proving to be a tremendous tool for guiding our efforts to manage TB in Nigeria, especially in the context of COVID-19”.

DATA MATURITY OF THE NOT-FOR-PROFIT SECTOR

Following 15 months and thousands of people using Data Orchard’s Data

IMPACT © THE OR SOCIETY

Maturity Assessment Tool, they crunched the numbers to analyse what the tool’s data tells us about the current state of data maturity in the not-forprofit sector. The results have been published in State of the Sector - Data Maturity in the Not-for-Profit Sector 2020. See http://bit.ly/3hQXnPG. What the data tells us about not-forprofit data maturity, inter alia, is that the cost of data is huge, hidden, and often wasted, that most leaders don’t see the value of data, that there’s lots of counting but not enough meaningful analysis, and that most not-for-profits lack skills, responsibility and support around data.

DOMO STREAMLINES EMBEDDED ANALYTICS PLATFORM

In June 2021 Domo, a cloud-based analytics vendor founded in 2010 and based in Utah, unveiled the latest

version of its embedded analytics platform, adding new capabilities while updating others and consolidating all of them under a single umbrella. When first released in 2017, the embedded analytics platform was made up of three distinct offerings. Domo White Label enabled customers to develop their own analytics applications and market them to their own customers under their own brand. Domo Embed enabled customers to build and embed basic applications such as reports and dashboards. And Domo Publish enabled users to share their analytic assets both with others in their organizations and with outside customers. The updated version of Domo Everywhere combines the three offerings into a single experience so users can more easily and quickly build and distribute embedded applications. In addition, the new iteration of Domo Everywhere enables users to build, market and share more advanced analytics applications— including machine learning models and other assets built using augmented intelligence capabilities—than they could using previous iterations of Domo Everywhere.

PEST CONTROL SIMULATION

Researchers at North Carolina State University have developed a simulation model to predict when and where pests and diseases will attack crops or forests. Their computer modelling system works by combining information on climate conditions suitable for spread of a certain disease or pest with data

on where cases have been recorded, the reproductive rate of the pathogen or pest and how it moves in the environment. Over time, the model improves as natural resource managers add data they gather from the field. This repeated feedback with new data helps the forecasting system get better at predicting future spread. “We have a tool that can be put into the hands of a non-technical user to learn about disease dynamics and management, and how management decisions will affect spread in the future,” lead author Chris Jones said. A report is available in the journal Frontiers in Ecology and the Environment at https://doi. org/10.1002/fee.2357.

IT’S THE CONFERENCE SEASON

There has been a flurry of conferences recently. Not the political party conferences, but O.R. conferences: first the regional (EURO) in July, then the international (IFORS) in August and finally the national (OR Society) in September. The first two were hybrid, but all could be attended virtually. There were many examples presented of O.R./analytics making an impact, including several UK-based contributions.

The finalists in the EURO Excellence in Practice Award included Nav Mustafee and John Powell’s work, NHSQuicker, which was featured in the Spring 2021 issue of Impact and the winning entry of Gilberto Montibeller and Alberto Franco. Their work on facilitated decision analysis for emerging

health threats enhanced the quality of health experts’ recommendations to the UK Department for Environment, Food and Rural Affairs leadership in the prioritisation of animal and human emerging health threats and informed new international standards for the Food Standards joint programmes of the Food and Agriculture Organization of the United Nations and the World Health Organization. The finalists in the IFORS Prize for O.R. in Development competition included Paola Scaparra, who presented the work of a team of researchers concerned to minimize flood impact on the road infrastructure in Vietnam. Their work, funded by the Global Challenge Research Fund through the British Academy, identified costefficient measures to mitigate the impacts of urban flooding and created an optimisation tool embedded into a Decision Support System which was applied to generate a 20-year plan of investments. More than half of the total reduction in congestion and damage was found to be achievable with less than 25% of available budgets. The project has catalysed interest in Vietnam and in neighbouring countries to develop capacity in O.R. to address development challenges.

At the OR Society conference, Lucy Morgan spoke about modelling of waiting lists for diagnosis and treatment of chronic heart failure following the Covid-19 pandemic, which has disrupted access to health services for patients with non-Covid-19 conditions due, not just to increased demand, but also as a result of the associated lockdowns and social distancing measures. A discrete event simulation model was built to

describe the impact of the pandemic and associated societal lockdowns on access to diagnosis procedures. Clinicians at Leeds Royal Infirmary were involved and it is hoped that the interventions found to reduce waiting lists will be put into practice by them. She also reported on work carried out with the Liverpool School of Tropical Medicine to investigate the impact of decentralised treatment pathways for multidrug-resistant TB (MDR-TB) patients in Ethiopia. Reducing costs incurred by MDR-TB patients is a priority of Ethiopia’s National Tuberculosis Programme. Lucy presented a simulation model representing MDRTB patient treatment pathways, which was used to investigate the impact of patient-centred and hybrid treatment strategies on patient costs during treatment and compare this to costs incurred under the current centralised treatment strategy in Ethiopia.

EDGE ANALYTICS

As Impact makes clear, analytical insights can be key for organizations to make decisions. Real-time data requires realtime analytics for better decision-making processes. And with growing data volumes from various applications and locations, they need to be effectively managed and analysed. Edge analytics deals with this. Edge analytics ensures automatic, real-time analytical computation of data without sending the data to the centralized data server for analytics or storage. With increasing numbers of connected IoT devices, businesses are making use of edge analytics and leveraging the solutions offered by cloud providers. Using next-generation data

IMPACT | AUTUMN 2021

DIRECT FLIGHTS SAVE LIVES

New research in the INFORMS journal Management Science tackles the transportation part of the kidney transplantation problem (see https:// doi.org/10.1287/mnsc.2021.4103). “Airline transportation limits the flexibility of organ transplantation, but new, more direct airline routes can increase the number of kidneys shared between regions connected by these routes by more than 7%,” said Tinglong Dai of Johns Hopkins University. “Operations Research and analytics is trying to save lives and allow kidneys to

IMPACT | AUTUMN 2021

be more readily available to those who need them when they need them. Too often, viable kidneys are wasted because they can’t reach a patient in time.” The research identifies how new airline routes can provide the necessary efficient airline transportation needed for the time-sensitive nature of kidney transplantation and reduce the number of viable kidneys being wasted because they didn’t reach the patient in time. The authors analysed U.S. airline transportation and kidney transplantation datasets and used the data to track the evolution of airline routes connecting all U.S. airports. Then they looked at kidney transplants between donors and recipients connected by these airports. “Transportation plays a major role in providing patients with available donations, if new airline routes can increase the volume of shared kidneys by 7.3%, think of how many lives could be saved,” continued Dai. “We also find the post-transplant survival rate remains largely unchanged. It’s a step forward in organ donations thanks to O.R. and analytics”.

E2OPEN TO MODERNIZE TESCO’S GLOBAL LOGISTICS MANAGEMENT

E2open, a leading network-based provider of 100% cloud-based, mission-critical, end-to-end supply chain management software, have announced that they are working with Tesco PLC. Tesco will replace its legacy transport management system and utilize the E2open platform for increased visibility, transport execution, invoice processes and supply collaboration. E2open’s network and applications manage all tiers of production, inventory, logistics, global trade and channel activity from a single platform in the cloud. Tesco will leverage this networked-platform to manage their large global community of suppliers and logistics partners as they serve nearly 4,000 stores around the world. “With the complexity of today’s supply chain there is simply no room for error,” said Michael Farlekas, president and chief executive officer at E2open. “From procurement to logistics to payment, our expansive, integrated platform will provide Tesco with the highest level of visibility into their global supply chain, reducing costs and improving efficiency while providing greater oversight to help avoid consumer shortages.” E2open will help Tesco improve efficiency through standardizing disparate processes, resulting in better utilization, reduction in detention and demurrage fees, and improve overall stock management. E2open’s platform will also facilitate ordering and ensure that Tesco reaches a better fill rate of its shipping containers and other assets.

analytical tools, cloud providers like Google, IBM, and Amazon Web Services (AWS) are allowing businesses to unleash greater value from their data. Despite the challenges of the pandemic, the global market for edge analytics is estimated to reach US$25.4 billion by 2026 according to an industry report by Global Industry Analytics Inc. Edge analysis solutions are expected to grow the fastest with more businesses now realizing the advantages they can have by leveraging data closer to the edge. The world’s leading economies, the US and China, are expected to adapt edge analytics solutions the most. In addition to several large manufacturers and telecommunication companies, more small and mid-sized companies in the US are contributing to the growth in edge analytics. Smart cities that are extensively fitted with sensors tend to rely heavily on edge analytics for the best experience and service. More at: http://bit.ly/EdgeAnalytics1

SUPPORTING THE UK’S OV E R S E A S T E R R I TO R I E S D U R I N G T H E COV I D -1 9 PA N D E M I C

PETE BAILEY AND STRUAN MILLAR

DURING THE EARLY PHASES OF THE COVID-19 PANDEMIC, the Government needed to consider not just the needs of the UK, but also of its Overseas Territories and our military personnel deployed abroad. Ministry of Defence (MOD) support to these overseas locations was known as Operation BROADSHARE. The Defence Science and Technology Laboratory (Dstl) was asked to provide Operational Research (referred to as Operational Analysis within MOD) to assess the likely numbers and implications of COVID-19 patients

presenting at medical facilities (such as hospitals) in these locations. This article will describe the modelling and analyses conducted by Dstl, covering the process from the collection of data, the tailoring of extant methods to better represent the problem of COVID-19, the types of metrics assessed and the impact of the work conducted. This included the deployment of RFA Argus (pictured here) to provide a medical response capability prior to the hurricane season as well as in support of the pandemic response.

OPERATIONS RESCRIPT AND BROADSHARE

For MOD, the pandemic was expected to affect operations overseas and at home, military training, overseas bases and normal departmental business as Defence observed government guidance on working from home where possible, social distancing and essential movement. The Department’s response, therefore, was not only the contribution to the Government’s effort to mitigate the effects of COVID-19 but encompassed its internal resilience activity. Two operations were activated: • Op RESCRIPT to deliver COVID-19 activity in the UK and Crown Dependencies, including Departmental business continuity and resilience, Military Aid to the Civil Authorities, and other support to government departments and local responders. • Op BROADSHARE delivering operations and activity outside the UK, including the overseas bases, the Defence overseas network, overseas training teams, support to the Overseas Territories, and government overseas operational and business continuity activity. MEDICAL MODELLING WITHIN MOD

The MOD has a duty of care to its service personnel, plus a remit to provide humanitarian aid to those in need. Understanding potential numbers of casualties, alongside the effectiveness of current defence medical capabilities in different operating environments, is vital to ensuring our deployed forces have access to the right medical facilities at the right time. This improves treatment times and, ultimately, survival rates.

IMPACT | AUTUMN 2021

The Ministry of Defence has a duty of care to its service personnel, plus a remit to provide humanitarian aid to those in need Dstl developed the ZEUS medical model in 2018 with support from Defence Medical Services (DMS). It is a decision support tool, used to assess medical plans (typically referred to as laydowns) at the operational level, in order to identify potential bottlenecks and limitations within the deployed medical system. This medical system is technically referred to as the Operational Patient Care Pathway and is detailed in Allied Joint Doctrine for Medical Support (AJP-4.10 Edition C Version 1, dated Sept 2019). ZEUS tracks patients (on a minute-by-minute basis) from the Point of Injury (or Illness – for the purposes of this article, it will be assumed that the term injury will include illness), through movement to and treatment at a medical treatment facility (the larger, more established of these can be considered to be small hospitals) to release (such as by returning to their original unit or being evacuated back to facilities within the UK). It was designed to assess questions such as: • Do we have the correct mix of medical capabilities to support the operation in question? • Are those capabilities appropriately located and do they have sufficient capacity to meet the clinical timelines? The clinical timelines provide planning guidance for the receipt of different levels of treatment, for example, the UK aims to provide advanced resuscitation and pre-hospital emergency care within one hour of injury or onset of acute symptoms. • What are the resource (e.g. personnel and equipment) and logistic (e.g.

