Impact case study database
Search and filter
Filter by
- University of Newcastle upon Tyne
- 10 - Mathematical Sciences
- Submitting institution
- University of Newcastle upon Tyne
- Unit of assessment
- 10 - Mathematical Sciences
- Summary impact type
- Technological
- Is this case study continued from a case study submitted in 2014?
- No
1. Summary of the impact
A mathematical sciences research team at Newcastle University has developed and implemented an original Bayesian statistics forecasting solution for gas demand within Northern Gas Networks (NGN), impacting upon 2.7 million homes and businesses. The research has delivered economic, operational and educational impact over a large geographical scale within the gas distribution sector.
Specifically, the research has corroborated impact in:
Reducing gas bookings by 16%, resulting in capacity savings of over £1M per annum;
Maintaining reputation and avoiding charges of £1M per day during extreme weather conditions;
Optimising network inspection planning, avoiding wasted expenditure;
Mitigating gas storage errors, maintaining system balance and supply;
Increasing understanding and education within the gas distribution sector.
2. Underpinning research
Background
NGN deliver gas to 2.7 million homes and businesses throughout Northumberland, Northern Cumbria, Tyne & Wear, County Durham, and Yorkshire. NGN supply gas from local sites (offtakes) connected to the national network. They have a requirement to accurately forecast gas demand annually, daily and hourly. The energy sector in the UK and globally is changing to accommodate greener technologies in line with growing economic, environmental and public health concerns. Operational efficiency and decision-making can be enhanced through the improved forecasting of gas demand – the core objective of this research project.
Historically, NGN used information from Xoserve, the central service data provider for the UK’s gas market, to inform demand predictions. In 2014, NGN selected Newcastle University (NU) as a research collaborator and contributed funding [G1] to deliver a more independent, robust, and specific solution and improve the accuracy of the predictions [E1]. NU was chosen for their experience in delivering statistics-based solutions within the energy sector, and their academic expertise in modelling time-dependent data.
The research challenge
Gas demand is primarily influenced by weather and work-life patterns [P1]. There are annual and weekly periodic effects and there is a need to nest short term daily predictions within fixed annual forecasts. Public holidays have a pronounced effect on demand, typically leading to an increase in household usage but a decrease in business use. This effect often spreads into neighbouring days, referred to as a proximity effect. Traditionally, in gas demand modelling, the days over which the proximity effect occur are pre-specified in fixed windows around each holiday. This is an inflexible approach as it relies on an arbitrary decision about the existence, size and position of the windows.
[P1] describes the Newcastle development of a fully Bayesian approach to the forecasting problem. The modelling assumed a non-homogeneous hidden Markov model, whose states account for normal, holiday and prior- and post-holiday proximity effects, and a state-dependent vector autogression for demand, with seasonal and day-of-week terms and non-linear effects of a time-varying composite weather variable. Apart from holidays themselves, which are observed, duration in the other states is hidden. This was modelled as a Markov process with time-varying transition probabilities depending upon the time intervals from the most recent public holiday and to the next one, and also on the particular holiday type, with Christmas, Easter and other holidays allowed to have different effects. Time-varying precision matrices were modelled using the Cholesky decomposition approach developed for a related non-homogeneous hierarchical Markov model used by Heaps, Boys and Farrow for spatio-temporal rainfall prediction [P2]. Priors were given careful attention and Hamiltonian Monte Carlo applied for estimation. The model was calibrated using high dimensional daily demand data collected over 9 years from each of the two NGN sub-regions: the North and the North East.
Improved forecasting
An evaluation of model accuracy (predictions vs. actual data), pre-Newcastle University research and post-Newcastle University research, has demonstrated a significant improvement in accuracy of 10% (in the mean absolute daily forecast error) in the North East region and 13% in the North region [E2]. Another improvement was that the posterior distribution for the states demonstrated different patterns in the identification of proximity days between different public holidays and across different types of customers. This highlights the advantage of not fixing (allowing uncertainty in) the identification of proximity days, superior to the traditional, inflexible approach described previously [P1][P3].
The Newcastle approach to forecasting provides probabilistic distributions. This provides quantifiable uncertainty (a probabilistic range) in the forecasts, rather than simple point forecasts which have limited value.
Another feature of the modelling is forecast stability, meaning that small changes in either input covariates or in parameter values do not lead to large changes in forecast distributions. This is important in long-term forecasting, where failure to enforce stability can lead to volatile or oscillating forecasts with ever-increasing variances.