medications, blood, oxygen, Personal Protective Equipment [PPE]) requirements? As such it is aligned to the NATO Principles and Policies of Medical Support (MC 0326/4, August 2018): ‘The aim of operational medical support is to ensure that every casualty gets the right treatment in a timely manner and at an appropriate facility.’ ZEUS was developed primarily in R and uses Microsoft Excel to manage the user input and the R Shiny package to provide visualisations of the outputs (Figure 1). Early development of the model requirement identified a number of factors, which drove the design, such as: • The need to represent, and react to, surges in demand on the medical laydown led to the model being designed to simulate random, distinct injury events. For example, the medical system may be able to treat 60 patients distributed equally over a 24-hour period, but how might it cope if many of those patients arrived simultaneously? • The need to represent future concepts of operation for the medical force (which may not yet exist) meant that the model needed the ability to represent these via data rather than as fixed structures. The deployed medical laydown will typically include facilities that expand and/or contract over time and those facilities may move at various points in the operation. • The requirement to recreate all of the decisions made within the model to enable the verification of its operation. In order to mesh the requirement for the inclusion of random elements with the requirement for transparency,

the core simulation is deterministic but includes a pre-processor that generates multiple patient streams, each of which details the number of patients, the timing of their injury and their individual medical requirements. A replication (i.e. an individual model run) is hence the processing of one of these patient streams by the core simulation. The model itself has two modes of operations: • Unconstrained, where the model assumes that sufficient resources are available to meet the patient demand, and • Constrained, in which the model will block the onward movement or treatment of patients, for example due to the lack of beds or ambulances. TAILORING ZEUS FOR COVID-19

In March 2020, Dstl was requested to support Defence Medical Services’ contribution to Op BROADSHARE with analysis to aid the understanding and assessment of future responses and potential

medical provision for a number of overseas bases, with a focus on: • The likely inflow of COVID-19 cases (categorised by severity of infection) presenting in each of those locations. • The numbers of Low and High Dependency beds potentially required in each location, where: ∘ Low Dependency represents a General Ward bed, and ∘ High Dependency represents a Critical Care bed, which may be equipped to provide mechanical ventilation. ZEUS had previously been identified as a potential model to address these issues but it required some modification of its input data to adapt it to better represent COVID-19, in particular how the disease affected different age groups and the escalating treatment requirements for increasing levels of severity. Data collection focussed on extant sources, including open source, supplemented with Subject Matter Expert judgement, and included:

• The ‘Population At Risk’ in each location (stratified into various age bands), which was provided by the analysis requestor. • Severity of infection was based upon papers from Imperial College London and the World Health Organisation. • Rates of infection (varying by week) were based upon the Scientific Advisory Group for Emergencies (SAGE) Realistic Worst-Case scenario or Defence Medical Services modelling, supported by Defence Statistics (Health) data on reported cases. • Medical treatment pathways, which detail the types of treatment required and the durations of those treatments, for different categories of patients, were developed in conjunction with Defence Clinical Advisors (who also have roles with the NHS). The analysis was conducted at pace and the outputs were presented less than 72 hours after the initial request was received. This resulted in a number of actions, including the deployment of medical staff and other resources to the Falkland Islands to supplement the facilities in that location. Figure 2 shows an oxygen tank being loaded prior to delivery to the King Edward Memorial Hospital in the Falkland Islands. This initial analysis was well received, and follow-on work was requested, both to investigate the initial analysis in more depth and to assess other locations. OVERSEAS TERRITORIES (OT)

The UK’s Overseas Territories are fourteen territories that do not form part of the United Kingdom itself, but which retain a constitutional and historical link with the United Kingdom. Most of the

IMPACT | AUTUMN 2021

medical facilities was sufficiently compelling that SJFHQ advised that the extant lockdown measures be maintained, which they were.

INVESTIGATING RISK DUE TO LACK OF CAPABILITY

FIGURE 2 OXYGEN TANK BEING LOADED PRIOR TO DELIVERY TO THE KING EDWARD MEMORIAL HOSPITAL IN THE FALKLAND ISLANDS

permanently inhabited territories are internally self-governing, with the UK retaining responsibility for defence and foreign relations.

This resulted in a number of actions, including the deployment of medical staff and other resources to the Falkland Islands to supplement the facilities in that location Standing Joint Force Headquarters (SJFHQ), leading MOD’s response, worked closely with the Overseas Territories to ensure they received the help they needed to manage outbreaks of COVID-19. Having been informed of Dstl’s analysis in support of the overseas bases, in early April they requested analytical support to assess the healthcare capabilities of the Caribbean territories: • Anguilla, which has a population of approximately 15,000 • Bermuda, approximate population 64,000 • British Virgin Islands, approximate population 30,000

IMPACT | AUTUMN 2021

• Cayman Islands, approximate population 66,000 • Montserrat, approximate population 5,000 • Turks & Caicos Islands, approximate population 37,000

INVESTIGATING SCALE OF CAPABILITY REQUIRED

As for the analysis of the bases, the initial request was for an assessment of the likely scale of medical capacity required within the territories, assuming different potential rates of infection (based upon updated data and modelling from SAGE, DMS and Defence Statistics). At the time, a number of the territories were considering whether to lift their local lockdown restrictions and SJFHQ wished to understand the potential impacts of this upon the numbers of patients presenting. Whilst the majority of the cases may have been expected to develop mild (or no) symptoms, a sufficiently large number may have been expected to require hospitalisation. The knowledge generated (via the ZEUS modelling) of the potential demand upon the

For the second phase of the analysis, we received up-to-date information on the current medical capabilities and capacities within each of the territories. As ZEUS models individual patients, we were able to use its Constrained mode to calculate how many of them had their treatment delayed waiting for a specific resource to become available (for example, a Ventilated Critical Care bed) and what was the duration of each of those delays. Whilst we could not say with any certainty what the specific medical implications of those delays in treatment could have been (in this case, what was the increased likelihood of mortality?), we could infer that we should be aiming to minimise both metrics should any delays occur. For a number of the territories, whilst the majority of the patients requiring Ventilated Critical Care beds may have been expected to suffer no delays in receiving necessary treatment, a number may have faced significant (and potentially unsurvivable) delays. Using this information allowed SJFHQ to make judgments on the balance of medical risk across all of the territories and identify which had urgent requirements for supplementary resources and supplies. IMPACT

The various analysis tasks undertaken during this period highlighted the value of building flexibility into the development of our models. The original requirement for the ZEUS model was based upon it being used to support the development of

medical plans for deployed military forces where many of the casualties would come as the result of combat operations or their supporting activities. However, by keeping its functionality as generic as possible, we were rapidly able to adapt the input data to represent the spread of the virus through civilian populations and to model their treatment whilst the knowledge of that was also being developed. The range of analysis conducted during this period had significant and, in some cases, near immediate impact. For the initial analysis of the overseas bases, the analysis was requested on a Monday afternoon, the outputs fed into the senior decision makers meeting on the following Thursday morning, the medical team was tasked on the Saturday and they arrived with the additional equipment required to increase the Critical Care capacity within the Falkland Islands by the next Tuesday. The impact of the analysis for the Caribbean Overseas Territories was not quite so quick but still rapid – the delivery of the equipment and supplies required to significantly increase the medical capacity within the Turks & Caicos Islands took place within 2 weeks of the analysis reporting at the end of April 2020. In addition, RFA Argus (Figure 3) sailed to the Caribbean in preparation for the hurricane season and to support the response to the COVID-19 pandemic if required. She has a fully equipped 100-bed medical complex on board, which includes an emergency department and resuscitation facilities.

FIGURE 3 RFA ARGUS (A PRIMARY CASUALTY RECEIVING SHIP) IN THE CARIBBEAN SEA, SUMMER 2020

Major General R T H Jones, Standing Joint Force Commander noted that ‘. . . the modelling drove many of the key planning assumptions, providing clear targets and timelines for individual OT healthcare capacity and capability requirements (for example, ventilator numbers, staffing levels and PPE requirements). . . ’.

the modelling drove many of the key planning assumptions, providing clear targets and timelines for individual Overseas Territories’ healthcare capacity and capability requirements

Pete Bailey is a Principal Analyst within Dstl’s Exploration Division. He was the technical lead for the analysis in support of Op BROADSHARE, having

previously led the development of the ZEUS model. Struan Millar is a Senior Engineer within Dstl’s Platform Systems Division. He was seconded to SJFHQ during Op BROADSHARE, as a deployable operational analyst, to provide real-time decision support and enable effective communications back to the wider science and technology community. © Crown copyright (2021), Dstl. This material is licensed under the terms of the Open Government Licence except where otherwise stated. To view this licence, visit http://www.nationalarchives. gov.uk/doc/open-governmentlicence/version/3 or write to the Information Policy Team, The National Archives, Kew, London TW9 4DU, or email: psi@nationalarchives.gov.uk

IMPACT | AUTUMN 2021

S U P P O R T I N G YO U R EFFORTS ON DIVERSITY Nicola Morrill

Not everything that counts can be counted and not everything that can be counted counts. Albert Einstein

Numerous studies have shown the benefits of a diverse, equitable workplace for business performance, innovation, customer loyalty, and employee engagement. Most, if not all, organisations and institutions, no matter their size, will have activity underway related to Diversity. The discipline of Operational Research (O.R.) and Analytics can support organisational efforts in Equity, Equality, Diversity & Inclusion (EDI) in a number of ways. This column briefly outlines several ways that O.R. can help with a range of areas organisations are likely to be covering as part of their EDI work, focussing on a number of common issues.

The Strategic O.R. area has several methods that are of use to these activities. In 2011, a special edition of the Journal of Operational Research Society (Volume 62, Issue 5) on O.R.’s contribution to supporting strategy provided an insight into a range of these. One “soft” O.R. method that lends itself very well to helping create your EDI strategy and also in supporting the design of the programme of work is Causal Mapping, often referred to as Strategy Mapping. A Causal Map is hierarchical in structure (linking means to ends) and built with a focus on achieving goals. The process of creating the maps is ideally a group process and this in itself will add lots of value to a collective understanding of goals around EDI, what is required to achieve these and some of the potential challenges around this. Figure 1 is an example of a causal map (taken from the Journal of the Operational Research Society – see http://bit.ly/ CausalMap) about carrying out a successful piece of work for a client. The map concept at the bottom of an arrow has an impact (positive or negative) on the concept at the head of an arrow. So, for instance, one way to “alleviate client lack of credibility/trust in a method” is to “resolve the client’s lack of familiarity” (with the method). Another way is to “carry out successful interventions”. The method could be used to build a similar picture but based around how to achieve the EDI goals of an organisation. For instance, a desire to create an inclusive organisation may require the barriers to inclusion to

AGREEING AND DELIVERING YOUR EDI AMBITION

An important place to start on your journey is to determine your EDI ambition and agree a strategy and programme of work to achieve this. All areas are important but to have impact, focus is needed.

FIGURE 1 A CAUSAL MAP

IMPACT © THE OR SOCIETY

be understood and there are many ways this could be achieved. Creating the causal map will help understand the different activities as well as some of the richer context around your organisation and what may or may not work.

UNDERSTANDING THE CURRENT PICTURE

A key part of developing your strategy and programme of work is to understand your current picture and this can be achieved in a number of ways, focussing on different aspects of understanding and taking different approaches to do this.

Creative Approach

Soft Systems Methodology (SSM) is a powerful tool to gain broad insight into how EDI is working (or not) in your organisation in comparison to how it should ideally be working. There are many aspects to SSM and a particularly accessible part is something called a Rich Picture. This is a drawing, usually produced in a group setting, that enables a view of a situation to be shared. It embraces complexity and includes relationships, feelings and anything else that the group believe may be relevant to the situation being explored. Rich Pictures are a pretty ideal tool to understand the current picture within your organisation or team and, importantly, it includes softer factors too. A picture really does paint a thousand words! Figure 2 is an example drawing about planning for organisational success. While this includes business processes and activities it also highlights how different parts of the organisation may be feeling. This link, https://bit.ly/

RichPictures, provides a short overview of the value of Rich Pictures.

Picture through numbers

A key part of your EDI picture is to understand how diverse your organisation is and, in order to do this, data is required. The use of data could focus on descriptive statistics to provide a snapshot view of diversity. For instance, counting the number of females and, with intersectionality in mind, also the number of disabled females within different areas of an organisation as well as overall. EDI best practise suggests capturing trends on the data of interest, rather than acting on a snapshot view. One step further is to capture data around the processes within the organisation, such as recruitment and promotion. This can be used to pinpoint where, if at all, there may be a challenge related to onboarding or promoting diverse talent. O.R. as a discipline is able to advise on what it may be useful to collect – not always what is easiest to count or, indeed, numbers. Where data (numeric and non-numeric) is stored in lots of different areas, O.R. can help with pulling this all together in an efficient way. For larger organisations, the challenge of pulling data together can be quite onerous and, again, something that O.R. is well versed in doing.

ANTICIPATING CHANGES AND UNDERSTANDING WHAT SUPPORT WORKS ‘BEST’

The analytics side of O.R. has many methods available to support organisations getting upstream of the diversity challenges in their organisation by anticipating where issues may arise. For instance, based on recent trends, what area of the workforce are likely to leave and how will this impact diversity? Figure 3 (see http://bit.ly/DashBoard-Example) uses Machine Learning techniques to explore who is likely to leave an organisation and presents the outputs in a dashboard. O.R. also has methods that will help understand what measures are most likely to reduce the likelihood of a particular demographic leaving.