Bi-annual research implementation
The modelling and data analysis produce parameter estimates which are used for future demand forecasting. These parameter values are now bi-annually updated by NGN and embedded within their operational processes to provide a more robust, ten-year prediction at a daily level.
3. References to the research
Publications
[P1] Heaps, S.E., Farrow, M. and Wilson, K.J. (2020) Identifying the effect of public holidays on daily demand for gas. Journal of the Royal Statistical Society: Series A, 183(2), 471-492. doi.org/10.1111/rssa.12504
[P2] Heaps S.E., Boys R.J., Farrow M. (2015). Bayesian modelling of rainfall data by using non-homogeneous hidden Markov models and latent Gaussian variables. Journal of the Royal Statistical Society: Series C (Applied Statistics) 64(3), 543-568. doi.org/10.1111/rssc.12094
[P3] Wilson K.J., Heaps S.E. and Farrow M. (2016) . Demand forecasting over complex geographical networks: the case of Northern Gas Networks. In: 26th European Safety and Reliability Conference. Glasgow: CRC Press. doi.org/10.1201/9781315374987-66
Grant
[G1] NGN funding to develop the research methodology (£42K).
4. Details of the impact
The statistical research by Newcastle University has delivered economic, operational, and educational impact on a large geographical scale.
“There have been three implementations of the research within our systems in 2015, 2017 and 2019.” [E1]. The next application is due in the first quarter of 2021. The bi-annual implementations have “…demonstrated a significant improvement in accuracy of 10% in the North East region and 13% in the North region” [E2], allowing the network to operate more efficiently and prevent costs [E1]. NGN documentation, outlining operating processes, includes details of the statistical methodology embedded within NGN’s forecasting activities [E3, E4]
Impact on capacity bookings
The gas demand forecasts produced by the Newcastle University research team have impacted upon gas bookings. NGN has reduced actual gas capacity bookings by 16%, year-by-year, from 612 GWh/day to 514 GWh/day, to match the statistical forecasts [E1]. “Avoiding paying for gas that is not used has resulted in reduced capacity costs of over £1M.” [E1] *. This reduction has enabled NGN to release capacity back to National Grid which can then be allocated to other customers.
Impact during challenging periods including extreme weather
The research methodology has been robust during challenging periods, such as unseasonable weather, changeable weather and bank holiday weekends which enables NGN to avoid charges of £1m per day [E1].
In 2018, the effects of the “Beast from the East” https://www.bbc.co.uk/news/world-europe-43218229 were mitigated as a result of the statistical research. During the commencement of this unseasonable weather, NGN increased their maximum day forecast using the holiday factors. Failure to apply the increase would have resulted in NGN under-forecasting the North East local distribution zone maximum day estimate. The unadjusted forecast was 245 GWh/day, the increased maximum prediction using the holiday factors (Newcastle research) was 258 GWh/day. The actual gas demand on the 1st March 2018 was 254 GWh/day. NGN booked their gas capacity in line with the Newcastle research-estimated maximum day demand. If NGN had not increased their estimated peak demand and gas capacity being allocated accordingly, overrun charges would have occurred.
“…we have used more accurate and precise forecasts to…ii) avoid charges during operationally challenging periods such as extreme weather e.g. overrun charges of £1M could occur on one date alone if a repeat of the 2018 Beast from the East occurred under the new pricing structure”. [E1].
Impact on gas network inspection planning
The statistical research has had a direct influence on scheduling cost-effective pipeline inspection and *“… avoiding wasted expenditure.” [E1]. The research has allowed NGN to identify the optimum times of year to plan and launch pipeline inspections and conduct inline assessments. A more accurate understanding of the demand at a nuanced, daily level has allowed this to occur. Without accurate forecasts, inspection activities would be sub-optimally executed resulting in wasted expenditure - confidentiality prevents the presentation of financial cost savings.
Impact on gas storage errors
Through more accurate forecasts, the “… research has mitigate(d) gas storage errors, maintaining system balance.” [E1]. Over-forecasts and under-forecasts lead to problems within the regional storage system which can lead to detrimental impacts on the national UK storage position. This can result in less efficient systems, unavailability of resources, higher costs and reduced flexibility for the UK energy system. Underestimating or overestimating demand can impact upon the quantity of carbon gas entering the system from facilities such as bio-methane plants, and hydrogen blending facilities. The effects of over- and under-forecasting have been reduced by the Newcastle research, though it is difficult to estimate a direct quantitative improvement.