MONITORING PROGRESS

FIGURE 2 A RICH PICTURE

IMPACT | AUTUMN 2021

Monitoring progress against your EDI strategy is important and I think the following quote from an article written by Siri Chilazi and Iris Bohnet (http://bit.ly/ ChilaziBohnet) sums it up well: “Manage DEI in exactly

WANT TO LEARN MORE?

The OR Society runs training courses on much of the above if you want to bolster your inhouse team. For instance, training in methods such as forecasting, multivariate statistics, data science and systems modelling. A number of other training establishments and professional societies also provide training across some of these areas too.

SHAPING MY NEXT PIECE FIGURE 3 AN EXAMPLE OF A DASHBOARD

the same rigorous and data-driven way you manage the rest of your business. Achieving DEI objectives requires no more and no less than the use of the same planning, feedback, and accountability processes that are deployed to reach targets in sales, product development, and budgeting. Data drives targeted action and creates accountability in these domains, and so it should in DEI as well.” And O.R. as a discipline has much to offer!

If there is something, related to O.R., that you would like me to consider for my next ‘column’ in Impact then please get in touch. I’ll be pondering what to write about over the coming months…. The goal is to share the discipline with users/potential users of O.R. by highlighting how it could support ‘business’ challenges they may be facing. Nicola Morrill is a Systems Thinking Consultant at Dstl, a certified coach and the current Diversity Champion of the O.R. Society. She writes in a private capacity – all views expressed are her own and all examples are available in the open domain. You can contact her on Nicola.impact@gmail.com

IMPACT | AUTUMN 2021

R OYA L M A I L’ S E S T I M AT E D D E L I V E RY W I N D OW BETTY SCHIRRMEISTER

IN 2019 ROYAL MAIL LAUNCHED ITS ENHANCED ESTIMATED DELIVERY WINDOWS providing an increasing number of customers with 2-hour delivery windows (see https://www.youtube.com/ watch?v=RpBsNQxBtU4). The model behind this high performing product was built by Royal Mail’s data science team and has been successfully put in production thanks to great collaboration and communication across the business. Royal Mail is a business with history. It can trace its origins to Henry VIII, followed by over 500 years during

which it has influenced postal services around the world. In 2015, state ownership came to an end and Royal Mail became fully privatised. Today Royal Mail is one of the largest employers in the UK, providing one in every 194 jobs, with 140000 employees, 90000 of which are postmen and women. The service Royal Mail provides today still follows the “one-price-goes-everywhere” universal service obligation on a range of letters and parcels going to 31 million households. This poses a tough challenge to Royal Mail in order to stay

considering the variability of workload delivering to 31 million households, an algorithmic approach to predicting doorstep delivery time slots was required

The parcel business is becoming more important with revenue increasing year on year. This trend is accompanied by a shift in customer needs, such as the wish for trusted updates regarding deliveries. In April 2019, Royal Mail introduced higher precision estimated delivery windows, with time frames as narrow as two hours. “Notifying customers of their expected delivery time the day before we deliver is just one of the ways we are helping to make our customers’ lives easier. Coupled with a shorter estimated delivery time, of as little as two hours, creates a step change in convenience for online shoppers. … This latest innovation is part of our major investment in changes that increase convenience for our customers and their recipients.” (Royal Mail spokesperson) The data science challenge we were facing was how to provide a delivery window estimate for customers while not being able to track parcels, as seen in other delivery companies who follow a delivery manifest. In Royal Mail, parcels are part of a broader traffic mix of which a large portion is not ‘digitally visible’. In addition, postmen and women currently still deliver parcels as part of their normal route, which includes a variety of untracked items, such as letters and

hence makes it difficult to evaluate workload beforehand. Additionally, postmen may switch routes and approach journeys in different sequences or at different starting times, again depending on workload. Consequently, considering the variability of workload delivering to 31 million households, an algorithmic approach to predicting doorstep delivery time slots was required. Applying cloud solutions enabling clustering and filtering of a vast amount of unlabelled GPS data in combination with regression techniques and an additional filtering based on confidence levels, has enabled us to nevertheless provide reliable delivery windows to the vast majority of our customers. This work was heavily reliant on cross functional solutions, keeping the business in the loop about progress along the way.

INNOVATION NEEDS DATA SCIENCE

At Royal Mail we have placed big data and data science at the core of our business growth, aiming to be increasingly data driven in our decision

making. It is applied data science, which enabled us to provide estimated delivery windows, originally four hours wide, and after a cycle of algorithm improvement reducing the window sizes to estimates as narrow as two hours. Royal Mail’s data science team is working on a range of projects that either strengthen key services or generate data-driven products for Royal Mail. As for the estimated delivery windows, the model is taking into account a set of features ranging from weekdays, time series to geo-locations. The model will generate a confidencevalue for each address. Based on these values addresses will receive an estimated delivery windows, for which we can guarantee an exceptional level of accuracy and coverage to our customers. Additionally, performance of the algorithm can be easily accessed via dashboards and is continuously monitored. However, implementing great data science work in a business of Royal Mail’s size can be challenging and relies on multiple factors: (1) advanced data acquisition, storage and maintenance, (2) communication across the business © Royal Mail

competitive in a market that has, since 2006, been increasingly filled with other postal providers.

IMPACT | AUTUMN 2021

and (3) coordinated cross-functionalwork. Looking elsewhere, as few as 13%, 1 in 8, data science projects make it into production according to Deborah Leff, CTO for data science and AI at IBM. In her opinion, the main reasons for this are (i) missing access to data, (ii) a lack of communication, and (iii) a poor collaboration across teams (see http:// bit.ly/DeborahLeff ).

BIG DATA

Royal Mail’s operations are generating raw GPS data via delivery routes every day with GPS coordinates fed into our systems every few seconds. These data are being made accessible, cleaned and filtered using cloud services, including tools such as Hadoop, Spark and GCP. Additionally, we have access to a well-maintained data warehouse, with historic data of the past years. Within Technology, the data and engineering team (D&E) combines data analytics, data science and data engineering to facilitate the use of our data for innovative products within our company. With easy access to data of good quality, we were able to combine months of filtered GPS data with years of historic tracked parcel data to create a solid, if still quite noisy, training dataset for the estimated delivery window (EDW) algorithm. Based on patterns observed in this historic training dataset, our algorithm could start building predictions. Having access to enough good quality data for a model to train on is key for the start of a successful data science project.

Having access to enough good quality data for a model to train on is key for the start of a successful data science project 18

IMPACT | AUTUMN 2021

BUSINESS UPTAKE

Good communication with the business, building up trust with your stakeholders is key on the journey to delivering a data product, seeing it run live in the business, and consequently being used as a ‘business as usual’ tool on a daily basis. Our data scientists have kept up a continuous communication stream with stakeholders, and other teams involved in the EDW production pipeline, explaining the science behind the model and instilling confidence in our ways of working. To guide the business through our development phase we have generated an interactive dashboard where parameters can be tuned and model outputs directly influenced by stakeholders. Providing stakeholders with a more active role during the model building process while guiding their decisions around implementing and using results has contributed towards enabling trust in the data science work and confidence in our solution. We believe that data science should not be a black box that has to be trusted blindly.

We believe that data science should not be a black box that has to be trusted blindly

CROSS-FUNCTIONAL TEAMWORK

Data science work is but one step towards implementing data driven products within a business. The ability to communicate and collaborate with different teams is important. Finding successful, agile ways of working that can be applied across teams are essential for a successful product delivery. For the estimated delivery window, we have paired data scientists and data engineers to help translate the science

into an application that can be easily used by other parts of the business. While some data scientists have worked on algorithm improvements, others have worked closely with data engineers to generate clean, fully tested code and improve the application design.

the estimated delivery window development for Royal Mail parcels has been a data science success

A SUCCESS STORY

As a result of the Covid pandemic, historical delivery patterns are currently less reliable for predicting future delivery patterns. This is due to an exceptional rise in parcel usage, significant reductions in letters and new working methods that prevent our delivery staff sharing a van. Nevertheless, the estimated delivery window development for Royal Mail parcels has been a data science success, which, aside from a great data science team, has been based on (1) excellent data quality and quantity made available in our infrastructure to work with big data, (2) continuous communication across the business, and (3) close collaboration across different teams, particularly within technology. Dr. Bettina E. Schirrmeister is Principal Data Scientist at Royal Mail. She has completed a PhD in Computational Biology at the University of Zurich, worked as a senior research fellow at the University of Bristol and joined Royal Mail as a data scientist in 2016. Since her start at Royal Mail, she has led and delivered various data science projects around forecasting and predictive analytics, as well as schedule optimisation and is leading data product innovation for Royal Mail.

R E N AU LT ’ S S U P P LY C H A I N A N D M A N U FAC T U R I N G ALAIN NGUYEN

RENAULT’S SUPPLY CHAIN AND MANUFACTURING span over 40 plants in 17 countries, 3500 supplier sites, 5000 delivery destinations. It manages 300,000 parts references and uses daily 3500 trucks to supply parts to factories, and 2000 trucks to deliver cars. The car seen above is Renault’s best seller first compact SUV.

O.R. has been applied to Renault’s supply chain and manufacturing for more than 20 years

O.R. has been applied to Renault’s supply chain and manufacturing for more than 20 years. During this time, a wide variety of issues have been addressed. Often based on in-house developments, O.R. tools help the business make breakthroughs in terms of cost performance. These tools are designed by a centralised O.R. team from the new digital transformation department, located in Paris. The team is composed of four permanent staff, one PhD and one intern. We focus solely

on optimisation engines, while connections to databases and graphical user interface are dealt with by IT departments. In order to take into account thoroughly the business rules in our algorithms, we develop a deep knowledge of the supply chain and manufacturing business. This dual mindset (business and O.R.) is critical to elicit the, often implicit, business rules, and to convert them into optimisation constraints and objectives. Another important asset is that, since we all come initially from IT departments, we also have a good knowledge of Renault’s information systems and IT infrastructure, which is key to a fast delivery into production of our algorithms. We detail hereafter a few cases that have been addressed.

the trucks, as shown in Figure 2. In Figure 2, the first truck from day D + 1 has been cancelled thanks to the anticipation of all its orders from D + 1 to D. The second truck of day D + 2 has been cancelled thanks to the move of all its content to the first truck of D + 2. The arrows represent the anticipation of orders. The delivery orders sent to suppliers define the quantity to be delivered but also the date-timed truck in which the parts should be loaded.

FIGURE 1 LOADING PLAN ON THE FLOOR

optimisation method which beat all bin packing specialists’ algorithms. It was a great example of partnership with researchers and delivered results beyond our wildest expectations.

CONTAINER LOADING

IMPACT | AUTUMN 2021

TRUCK LOADING AND DELIVERIES ANTICIPATION

Everyday 3500 trucks deliver parts to our plants worldwide. The filling rate of these trucks is critical, since the inbound transportation yearly budget is over several hundred millions of euros. The specificity of Renault is that transportation is synchronised with production control. Deliveries can be anticipated two to three days beforehand in order to better fill

We experimented with optimisation software from major suppliers, but all failed to cope with our very specific process and huge data volume (7-weeks horizon). We therefore implemented a heuristics-based algorithm to manage deliveries anticipation and coupled it with a 2D placement method based on best-fit strategies. To cope with the huge data volume, we run the optimisation in Google Cloud Platform, with a dedicated server per plant and with lots of tuning to improve performance, all of which could not be done by standard commercially available software. The tool was a major success in terms of cost benefits, though quite disruptive for suppliers. For the very first time, © Renault

Renault have to send parts from Europe to factories worldwide (Brazil, India, China, Russia). In addition to the daily loading of containers, logistics platforms must estimate every week the number of containers needed for the following weeks. Therefore, a loading tool is at the heart of the activity of logistics platforms. The minimisation of the number of shipped containers is critical since a gain of one cubic metre per container can generate huge annual savings for logistics platforms. A large quantity of small/ medium/large items have to be packed into containers of different sizes, as can be seen in Figure 1. We first developed a local search algorithm, but it could not surpass experts’ loading skills. So, we ran a European academic challenge, then contracted with the winner team to integrate their algorithm and roll it out in our logistics platforms. The winning team (two researchers from a French university) designed an original

Everyday 3500 trucks deliver parts to our plants worldwide

FIGURE 2 ANTICIPATING DELIVERIES

LINE BALANCING FOR POLISHING WORKSHOP

where investments are severely capped, O.R. provides a golden opportunity to increase productivity A column generation method of mathematical programming was implemented and enables workshop managers to simulate several configurations with fewer operators, which is impossible to do manually. The tool is also used by headquartered process engineering teams to benchmark the quality of the line balancing done manually in some plants. Unsurprisingly very experienced workshop managers generate very cost-effective line balancing (nearly © Renault

Line balancing is very time consuming for workshop managers. The objective is to assign the list of production tasks to the minimal number of operators, with the main constraint that every operator should perform all their tasks within the cycle time, i.e. the time during which a vehicle goes through the operator’s workstation (usually one minute). In the polishing workshop of the paint line, the tasks consist in filling the holes under the car body with polish, as seen in Figure 5. Parts will be clipped under the body, thanks to these holes. We have to take into account operation times and moving time from one hole to another. Operation times were easy to obtain. But

they can visualise loading plans with a Renault’s web tool (see Figures 3 and 4) and have to follow them. In Figure 3 the three truck views represent the left-side, right-side and top view. Colours represent different suppliers loaded in the same truck, there are two suppliers in the displayed truck, the blue one and the red one. In addition to 2D views, the 3D view, as seen in Figure 4, gives a more comprehensible representation of the loading plan. Since truck filling rates are optimised, there is no longer room left for suppliers to add late deliveries surreptitiously! Also, the tool becomes very intrusive in the way packages should be regrouped into stacks and loaded into the trucks, whereas before suppliers and drivers were free to load as they want. The change management needed huge resources: there were more than 10 process engineers on the job, but only one engineer on the optimisation algorithm.

moving times were harder to determine, since they were not available before the development of the tool. It would have been excessive to ask managers to measure all the moving times of their operators. As a result, we made gross approximations of the moving time between holes, based either on the distance between the holes, or on the times between blocks when the holes do not belong to the same block of the car floor (which is divided into six same-sized blocks).

FIGURE 4 WEB TOOL TO VISUALISE LOADING PLANS FOR SUPPLIERS – 3D VIEW

FIGURE 5 OPERATORS FILL THE HOLES UNDER THE CAR BODY

IMPACT | AUTUMN 2021

It is a huge improvement compared to the existing process where layout specialists had to implement physically the kitting area to check the efficiency of the layout.

CONCLUSIONS FIGURE 6 VIEW OF ALL THE HOLES TO BE FILLED

equivalent to the tool) while junior managers are not as effective as the tool. It was critical to provide a graphical and intuitive representation of the solution, as shown in Figure 6, (where each colour represents holes assigned to an operator, and edges display the operators’ movements) so that workshop managers can validate the solution’s feasibility and quality.

The objective of the kitting layout model is to minimise the layout surface and the kitting operators moving times while satisfying very specific constraints of each plant (e.g. constraints associated with Automated Guided Vehicles used to move kitbox). The difficulty was to develop a generic tool for all plants, and not a list of workshop-specific algorithms. It required time-consuming discussions with the plants’ layout specialists. The roll out of the tool also required a fulltime resource in every plant. The trick is to convince the layout specialists that although the tool does not produce a perfect kitting layout, the tool provides a layout finalised for 80-90% of the parts (the user may have to modify the layout of 10-20% of parts) and enables the user to simulate several scenarios of layouts. Once the layout is calculated, a second tool enables the simulation of the movements of operators with virtual reality headsets (see Figure 8).

The kitting concept is designed for high diversity parts. Instead of locating these parts along the assembly line, which would require considerable space, the kitting concept requires that these parts are located in kitting areas, where operators will prepare kits of parts dedicated for each vehicle, in a kitbox (see Figure 7). Then a train of kitbox is moved to the assembly line, and when a vehicle arrives at the operator’s workstation, only the kitbox associated with this vehicle is brought to the operator.

FIGURE 7 A KITBOX

IMPACT | AUTUMN 2021

KITTING LAYOUT

FIGURE 8 3D SIMULATION OF THE KITTING AREA WITH VIRTUAL REALITY

O.R. demonstrates major benefits in terms of cost reductions for the supply chain and manufacturing. In a context where investments are severely capped, O.R. provides a golden opportunity to increase productivity. Although O.R. tools already cover a large array of topics, much remains to be done on the field. The current projects involve the network design for the delivery of parts from suppliers to plants and the empty packages dispatch from plants back to suppliers, which are two very complex issues, because there are many contradictory objectives, and also because processes are not well formalised. The main hurdle is a lack of ‘facilitators’, middlemen who understand O.R. principles and benefits as well as the business whereabouts. We need these facilitators to interact with field managers, formalise their needs and assist users in using the O.R. tools, which is not the easiest part! Alain Nguyen is the combinatorial optimisation expert at Renault, where he has spent all his career. He started in the AI team in the 1990s developing expert systems, then moved to the O.R. team in the 2000s, where he contributes to the development of the major tools for the supply chain and manufacturing. He has been a member of the bureau of the French O.R. Society since 2020.

I M P R OV I N G M E D I C A L S U P P LY C H A I N S IN CHALLENGING ENVIRONMENTS SARA VALIÑO, WOUTER SMEENK AND KOEN PETERS

2021 HAS SEEN SOME OF THE LARGEST LOGISTICAL OPERATIONS IN THE HISTORY OF MEDICINE, with governments all over the world rushing to roll out the various COVID-19 vaccines to their citizens. It is exciting to see operational research (O.R.) playing a pivotal role

in many of the distribution plans in high-income countries, but it is not (yet) a tool that people tend to look to in developing countries. Complex analytics are hard to apply in environments with lower data maturity after all, but with pragmatic and creative approaches a lot of value can still be unlocked.

The United Nations World Food Programme (WFP) has a long history of providing humanitarian assistance in challenging environments. Last year as many as 115 million people were provided with food assistance in more than 80 countries. WFP has increasingly been looking to analytics and O.R. to improve and manage its life-saving operations, and was awarded the 2021 Franz Edelman Award for Achievement in Advanced Analytics, Operations Research and Management Science for its use of O.R. to deliver food assistance amid emergency responses.

WFP has increasingly been looking to analytics and O.R. to improve and manage its life-saving operations Many of the industry standard models and tools fall short when exposed to a humanitarian setting, so in order to be able to utilise analytics in remote parts of the world where it works, WFP has had to come up with pragmatic approaches that balance the need for modelling complex operational environments with the availability and quality of data. This challenge is not just WFP’s alone, but spans the entire humanitarian sector. Increasingly WFP is being requested to support governments and NGOs in improving their supply chain operations through analytics, building on the methodologies that were developed for its own operations. Through its Health Supply Chain Strengthening team, WFP has been exploring how it could use the analytics it was applying to its own supply chain to improve and manage medical supply chains in challenging environments, working with partners such as Takeda, The Bill & Melinda Gates Foundation,

IMPACT | AUTUMN 2021

and The Global Fund. This article will look at one of these collaborations in more detail.

OPTIMISING DELIVERY ROUTES FOR THE CENTRAL MEDICAL STORE OF CÔTE D’IVOIRE

Public health distribution networks traditionally mirror administrative structures, which is not always beneficial for the supply chain execution and can lead to a surplus of storage points, greater distances driven, and higher numbers of vehicles in use. Sub-optimal supply chain networks not only tend to push up costs, but also increase management complexity and impact overall performance. Questions such as “how many warehouses? In which locations? How should we best route products for delivery?” are common for supply chain professionals trying to carry out these complex operations. When so many parameters are involved, it’s difficult for the human brain to find ideal solutions. This is where mathematical optimisation helps to rigorously explore the options and their impacts, allowing for better supply chain decisions. In the public health space, where resources are often scarce, such approaches can help unlock important value.

Mathematical optimisation helps to rigorously explore the options and their impacts, allowing for better supply chain decisions In 2019, as part of a broader collaboration between the Ministry of Health in Côte d’Ivoire and WFP’s Health Supply Chain Strengthening team, WFP began working with the Central Medical Store of Côte

d’Ivoire (NPSP) to use its analytical capabilities to help optimise a public health distribution system. As part of a decentralisation strategy, the NPSP opened a regional distribution hub in the central city of Bouaké. Truck delivery routes across the country needed a complete redesign to operationalise this new hub and so WFP worked alongside the NPSP to define an optimal delivery network blueprint. With the launching of its second distribution hub in Bouaké, the NPSP aims to improve service levels for the health facilities it serves. Originally more than 450 facilities were supplied from the Central Medical Store in Abidjan once per month. The NPSP developed a baseline delivery plan for a two-hub operation and invited WFP to challenge it and potentially come up with more optimal delivery routes. The objective? Minimise distances travelled by trucks while delivering to all client sites within the current onemonth cycle, whilst in parallel analyse solutions that could allow a future reduction of the delivery cycle from monthly to twice per month.

THE APPROACH

In collaboration with NPSP technical experts, WFP prioritised, structured, collected, and analysed data as well as applied various algorithms to evaluate different scenarios within the given constraints and data availability, but this wasn’t without its challenges. One of the major hurdles was the lack of data that optimisation experts typically expect when conducting analyses around facility allocation and vehicle routing decisions. For example, there was no geolocation available on NPSP’s client sites, and historic data

on demand and delivery volumes was specified in numbers of parcels (with unknown contents, volume, and weight for those parcels). The former was addressed through a location mapping exercise (using a mobile geo-tagging app), and to address the latter the parcel deliveries had to be used as a proxy for actual demand volumes – critical if you want to identify the most appropriate delivery routes. Data on routes between each health facility was collected using the Google Maps API once each facility was properly geo-tagged.

One of the major hurdles was the lack of data that optimisation experts typically expect when conducting analyses around facility allocation and vehicle routing decisions The next step was to identify zones and circuits that could be used to replenish the 450+ health facilities, based on their distance to the two hubs, the geographic layout of Côte d’Ivoire, and insights from the operational team. NPSP was planning to allocate a quarter to a third of the total volume to the new

hub in Bouaké, so a clustering algorithm was used to allocate a sub-set of facilities accordingly (see Figure 1), based on their estimated demand and the relative distance to each hub. Circuits were then developed together with logisticians on the ground based on available truck sizes, demand estimations, and the major road networks in each region. One important consideration was that NPSP should be able to run analyses independently, so the entire process was captured in an Excel-based tool that they could use to redefine demand, circuits, and routes and automatically assess the performance of their new plan across a wide range of KPIs. Examples include average truck utilisation, number of kilometres driven, the logistics cost, and the overall number of trips required to fulfil all operational demand.

RESULTS

Besides evaluating the performance of the baseline (the one-hub model) and the new NPSP proposal, the optimisation approach was used to explore strategies for different

network configurations and demand projections. Two scenarios in particular were explored: an optimised scenario following the current approach of monthly delivery cycles, and an optimised scenario with 15-day delivery cycles for each facility. Each of the scenarios represents a delivery plan, allocating specific client sites to Abidjan or Bouaké agencies, and grouping sites into six new delivery rounds. The NPSP performed targeted field tests to ensure proposed delivery routes were feasible, and when needed, adjustments were made. The Excel implementation meant that the whole process was very transparent and easy to adjust. In the current supply chain set-up (the one-hub model) the total distance that has to be covered to supply all zones is approximately 47,000km across 91 trips. Opening the new warehouse in Bouaké is expected to decrease this to approximately 45,000km without any changes to the number of trips. These gains were not as high as expected because the proposed circuits and zones were not yet optimised for the new set-up. After applying the optimised circuits and

FIGURE 1 ALLOCATING HEALTH FACILITIES TO HUBS (LEFT), TRUCK CIRCUITS (RIGHT)

IMPACT | AUTUMN 2021

IMPLEMENTATION

FIGURE 2 PROJECTED KPI IMPROVEMENTS OF THE OPTIMISED SOLUTION COMPARED TO THE CURRENT ONE-HUB NETWORK SETUP AND THE INITIAL PROPOSAL FOR THE TWO-HUB NETWORK

zones, the total driving distance is expected to drop by another 12,000km to 33,000km, representing a reduction of 27 percent. Furthermore, the expected number of trips necessary for one delivery cycle is expected to drop from 91 to 85 and the average truck capacity utilisation increases from 49 percent to 88 percent if the optimised circuits and zones are implemented (see Figure 2). Additionally, WFP Supply Chain Planners explored the potential of moving from 30-day delivery cycles to 15-day cycles. An operating model where the frequency of delivery is increased would improve the service

IMPACT | AUTUMN 2021

level and benefit big customers especially, with high quantity orders every cycle. Compared to the initial NPSP proposal, the optimised 15-day cycle scenario still represented a 14 percent reduction in distance covered (38,896km) and 33 percent increase in vehicle utilisation (82 percent), but the number of required trips (per month) would increase from 91 trips to 103 trips. A hybrid model where only the biggest facilities (representing 20 percent of demand) would receive two deliveries per month was found to be the most robust, also accounting for expected increases in operational demand.

With this powerful analysis and encouraging results in its hands, the NPSP was empowered to pick and implement the scenario that it sees best fit to serve its clients and reach its goals. Handover sessions took place with NPSP focal points on how to use and configure the Excel tool. While the tool is not a structural solution to manage medicine distribution networks (e.g. if a new hub opens in the country the core optimisation exercise of clustering locations and defining the delivery zones should be produced again), it has proven to be a useful offline tool to explore small changes in the network (e.g. manual modification of a circuit) and to perform quick comparisons of scenarios. The NPSP undertook field missions to confirm the operational feasibility of proposals developed with WFP and adjustments were made to the models to fully align with realities in the field. The additional scenarios that were developed with NPSP experts allowed them to explore feasibility of future increase of service levels (e.g. the 15 days scenario) and operational demand. Delays in the Bouaké hub project were faced due to the ongoing COVID-19 response, but NPSP leadership had all the necessary analysis required to make the best possible strategic network design decisions available. Finally, in November 2020 the new Bouaké hub opened and NPSP confirmed that the study was used for the routing design of most of the new itineraries. As recommended by WFP, a second feasibility check was done before the opening and some of the itineraries were adapted due to access constraints. Also following WFP recommendations, the NPSP implemented a hybrid

model: a 15 day delivery cycle for key customers with high rotation and a one month delivery cycle for remaining customers.

REFLECTION

Although data availability and quality are still challenging in many parts of the world, there is no question that analytics can still bring business value. Analysing operations in such a context requires a softer and more pragmatic approach than we are used to in the field of operational research, but when we let go of our expectations and ideal scenarios there is a lot that can be done with the data that is there. That’s why it’s imperative to continue working with NGOs and governments to help them unlock the untapped potential of

their data, and help them to do more good with their limited resources.

It’s imperative to continue working with NGOs and governments to help them unlock the untapped potential of their data, and help them to do more good with their limited resources

Wouter Smeenk has been working with WFP’s Supply Chain Planning and Optimization service for four years, developing and applying operational research tools to support the operations of WFP partners and governments, including the control tower backing WFP’s COVID-19 response. He has a Master’s degree in Operations Research from Tilburg University.

Sara Valiño has been working with WFP’s Supply Chain Planning and Optimization service for three years, strengthening the supply chain planning practices of WFP’s partners in the health sector and supporting WFP’s operations in Latin America with advanced analytics. She has a Master’s degree in Supply Chain Management from the MITZaragoza Logistics Center.

Koen Peters has been working with WFP’s Supply Chain Planning and Optimization service for seven years, developing and applying operational research tools to support the design and management of WFP operations. He has a Master’s degree in Operations Research from Tilburg University and is currently pursuing a PhD in humanitarian logistics at the Zero Hunger Lab.

IMPACT | AUTUMN 2021

NET ZERO EMISSIONS – FROM WHY TO HOW BEVAN FREAKE

MOST OF US UNDERSTAND THE NEED to drastically reduce man-made greenhouse gas (GHG) emissions to limit the effect of climate change and its damaging consequences. The need to achieve net zero emissions has been much publicised, but what does this actually mean for how we live our lives and the changes that will need to be made? How can we heat our homes, fuel our transport, and supply our electricity? How different will 2050 need to be? To help answer these questions, the UK Government’s Department

for Business, Energy & Industrial Strategy (BEIS) have developed two new online energy and emissions Calculators for creating pathways to net zero, called My2050 and the MacKay Carbon Calculator. At a launch on 3 December 2020, in his opening statement the then Minister, and now Secretary of State, for BEIS, Kwasi Kwarteng, said ‘I believe a green recovery can lay the foundations of a clean and prosperous future, and the Calculator gives a common language for us to represent that shared vision.’

THE POLITICAL LANDSCAPE

Governments across the globe recognise it is essential to cut emissions. In 2015, 196 Parties came together under the Paris Agreement to transform their development trajectories and set the world on a course towards sustainable development. The agreement aims to limit warming to 1.5–2 °C above pre-industrial levels. In 2019 the UK became the first major economy to commit to a ‘net zero’ target, requiring the UK to bring all greenhouse gas emissions to net zero by 2050 under the Climate Change Act. Following this, in November 2020, the UK Prime Minister announced a 10-point plan of action for 2030 and in early December, he announced the UK’s first Nationally Determined Contribution (NDC) to reduce greenhouse gas emissions by 68% by 2030. The UK also jointly holds the presidency for the UNFCCC Conference of Parties meeting in Glasgow during November this year (better known as COP26) to push for greater international commitment in the fight against climate change.

model themselves, both at high level for the non-expert, and intricate detail for the expert. They are designed to aid understanding of the options for meeting our emissions targets, identify any crucial decision points and engage the public in the debate by helping them to understand the trade-offs. The problem is broken down and presented to you through simple choices for ‘levers of decarbonisation’. For each lever, you set the level of ambition anywhere on a scale from minimal effort (level 1) to the maximum effort deemed plausible (level 4). For example, the proportion of cars and vans that are electric vehicles remains at today’s level under level 1 but reaches 100% under level 4. Levels of ambition were set in consultation with experts across industry and academia so they should not be regarded as only being a government view. The original UK 2050 Calculator was developed in 2010 and the simplified My2050 version was launched in 2012, aimed at the general public. Over 20,000

people submitted pathways using the tool, providing a unique insight into the public’s preferences for a lowcarbon future. Guidance was produced for teachers so it could be used in the classroom as well. The inspiration and driving force behind the original Calculator was the late Sir David MacKay and the principles in his book Sustainable Energy – Without The Hot Air. He stressed the importance of quantifying the challenges we face with numbers rather than adjectives and breaking the problem down into manageable chunks. The new tool was designed with the same philosophy and was named after him in tribute.

MY2050

My2050 (https://my2050.beis.gov.uk/) is about helping the public and schools to engage with net zero and the ways in which individuals can help through their lifestyle choices. You decide the level of ambition under 15 levers, covering both

Department for Business, Energy & Industrial Strategy

the UK Government’s Department for Business, Energy & Industrial Strategy (BEIS) have developed two new online energy and emissions Calculators for creating pathways to net zero

WHAT ARE THE CALCULATORS?

The Calculators are analytical models for identifying and exploring the range of physically possible scenarios for cutting greenhouse gas emissions. They are accompanied by interactive online tools allowing individuals to use the

MY2050 HOME PAGE.

IMPACT | AUTUMN 2021

Department for Business, Energy & Industrial Strategy MACKAY CARBON CALCULATOR UNDER ESC (ENERGY SYSTEMS CATAPULT) ILLUSTRATIVE NET ZERO PATHWAY.

energy demand and supply across sectors such as transport, buildings, industry and electricity generation. Clicking any lever name opens an information page on the technology and the lever options to allow more informed choices. Immediate feedback on the implications of your choices is provided through an eye-catching animation and a greenhouse gas (GHG) meter showing the percentage emissions reduction. The tool has a game-like feel designed to reinforce the goal of meeting net zero emissions. Although 15 independent levers allow a lot of possible combinations, only a few of them lead to net zero.

The Calculator confronts issues head on putting the power in your hands to choose between often competing options

THE MACKAY CARBON CALCULATOR

The MacKay Carbon Calculator (https://mackaycarboncalculator. beis.gov.uk/) is a much more detailed

IMPACT | AUTUMN 2021

tool aimed at a better informed user. It offers both greater flexibility with 45 decision levers and more comprehensive feedback with 30 interactive graphs focussed on key metrics for different areas of the energy system. It also introduces new features in the more advanced ‘2100 mode’, where the timing of action is brought under consideration. You are now able to customise when you think each system change should occur by choosing the start and end point for the action. This gives you the opportunity to explore options for when we could reach net zero or to prioritise which areas decarbonise early and which come later. If the number of choices feels overwhelming, you can start from a pre-set example pathway, ‘ESC Net Zero Illustrative’, provided as an independent balanced pathway to net zero by partners Energy Systems Catapult (ESC). Both tools are based on the same spreadsheet model which has also been published for complete transparency and to allow full

customisation. This is more detailed still with over 160 levers.

BRINGING OUT THE KEY CHOICES

Reaching net zero emissions requires difficult choices and there is no single right way of doing it. The Calculator confronts issues head on, putting the power in your hands to choose between often competing options. For example, should our buildings be converted so that heat is supplied from district heating networks or heat pumps? Or should we continue to use gas boilers, but fuel them using a gas grid that has been decarbonised with low carbon hydrogen or biogas? Should we supply zero carbon electricity from nuclear or focus on renewable technologies such as wind and solar? Should we use more land for forestry and bioenergy, and if so, will this require changes to our diet in order to free up land currently used for livestock? How should we balance behavioural and technology change? The Calculator emphasises that net zero is a whole system problem.

The importance of both demand and supply side actions is a core feature. Switching to hydrogen will only reduce emissions if you ensure that hydrogen is supplied from low carbon sources. The impact of converting from petrol and diesel to electric vehicles may seem surprisingly small if you do not supply the additional electricity demand with low carbon generation. The tool forces you to make conscious choices so you gain a more rounded understanding of the problems at hand.

The Calculator emphasises that net zero is a whole system problem

When must action start? 2050 may seem far off but when you consider a car may last 10–15 years, a boiler 10–20 years and a power station up to 50 years the choices we make today are hugely important. How quickly can we deploy new technology and adjust behaviour? Change will not happen overnight so

net zero is a problem for now, not just for 2050.

GLOBAL IMPACT AND INFLUENCE

Climate change is a global concern, and the UK is not the only country facing the issue of how to decarbonise our energy system. Analysts in other countries became interested in the Calculator approach very quickly after the original was published. With the launch of the EU Calculator in 2020, 61 Calculators have now been or are being developed (55 countries, four territories and two cities). Each one has taken the overall approach developed by BEIS and adapted it to local circumstances. Since the creation of the first UK Calculator in 2010, the UK Government has been supporting other countries to create their own Calculators through BEIS-funded international outreach work. The 2050 Calculator international outreach programme is part of UK International Climate Finance, and in its first phase supported ten developing countries to create Calculators. It has been extended due to demand, and is currently supporting projects in India, Malaysia, Nigeria, the Philippines, Thailand and Vietnam, with proposals in development for Colombia, Kenya and Zimbabwe. Around the world, BEIS-supported Calculators have been used to develop NDCs (Nationally Determined Contributions) as commitments to the

Paris Agreement (in India, Vietnam, Colombia, and Nigeria), national energy plans and sectoral action plans. They have also informed policymakers, the general public and industry on decarbonisation strategies as well as raising awareness of key issues. All this is possible because the Calculators provide an open, trusted and transparent forum for policymakers to co-design, develop and build solutions to climate, energy and land use issues. If you would like further information about the Calculators, there is a Government carbon Calculator page (https://www.gov. uk/guidance/carbon-calculator), or you can email the Calculator Team at mackaycarboncalculator@beis.gov.uk. To support use of the tools in schools, the Royal Geographical Society has created GCSE and A-level teaching resources which are available on its website, or via the gov.uk page. We hope the tools will make the conversation around net zero more accessible to a wider audience and support debate on planning our low carbon future.

Bevan Freake is a member of the Government Operational Research Society (GORS) working for the Department for Business Energy & Industrial Strategy (BEIS). He was the modelling lead in developing the MacKay Carbon Calculator and has experience working on other whole energy system models for projecting UK energy and emissions and cost optimisation to meet emissions targets.

IMPACT | AUTUMN 2021

U N I V E R S I T I E S M A K I N G A N I M PAC T EACH YEAR STUDENTS on MSc programmes in analytical subjects at several UK universities spend their last few months undertaking a project, often for an organisation. These projects can make a significant impact. This issue features reports of projects recently carried out at two of our universities: Cardiff University and London School of Economics. If you are interested in availing yourself of such an opportunity, please contact the Operational Research Society at email@ theorsociety.com THEATRE INSTRUMENT FLOW OPTIMISATION AT THE GRANGE UNIVERSITY HOSPITAL (Charlotte Marshall, Cardiff University, MSc Operational Research and Applied Statistics)

Grange University Hospital (GUH) is a new Specialised Critical Care Centre hospital in Cwmbran, South Wales which opened four months early in November 2020 within the Aneurin Bevan University Health Board (ABUHB). The new hospital deals mostly with emergencies, so having the correct stock of surgical instrument trays in the theatres is crucial. Charlotte’s dissertation looked at optimising the number of trays which need to be stored at GUH to reduce the number that are not available. At GUH, there is a set amount of storage where the sterilised and decontaminated trays which hold surgical instruments can be kept. After use these trays are taken to a Hospital Sterilisation and Decontamination Unit (HSDU) to be sterilised and decontaminated before they can be used again, to reduce contamination and infection between patients. The model Charlotte created takes the mean and standard deviation of the daily demand for each tray and the mean and standard deviation for the lead time (turnaround time) for this tray to estimate the required number. This estimate depends on the service level, defined as the proportion of cases where 32

IMPACT © THE OR SOCIETY

the stock is sufficient to meet demand. In the model, the service level can either be specified (e.g. we want to be 99% sure we have sufficient stock to meet demand and turnaround variability) or it can be optimised using the model, to maximise the service level across all procedures within the constraint of the maximum storage available in the hospital. Charlotte’s dissertation highlighted the need for more data to be available for tray turnaround time, demand and whereabouts. This will enable a more accurate estimation of the number of trays needed not just at GUH, but also at the other hospitals within ABUHB. Overall, the optimisation results showed that there is sufficient storage space at GUH to allow for a maximum average service level of 99.988% for the emergency procedures. However, while this high service level can be achieved, the trade-off between service level and cost of equipment must be considered. If the surgical trays required are not available to be moved from other hospitals, they will have to be purchased. This may mean a large cost in equipment for a small increase in service level. Trish Chalk, Head of System Planning at ABUHB: “Charlotte’s dissertation has involved the move of

emergency and high acuity elective surgery to the GUH. Due to the pandemic the hospital was opened four months early which meant accelerating timescales that had been in place since the project conception. Charlotte’s work enabled the Division to test and understand if the planning assumptions made for Theatres were correct and if any immediate remedial actions were going to be required.” Terry Watkins, Senior Programme Manager at ABUHB: “When Charlotte joined the division, in the middle of the Covid 19 Pandemic, she very quickly identified the need to standardise trays because of the various types of procedures carried out on different sites by different surgeons who had their specific instrumentation. This was a pivotal development in the analysis and allowed the teams to significantly reduce the storage space required at the new hospital. Charlotte’s work has allowed us to evidence the changes needed to change the layout and storage configuration and also secure the funding for the new system. Charlotte’s work has been highly valued and formed an important analysis: without it we could have been in a very different position which would have affected service delivery.”

MEASURING TEAM DISRUPTION (Mohammed S. Agha, LSE, MSc Operational Research and Analytics)

The LSE Operational Research and Analytics Masters programme has a long-standing and valued relationship with Satalia, which has regularly hosted summer consultancy projects and employed several of its graduates. Satalia is a thoughtleader in AI, pioneering the future of work by combining technology with organisational psychology to create swarm-like organisations and many of the projects that students have worked on involved the development of novel approaches to the challenges Satalia’s clients are facing. In this case, the focus was on the potential use of social network analysis and data science to measure and minimise disruption to high-performing teams. Mohammed’s project explored two specific questions that arise when members exit a team: who are the critical members that may disrupt the team when removed, and what is the effect on the team when certain members are removed? The analytical approach, the use of a synthetic network generator for testing, and results are described in detail in a Satalia Blog post (see http://bit.ly/ SataliaBlog). LSE students who wish to apply for a particular project submit their applications to the LSE Projects Coordinator who compiles a shortlist

of candidates for each client to choose from. Mohamed was shortlisted and subsequently selected because his application showed good insight into the problems, because he had demonstrated the necessary technical abilities, and because his background, and motivation and character matched the needs of the project. The core was the application of methods taught on the course but exploratory projects like this one demand much more of the student than technical skills, especially when they must be delivered completely remotely due to COVID restrictions. The student must have the ability to succeed and be self-starting, but also have the confidence to take on a project where there is no predetermined methodology and which might, through no fault of their own, fail to produce a positive outcome. Laura Weis, Satalia’s project supervisor said:” Both our students this year were extremely professional and reliable and coped well with the complexity of their projects and some of the challenges we faced with regards to providing them with data. Mohammed’s experiments confirmed the usefulness of a wide variety of network metrics in measuring and preventing disruption and there are valuable results we could build on in further projects. We were pleased to

hear that he received a Distinction for this part of his degree and wish him well for the future”. The open brief had initially given Mohammed pause for thought, but the Satalia team had given a talk to students earlier in the year so he knew who he would be working with, and that they had a track record of successful projects with LSE. He commented “The project offered me a great chance to work with advanced techniques on real problems and test my analytical thinking and decision making. There were inevitably moments when we seemed to be going backwards instead of solving the problem, but I am really pleased with the way the project eventually worked out”. The aim of the projects on the MSc programme is to develop and test the consultancy and project management skills taught on the course as well as the students’ technical ability. This is demanding on the students and on the LSE’s supervision team, who coach and mentor them through the project. However, LSE believes the result is a more rounded and more employable graduate who has demonstrated more than technical competence. Mohammed is now building his career in the UK as an optimisation practitioner.

IMPACT | AUTUMN 2021

SECRETS OF SUCCESS WITH LOAN PRICING OPTIMISATION BARRY HONEYCOMBE AND MARC DROBE

HERE WE EXAMINE HOW LOAN PRICING OPTIMISATION WORKS and how it prompted a 29% increase in new sales for Home Credit Russia. Financial organisations looking to achieve their portfolio goals in these difficult times should investigate loan pricing optimisation. Increasing your precision around pricing can virtually guarantee a significant boost in profitability, without a massive IT project. A good example of this comes from Home Credit Russia, which operates across three

continents, mainly in markets with high growth potential. Petr Kapoun, Chief Risk Officer at Home Credit Russia used optimisation to achieve remarkable results: a 26% rise in portfolio profit and a 29% increase in new sales.

WHAT DID HOME CREDIT RUSSIA WANT?

The overall objective for Home Credit Russia was for its portfolio management to maximise profit within

defined constraints. Its initial target was set to achieve a minimum 10% increase at constant risk costs. Due to local regulations and potential impacts on regulatory capital, it also needed to control risk weighted assets as linked to interest rates threshold. Another core objective was to increase transparency and manageability. Home Credit Russia was looking to reduce complexity in parameters used for limit calculation. It wanted to move from risk based to risk and revenue-based pricing. The strengths behind loan pricing optimisation platforms were clear to Petr Kapoun and Home Credit Russia. Results can be achieved within a few months, there is no need for lengthy IT projects or long-term planning. Whilst optimisation can be deployed as a real-time process, the best practice in banking and financial services is to create the chosen strategies in an off-line planning environment, taking account of the portfolio and accountlevel constraints and determining a preferred scenario that meets the bank’s strategic objectives. Once this scenario is stress-tested to understand how it will hold up under a number of different economic scenarios, it is then output in a format that can be easily available for real-time deployment

FIGURE 1 THE TRADITIONAL APPROACH

in the bank’s existing systems. This approach ensures minimal IT integration. This was particularly important given the financial climate faced in 2020 and beyond, brought on by the pandemic. With resources already stretched thin, it was crucial to Home Credit Russia that the on-boarding and activation process for new accounts was streamlined and smooth.

Home Credit Russia used optimisation to achieve remarkable results: a 26% rise in portfolio profit and a 29% increase in new sales.

THE TRADITIONAL APPROACH

In loan pricing optimisation — as, indeed, in any decision modelling and optimisation project — the traditional approach to formulating a decision strategy gets flipped around. The traditional approach, illustrated in Figure 1, starts with the known and available information, e.g. credit bureau data, demographic data, customer application data, and the predictive models that assess factors like response

likelihood or risk level. Typically, the segmentation of the applicant universe into different tree branches, and the assignment of actions to each end node, is done judgmentally. Once implemented, the results of this strategy can then be measured along the defined KPIs, typically using metrics such as revenues, losses, costs, and capital. The problem is that there is a universe of possible decisions for each customer, and thousands of possible alternative strategies that cannot be evaluated within traditional champion-challenger approaches. In the example above, there are the price, amount, term, and accept/ reject decisions where each customer only gets one decision per category. However, the decision being made in constructing a decision strategy has multiple facets - where to set the cut-off, how to segment, which action to assign to which segment – and is multi-dimensional, as risk, revenues, take-up, etc. all need to be considered. For expert strategies, these are essentially estimates supported by unidimensional predictions.

WORKING BACKWARDS FROM THE OBJECTIVES

Rather than start from the data on the left and work toward the outcomes on the right, in optimisation we start with the objective and work backwards, as shown in Figure 2. The overall objective of a retail lending strategy is typically to maximise profit, but there can also be secondary goals, for example portfolio growth, keeping the accept rate constant, or reducing the loss rate. With decision modelling and optimisation, the starting point

IMPACT | AUTUMN 2021

So the question becomes, given that I have different, competing objectives, what is the impact on my KPIs if I move away from BAU and: • Maximise net interest income (NII)? • Maximise origination volume? • Can I increase both? FIGURE 2 THE DECISION MODELLING APPROACH

includes some known information. But rather than working from left to right, where objectives are present and strategies designed to achieve these goals, the process moves from right to left. It is then possible to define the key metrics that are driving the objectives and constraints can be calculated and measured. At this stage, rather than segmenting the portfolio and assigning one action, within the decision modelling and optimisation framework it is important to look at all possible actions that can be taken for each customer. Key aspects are action-effect models, which encode the relationship between a given action and the likely response, that are built into the solution. This provides the basis for optimisation algorithms to search through a gigantic space of possible alternatives in a way no person developing a strategy ever could, identifying optimal decisions for each customer.

ASKING THE “WHAT IF” QUESTIONS

The optimisation framework enables greater collaboration between the data scientists and the business manager. This is especially useful when you need to rapidly project the results of

IMPACT | AUTUMN 2021

scenarios in which certain variables or segmentation criteria are changed. The typical question a portfolio manager will ask is a what if – such as what is the impact on my portfolio if I change my strategy? In this example, we are using pricing decisions that drive revenue, but also need to consider price sensitivities that impact take-up and volume.

The typical question a portfolio manager will ask is a what if – such as what is the impact on my portfolio if I change my strategy? For the business as usual (BAU) strategy, i.e. what if I keep the same pricing, certain data should be immediately available: number of applications, approval rate, take-up rate, average loan amount, etc. What should also be known are the constraints that are defining the action space, i.e. the ‘room’ in which alternative strategies may be found. These can be regulatory constraints, such as capital; business constraints, such as funds available; or management guidance, such as ‘keep the market share’. No financial institution is operating in a completely unconstrained environment.

As you see here, the BAU strategy is our starting point, to which we are comparing different scenarios where we have optimised within the scenario parameters, or constraints. So, let’s look at the first question: what if I maximise NII? What is the most we can increase it by, and how much origination volume must we give up to achieve this? As you can see in the scenario shown in Figure 3, the NII will increase by approx. 20%, but origination volume will drop from 12 to 10.5 million. On the other end of the scale, maximising origination volume, we see a substantial drop in NII. All other points on the efficient frontier here are representing operating points balancing NII and originations volume to different degrees, constraining the origination’s volume. In addition to answering the “what if?” questions through a process of simulation, we combine this with optimisation scenarios which tell us what action, in this case pricing, we need to deploy to achieve a given level of profit within our constraints. The combination of simulation and optimisation scenarios and subsequent comparison enables the business user, such as Home Credit Russia, to determine their preferred strategy that meets the goal of the business within the constraints.

and its aftermath will linger for years, even once the vaccine has been globally administered.

With COVID came new working strategies, a plethora of changes to financial agreements, and confusion

FIGURE 3 FICO OPTIMISATION MODELER, LOAN PRICING DEMO

RESULTS AT MULTIPLE LEVELS

The results speak for themselves. Home Credit Russia has used FICO optimisation to achieve remarkable results: 26% increase in portfolio profit and a 29% increase in new sales. In addition, Home Credit identified loss making pockets in the portfolio that had been accepted before due to lack of granularity. Scenario analysis allowed them to deeply understand trade-offs, e.g. volume, interest and take-up rate. Kapoun said: “Activating the loan pricing optimisation framework has truly benefitted Home Credit Russia. There are two main reasons for this. Everyone faced a disruption or change to their financial position in 2020. From job losses to payment holidays, there was so much activity it would have been extremely difficult for us to provide the same level of service our customers expect. Optimisation crunches the numbers, inputs all the factors, and produces the correct offers.”

Optimisation establishes a clear structure where there have been many overlapping policy rules, different pricing tables, and difficult simulations of what happens if Home Credit Russia does change its policies. Kapoun and Home Credit also found that optimisation helped identify where the business can use analytics expertise to deliver significant business impact. With FICO’s optimisation tools, Home Credit began to see the benefits in its various teams and overall optimisation should usher in further cooperation. Business and risk will work together to find the best trade off, discussions are driven by data rather than assumptions, and there is more of a forward-thinking mentality. It appears that there has never been a more pertinent time to adopt loan pricing optimisation tools and platforms within financial organisations. With COVID came new working strategies, a plethora of changes to financial agreements, and confusion. But it is clear that COVID

With optimisation, all of the relevant parties at a lender come together quickly and efficiently to understand how their objectives can be managed together with a common objective based on fact rather than assumptions. For more information on decision optimisation and the Home Credit Russia case study, check out the video recordings of a roundtable webinar with Petr Kapoun and FICO’s Marc Drobe and Barry Honeycombe here: www.fico.com/blogs/loan-pricingoptimization. Barry Honeycombe is a principal consultant at FICO, responsible for optimisation solutions delivered to clients across Europe, Middle East & Africa. He has more than 25 years of experience in applying advanced analytics across the customer lifecycle including marketing, pricing, customer retention and collections. Marc Drobe is a principal consultant at FICO with 30 years of experience in various roles in the financial services industry. Marc is a trained banker and has worked both in line positions as a risk manager and in consulting positions advising customers on risk and revenue management. In his current position at FICO, Marc is advising customers on designing and implementing solutions based on the FICO Decision Management Platform, including decision modelling and optimisation.

IMPACT | AUTUMN 2021

CREATING NOVEL MOBILITY INSIGHTS IN A NATIONAL LOCKDOWN JORDAN LOW AND SAM ROSE

PRIOR TO THE PANDEMIC, more people than ever were travelling for work and leisure. Now though, it can be hard to remember a time when commuting into the office was commonplace and seeing other people involved deciding on where to meet, not which video conferencing software to use. For over a year, coronavirus restrictions of some form have guided where the UK public can travel to and

who we can see when we get there. So at some point you might have been curious whether this stay-at-home guidance was being followed on a large scale, what effect it has had on patterns of movement and commuter habits, or what sorts of events or occasions might make it more or less likely that people will travel? These questions might simply be interesting to most people, but for

UNDERSTANDING MOBILITY IN NEAR REAL TIME

Many openly available sources that are now online were not openly available at the start of the pandemic and were not representative enough for these purposes as they looked only at specific groups of people. Instead, DfT worked with O2 Motion to access data from the mobile phone network, which had not previously been used for this purpose in government. (NB. All data, insights and trends throughout are anonymised and aggregated, never allow identification or mapping of individuals, and operate within strict privacy guidelines). This data is anonymised and aggregated to local authority level. We then looked at other existing data sources such as the Office of National Statistics (ONS) census information and the resulting dataset formed the basis for a range of analysis and insights to support aggregated mobility monitoring both

IMPACT | AUTUMN 2021

during and after each lockdown, as well as enabling deep-dives into regional issues. Within days this became a daily reporting function alongside other leading indicators for individual transport modes that supported the monitoring and decision making in government.

included in these packs, showing changes in total mobility compared to pre-pandemic levels, which were later produced for different trip purposes and geographical areas (axis removed for publication). During the height of lockdown, road travel dropped by over 75%.

this became a daily reporting function alongside other leading indicators for individual transport modes that supported the monitoring and decision making in government

DEVELOPING CRITICAL ANALYSIS

On 23rd March 2020, the UK entered lockdown. From our first look at the mobile telecoms trip data on the same day, the department’s Analytics Unit paused existing project delivery and put out the first mobility insights pack to the operations teams working on the pandemic less than 48 hours later. The starting objective of this work was simply to understand patterns of national mobility, so these daily packs included the total trips being undertaken both across the UK and in geographical areas, benchmarked against typical trip levels from before the pandemic. Figure 1 is representative of charts

It rapidly became clear however that the dataset presented additional utility beyond these high-level measures. A close working relationship between the O2 Motion team and the DfT enabled additional datasets to be brought online as the project expanded, including hourly data to help understand the potential pressures on public transport at peak times, insight into commutes and leisure trips in travel patterns, and critically an Origin-Destination matrix, which provided mobility information not readily available from any other source. This became particularly important when England moved to a tier system and there was a requirement to look at cross border travel and identify potential future hotspots. Throughout lockdown the team continued to develop new insights through linking these new data sources to existing Government data both

those making critical decisions they’re essential information. Whenever decisions like the coronavirus travel restrictions are planned or implemented, decision makers want and need to monitor and understand their impact so that, if necessary, changes can be made. Where the coronavirus travel restrictions differ from many transport policy issues however, is that every mode of travel is affected, and we needed to know what was happening in near real time. Near real time mobility data across all modes wasn’t something which the UK Government’s Department for Transport (DfT) had been faced with before and so we needed a new approach.

FIGURE 1 AN EXAMPLE OF CHARTS PRODUCED TO SHOW THE IMPACT OF A NATIONAL LOCKDOWN ON POPULATION MOBILITY

nationally and regionally as well as additional open source data when it was available. Through the mobility data, we could investigate connectivity between hot spot areas and the rest of the UK. As cycling became more prominent in how people might travel, we linked total mobility estimates with other modal data to derive cycling estimates. This later became a new daily cycling series, which had not previously been created. And by bringing in more ONS census data we could move into predictive analytics, looking at key workers and shopping habits to estimate expected levels of mobility under key policies. Up to 32 different analytical products were being produced at peak, over four main areas, some of which are listed in the panel.

As cycling became more prominent in how people might travel, we linked total mobility estimates with other modal data to derive cycling estimates

A COMPLETELY NEW WAY OF WORKING

To enable rapid working against the intensely demanding timescales of the pandemic crisis, the team developed new ways of working on top of the already challenging move to remote working. Each time new analysis, data series or products were developed the team were quick to automate these and continued to improve them throughout. Enabled by shared software, we were able to ensure resilience by alternating who worked on

each task and through regular updates. Normally the Analytics Unit’s shortest projects last weeks to months, so this represented a completely new way of working for many in the team.

This work has produced evidence to support monitoring and decisions as the crisis has unfolded

FAR-REACHING IMPACT

This work has produced evidence to support monitoring and decisions as the crisis has unfolded. But in addition, while meeting the mobility information needs of the crisis, the team has ensured that the benefits of improved insight into mobility will be sustainable, meeting the longer-term needs of Government and

A WIDE RANGE OF ANALYTICAL PRODUCTS

The tempo of working, repeated needs from central government, and the department’s access to anonymised, aggregated insight from mobile telecoms data meant that at peak periods, the team were producing up to 32 different analytical products daily and weekly for different purposes, highlighting the range of possibilities that this sort of data presented. Products included:

Daily mobility updates

Looking at big trends in national mobility to answer questions on how mobility is changing as a result of lockdown, and throughout restart, as well as more focussed analysis on geographic areas and times of day.

Predictive Analytics

These products answered questions on what effect on total mobility should be expected due to the lockdown, so that this can then be compared to insights coming from the daily updates, and areas or times which were different than estimated can be identified.

Local lockdown connectedness

We also provided analysis for planners, identifying potential areas for future outbreaks and looking at how people are moving at different times of day.

Cycling

Data on trips by bicycle in near real time has not previously been available on a regular basis, but by combining the mobility data with other available sources the team derived a measure for the change in usage.

IMPACT | AUTUMN 2021

highlighting what is possible with this sort of data. We continue to work with stakeholders across the department to identify new areas where this data source can be used and with wider Government on how to source and use data like this in the future.

this project has meant that the DfT and wider Government have been able to support the UK community with more evidenced based decision making Feedback has demonstrated this wide-ranging impact, as “The support provided to us has been incredibly useful in understanding how patterns

IMPACT | AUTUMN 2021

of movement have changed during the course of the pandemic which has informed decision making… and has provided an important source of quality assurance on mobility trends from other new sources or novel methodologies”. Fundamentally, this project has meant that the DfT and wider Government have been able to support the UK community with more evidenced based decision making, furthering their use of new and innovative data sources and supporting a more robust response to what is the largest National Emergency in decades. Jordan Low is a senior operational research analyst within the Department forTransport’s Advanced Analytics Division.

Sam Rose is Deputy Director of Advanced Analytics and Head of Profession for OR in the Department for Transport. This work was awarded the OR Society’s President’s Medal in 2020. © Crown copyright 2021, DfT. This material is licensed under the terms of the Open Government Licence except where otherwise stated. To view this licence, visit http:// www.nationalarchives.gov.uk/doc/ open-governmentlicence/version/3 or write to the Information Policy Team, The National Archives, Kew, London TW9 4DU, or email: psi@ nationalarchives.gov.uk.

M A K I N G A N I M PAC T W I T H SOFT OR MARTIN PARR

THERE IS A CONTINUAL NEED WITHIN INDUSTRY AND COMMERCE to ‘follow a market’. A simple example of this would be presenting a department in a good light so that strong candidates would apply for jobs and further strengthen the department’s staff. Within government departments there are similar challenges in adapting to the changing needs of society or changes due to strategic events like Brexit and Covid. On a daily basis, alignment activities are done informally though conversations and actions that lead to adaptation. But some adaptations need more consideration. For these adaptations the ‘executive troubleshooter’ has been a familiar icon. A trouble-shooter will attempt to re-align a company that has drifted away from its market, its core strength or its ability to meet the demands placed on it. One of the keys to success in any alignment is balancing an organisation so that it is receiving the right information to control its operational system. The role of management is critical in interpreting information from, and interacting with, an external environment. The management system provides appropriate direction and advice to the operations system, listening to how well activities are progressing. Figure 1 shows a very simple model of a company and its environment, built from the most fundamental elements of a Viable Systems Model (VSM).

Soft methods provide a ‘bag of tools’, that can be applied to problems that are “resistant to outright solution” A vast wealth of resources is available from Soft OR when adaptation is needed. Soft methods provide a ‘bag of tools’, that can be applied to problems that are “resistant to outright solution”. The more frequently used tools in the kit-bag include Soft Systems Methodology (SSM), VSM and Strategic Options Development and Analysis (SODA). The tools focus on why a new activity (or change in activity), item or system is needed. The tools then consider all the ways that the new entity could be instantiated. The effect of each option on each of the different stakeholder groups is then assessed before a final option is selected that will best meet the needs of the entity. Complex adaptions are often ‘cross cutting’ (meaning it is beyond the control or ‘information horizon’ of

FIGURE 1 THE MOST FUNDAMENTAL ELEMENTS OF A VIABLE BUSINESS MODEL

staff that are best placed to improve the system). Sometimes a change is needed because the landscape of the business has changed, as happened to many businesses during the pandemic. Sometimes a new strategic direction has been set by an owner or a senior manager. An example of changing a complex system following a strategic change by the owner follows, in which a well-known military system has been enhanced by soft methods.

THE UK RESERVE RECRUITMENT SYSTEM

The reservists within the UK armed forces are highly trained staff, some of whom are specialists, chosen to enhance key military skills and knowledge. Following the publication of the Future Reserves 2020 paper in 2011 it was concluded that the UK’s reserve forces needed significant revitalisation and re-orientation. It was assumed that the existing system could make the correction that was needed but by 2014, the trained strength of the reserves was 13,000 short of the target of 35,000. There were problems in increasing the

IMPACT | AUTUMN 2021

reserve force levels by any significant amount, despite applying considerable resource to this activity. A number of initiatives were tried but these did not increase the trained strength levels greatly and by 2014 the situation was urgent. Soft OR was used considerably within a number of government organisations to review the reserve recruitment system holistically so that the real problems with the system were highlighted. This ensured that the system could be improved, rather than risk moving the problem around. The review was delivered in the following way: 1. A rapid review was performed to structure and plan the activity 2. An enterprise level review was developed to represent the system being formed using VSM. This highlighted which organisations were performing the governance, management, audit and coherence activities. The work also showed the links between each system and the information flows that were available to decision makers. 3. Some elements of SSM were

used to consider the purpose and role of some of the key system elements. 4. Stock and flow modelling then considered how the candidate reservists move through the operations system that consists of: (1) Advertising, (2) Application, (3) Eligibility Testing, (4) Selection, (5) Training and (6) Maintaining Force Strength. Analysis determined that less than 3% of candidates that completed an application form (end of stage 2) joined the trained strength (stage 6). 5. The marketing campaign was reviewed and re-targeted Some of the key recommendations from the systems review included: • Widen the target audience based on ‘word of mouth’ for the marketing campaign so that the campaign would appeal to a much larger population group and would ameliorate, at least some of, the high loss rate seen within the Testing, Selection and Training elements of the system. • At the same time reduce the losses in the recruitment pipeline through more streamlining and monitoring of the recruitment system. • Monitor and control the system using more targeted management information and use this information to make key decisions about the system: e.g. how decisions such as the tattoo policy affect losses in the recruitment system. In 2014 any visible tattoo would disqualify a candidate reservist, and yet no data was available as to how many potential reservists this affected.

THE OUTCOME

The recommendations were made, through the military staff who were improving the reserve recruitment system, to the Secretary of State for Defence. Following the systems review significant changes were made. Within six months the trained strength figures began to increase. While there were other activities within Ministry of Defence (MoD) that were looking into the reserves, this systems review certainly contributed to the improvements. By April 2015 a clear upward trend had been established and had this continued, on a linear trajectory, would have been expected to deliver close to the original target figure of 35,000 as shown in Figure 2, even with the substantial outflow figures from the reserves.

A BROADER PERSPECTIVE

Soft OR can be applied to many different activities when some form of realignment is needed. There are many case studies, but each application is almost unique so it is difficult to demonstrate exactly how an approach will be of benefit to a new client. It does take time to apply Soft OR well.

Often senior staff prefer to take the best available solution early in a project lifecycle and only consider Soft OR once it has proved to be difficult to implement this early solution. The journey of implementation for Soft OR begins with a diagnosis of the system. This step is essential if the real, underlying problem is going to be solved. The senior staff need to be involved throughout the lifecycle of the activity. In many cases there is a pre-determined view of what the final outcome should be. This view often changes. In one high profile case a board had a ‘default’ course of action in mind. The politics of the situation prevented them taking this action, and so the original remit of the system diagnosis was ‘find a way to get past the politics’. The diagnosis activity reviewed the evidence that supported the default course of action and found little more than hearsay. An evidence base was then developed, and this suggested a subtly different course of action. Initially the board did not welcome this outcome. It was tested independently, and a change of direction came when the board accepted that the evidence base was correct. This represented a key turning point for the activity.

FIGURE 2 TRI SERVICE VOLUNTEER RESERVE STRENGTH (FROM THE 2021 PUBLISHED QUARTERLY SERVICE PERSONNEL STATISTICS)

There are often sources of tension in implementing Soft OR. The first is in the amount of time that is needed from senior staff. It is essential that any approach to a complex problem is taken with the client and the senior managers. Any pathway towards a solution that has not been developed and agreed with the client will be doomed to failure. A second source of tension is in the evidence base that is needed to support a change. Often the investment of days or weeks to prepare this is questioned. Bringing reality to a situation can lead to tensions with ‘long held’ views of some senior managers. The purpose of the evidence base is to test these views, ensuring that the way forward is based on solid evidence, not just beliefs. A turning point is always reached in any activity when the evidence base and the systems approach begin to yield value for the client. Sometimes the turning point happens very rapidly – the acceptance of the ‘ground truth’ data being correct. The only occasion when Soft OR is highly unlikely to yield value is when the system owners and senior managers are not able to participate actively in the work, and it is always best to stop an activity that is lacking key stakeholder engagement. Within government, soft OR could be applied to support many key decisions. It can improve systems and highlight risks long before they surface. Often Soft OR will improve the likelihood of a project delivering a successful outcome, but it is not a panacea. Soft OR won’t provide a “magic wand” solution to immediately fix a bad project. It will highlight where problems are. It enables decisions about continuing with a project to be taken ideally when only very small amounts of money have been committed. Soft OR can also help enhance established systems. Soft OR is having

IMPACT | AUTUMN 2021

a considerable impact and can have much more if it is well integrated with other OR methods, with a thorough knowledge of the client’s business environment.

Soft OR is having a considerable impact and can have much more if it is well integrated with other OR methods, with a thorough knowledge of the client’s business environment

of a complex project will usually return considerable value. While there are many case studies to draw on, the individual nature of a client’s situation means that it is not possible to demonstrate value simply by finding an almost identical case. If we are to grow Soft OR there is a need to integrate it with other methods, demonstrate why a client should invest time and resource in these approaches and why it is best to invest so that a more strategic outcome can be found, in comparison with the current ‘best available solution’.

CONCLUSIONS

Soft OR is immensely powerful as a technique and one implementation of this has been described in detail. It takes time to implement Soft OR. A very small investment at the beginning

IMPACT | AUTUMN 2021

ACKNOWLEDGEMENTS

Thanks are due to Christina Phillips and Jim Scholes within the OR Society PSM SIG and also to Neal Moule and Nicola Morrill at Dstl for their support.

Martin Parr is a consultant with Dstl and Guided Systems Solutions, a company that provides consultancy and training for people to access soft OR. He is also an author writing both stories and technical articles to illustrate how important systems approaches are in life. Martin read Electronic Engineering at the University of Surrey and began his career researching intelligent machines and computer vision systems. In the past 15 years he has designed and developed soft analytical approaches to manage and improve systems, mainly in the public sector. Two of the systems that Martin has enhanced had annual budgets well above £1 billion. He is a Chartered Engineer, a Fellow of the IET, a Fellow of the Operational Research Society and has a chair at Kent Business School.

HOTT

Geoff Royston

Climate change is a hot topic. And getting hotter with the recent report from the associated Intergovernmental Panel, the build-up to the United Nations summit meeting in Glasgow this November – and Bill Gates’ new book, How to Avoid a Climate Disaster. Reading this prompts me to offer a few comments from an analytical perspective, centering on three areas – understanding the issue, developing options for solutions, and coping with deep uncertainty. Bill Gates’ book ranges widely – from cars to cement, from cities to cows (a surprisingly large contributor to global warming gas emissions) – but it starts with the observation that the underlying issue is the global need to continue to make more energy available to more people, not least for improving the lives of the poorest, without releasing any more carbon into the atmosphere. Which brings me to my first analytical point, about understanding the issue.

– their atmospheric stock level will still keep rising, and the world will continue to get hotter.

MITIGATION AND ADAPTATION

As yet the world, despite two major commitments to action (the Kyoto and Paris agreements), is nowhere near on track for attaining the target of 100% reduction in carbon emissions by mid-century (see Figure 1). In a world of constantly rising demand for energy, most of which is currently produced by fossil fuels (for understandable reasons, petrol is cheaper per litre than cola, and generates ten times as much energy per kilo as TNT), getting to net zero carbon is going to be hard. Especially to do quickly; past transitions in energy sources have all taken many decades. Which means, as Bill Gates notes, that there needs to be strenuous efforts not only at eliminating net carbon emissions – mitigation – but also on coping better with climate change – adaptation. How to Avoid a Climate Disaster discusses many approaches to mitigation and adaptation. Innovative devices, redesigned processes, new policies ……… (read the book!). Which brings me to my second analytical point, about the contribution of analysis and modelling to solutions.

INFORMATION AS ENERGY?

THE ATMOSPHERIC BATHTUB

Bill Gates mentions he is funding computer modelling of all the US power grids, and that this has shown that building a national grid and associated intelligent control systems would reduce resources needed for energy production by 30 percent. (The UK was first in the world to create such a grid, back in 1935!). More generally, closer matching of supply and demand for electricity, e.g. through dynamic pricing, can improve

Until as recently as 2019 the UK was committed only to reduction in its carbon emissions, not elimination. But only elimination will do. An early simple, but compelling, illustration of why that is so came from the system dynamics modelling community (which also deserves acknowledgement for an early warning on global environmental problems, including CO2 pollution, with the report The Limits to Growth, published way back in 1972). It centres on a simple ‘bathtub’ model: if you turn the taps to slow, but not stop, the flow of water into a bath, the water level will still keep rising (unless you pull out the plug). Similarly, with just reducing net emissions of carbon gases

FIGURE 1 CURRENT v REQUIRED CARBON EMISSIONS FOR LIMITING GLOBAL WARMING TO 1.5° (BASED ON FIGURES IN THE UN ENVIRONMENTAL PROGRAMME EMISSIONS GAP REPORT 2020)

UNCERTAIN ELEPHANTS?

Climate change deniers make great play of the fact that the projections of global heating and associated events are based on models entailing sizeable uncertainties. This is, well, undeniable, but such uncertainty (which is mostly about the speed and distance of travel, not about its general direction) is not an unacknowledged elephant in the room of climate science. The current pandemic has provided a salutary reminder that complex dynamic systems can be hard to model with any precision (especially when people’s behaviour is involved) – and climate modelling is no exception. Nevertheless, in both cases, there is no sensible choice but to try, and for policy makers to pay attention to the results. Uncertainties surrounding climate change mean that we have to be prepared for a range of possible future climates for the Earth, some compatible with human society as we know it, some not. Which takes me to my last analytical point, on scenario planning, an approach that can be particularly useful when probabilities are difficult or impossible to assign and classical cost-benefit analysis is not practicable.

PRECAUTIONARY TALES

The first task in thinking and planning for uncertain futures in any domain is to envisage some credible scenarios. Then to consider strategies that are robust to as many of these

IMPACT | AUTUMN 2021

as possible; including coping with the worst-case scenario. Climate change certainly has some worst-case scenarios, particularly those involving reaching climate tipping points via positive feedback effects (a stalwart of system dynamics modelling) such as melting ice caps reducing the reflectivity of the Earth to sunlight, thus further increasing warming, leading to more melting ……… In such cases there can be calls for invoking the precautionary principle – that faced with a potentially serious hazard but where little is known – or even realistically knowable – about likelihood or impact, steps should nevertheless be taken to mitigate the risk. However, the precautionary principle is not without difficulties – what is an appropriate level of mitigation for an uncertain level of risk, especially when the related precautions may be costly or themselves be of uncertain effect? That certainly looks like the sort of situation presented by some climate change scenarios: e.g. what role, if any, should geoengineering play? Cass Sunstein (of Nudge fame and also the person responsible, under the Obama administration, for work to develop a value for the economic cost of carbon emissions) addresses this sort of issue in his new book, Avoiding Catastrophe: Decision Theory for COVID-19, Climate Change and Potential Disasters of All Kinds. While warning of the dangers of dismissing quantified approaches too readily, and hence arguing that every attempt should be made to estimate at least ball-park probabilities and impacts, Sunstein also argues there are nevertheless a few situations where the precautionary approach looks appropriate. Situations, for example, where the worst case is grave, and the costs and risks of mitigation are not too high. This could well be the position for global heating, where the worst case looks irreversibly bad and there seem to be affordable mitigation options that could if necessary be reversed. As the title of another recent climate change book, by Mike Berners-Lee, reminds us: There is No Planet B. © Penguin Random House

efficiency. Information and analysis can reduce the requirement for energy and resources – in effect substituting for them. (Though improved efficiency in energy production may not reduce carbon emissions if it just lowers energy prices: it needs to be accompanied by regulation, especially carbon pricing.) Gates also observes that steps that would give sizeable reductions in carbon emission by 2030, e.g. building gasfired electricity generating stations to replace coal-fired ones, can be very different from the things we need to do to get to zero by 2050. Exploration of the interplay and best sequencing of steps towards net zero is clearly a task where analysis and modelling can contribute. And when it comes to adaptation, good information and sound analysis also play a vital role. For example, Gates mentions early warning systems for storms and floods; phone apps that can help farmers in low-income countries to identify crop pests and diseases; and information systems and logistics models that support emergency workers in disaster situations.

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.

THE OPERATIONAL RESEARCH SOCIETY

HOME TO THE

SCIENCE + ART OF PROBLEM SOLVING

Events for Analysts We are the professional membership body for operational research analysts and decision-makers. Operational research (OR) covers a huge range of problem-structuring, problem-solving and decision-support methods and technologies.

Come to our conferences, lectures and workshops: ۸ Discover new insights and solutions ۸ Learn new skills during workshops ۸ 3JY\TWP್\NYM QNPJRNSIJI್UJJWX್

۸ Explore interests with our special interest LWTZU RJJYNSLX್್ ۸ 2JJY QTHFQ 47 HTRRZSNY^ FY TZW WJLNTSFQ LWTZU XJXXNTSX್

+NSI YMJ NSXNLMYX ^TZѣWJ XJJPNSL FY www.theorsociety.com/events

Pro Bono OR: is a charitable initiative that connects operational research analysts with the Third sector.

Operational research can help with: » Strategic planning/review » Data analysis and insight » Options appraisal » Decision-making » Process improvement » Impact measurement

TT ‫ܪ‬SI TZY RTWJ theorsociety.com/ ProBonoOR

Get in touch if: • You’re an analyst wishing to contribute to society over and above the day job. • You represent a charity wanting to achieve better insights, better decisions or better ways of working.

Impact Magazine Autumn 2021

Articles inside

Universities making an impact

Hotter

Making an impact with Soft OR

Supporting your efforts on diversity

Seen Elsewhere