Impact on understanding & education: greater confidence and staff development
The research has had a significant impact on staff understanding and education within NGN.
The research methodology allows NGN to make informed decisions and improve situational awareness. “Management now authorise new demand forecasts with increased confidence, knowing that prior authorisations based on the outputs of the research have been successful.” [E2].
*“The research has educated some NGN staff members in the field of Bayesian statistics and provided an opportunity for continued professional development.” [E2].
Establishing UK wide gas distribution improvements
This research work resulted in the establishment of a UK-wide project in Autumn 2018 between all the UK gas distribution distributors (NGN, SGN, Cadent, Wales & West Utilities) and the School of Mathematics, Statistics and Physics at NU [E5]. The focus of this work was evaluating and improving the analysis of leakage at medium pressure across the UK gas distribution networks to ensure that the methods deployed are both accurate and cost-effective to the industry. The gas industry benefits from these statistical approaches in helping make the UK’s energy infrastructure low-carbon, sustainable and secure [E6].
5. Sources to corroborate the impact
[E1] Testimonial from the Supply Strategy Manager at NGN. Provides evidence of the financial impact and impact types.
[E2] Testimonial from a Strategy Analyst at NGN. Provides evidence of the accuracy improvement, understanding and education impact.
[E3] NGN procedure for daily demand profiling (see page 3). Provides evidence of Newcastle University’s research work embedded within NGN’s demand profiling approach.
[E4] NGN 2019 demand forecasting user-guide (see page 28). Provides evidence of Newcastle University’s research work embedded within NGN’s user guidelines.
[E5] Testimonial from an asset officer at Wales and West Utilities. Provides evidence of the UK-wide project involving all UK gas distributors.
[E6] Networks article https://networks.online/ published in February 2019. Provides evidence of the project and collaboration between NGN and Newcastle University.
- Submitting institution
- University of Newcastle upon Tyne
- Unit of assessment
- 10 - Mathematical Sciences
- Summary impact type
- Technological
- Is this case study continued from a case study submitted in 2014?
- No
1. Summary of the impact
A Newcastle University Mathematical Sciences research team developed and applied novel Bayesian statistical methods to create software for (a) predicting traffic collision hotspots and (b) evaluating site-based road safety interventions. Corroborated impacts have occurred on a regional, national and global scale, including:
a significant contribution to reducing average annual traffic casualties from 514 to 436 in North Yorkshire, with £22.5M estimated accident prevention savings;
influencing traffic and road safety policy in over 60 countries through the International Transport Forum;
impacting the design of a low emission zone in Lisbon, Portugal;
developing new software applications with a sales value of €1.1M for 140 global organisations in 40 countries through PTV Group, Germany.
The research benefits address the 2030 Agenda for Sustainable Development Goals 3.6 and 11.2 directly, and help towards achieving the United Nations-supported global initiative of Vision Zero - “ the aim of achieving a highway system with zero accidents or fatalities involving road traffic”.
2. Underpinning research
Background
Each year in the UK there are in the region of 200,000 injuries, 20,000 serious injuries and
2,000 deaths because of road traffic accidents (RTAs). Since 1998 the UK government has considered road safety (speed) camera placement to be an important part of its strategy to reduce RTAs and consequent casualties. Road safety cameras are known to
be effective in reducing RTAs but they are expensive to maintain and perhaps more importantly they are most effective when strategically and adaptively sited. This leads to the need for effective collision prediction analysis, to identify future hotspots and to determine the consequences of site-based interventions such as moving cameras between locations.
The research challenge
The assessment of the effectiveness of road safety measures necessarily requires the analysis of observational data, bringing with it consequent difficulties in interpretation. For example, the problem of selection bias is well known in the road safety literature. A new safety measure (e.g. a speed camera) implemented at a site because of an unusually high number of accidents or casualties might seem subsequently to be effective simply because of the effect of regression to the mean (RTM).
Empirical Bayes (EB) methods were therefore developed in the 1990s to forecast RTA frequency in the absence of intervention, usually based on a simple Poisson random effects model with gamma prior distributions and an exchangeability assumption across all sites being analysed. The EB approach to analysis became the gold standard internationally for separating RTM effects from genuine treatment effects of road safety measures.
Research at Newcastle University (NU) began in 2008 because of concerns regarding the effectiveness of the then-standard procedures. In collaboration with the Northumbria Safer Roads Initiative (NSRI) and the Newcastle upon Tyne Hospitals NHS Trust, Dr Lee Fawcett (School of Mathematics, Statistics and Physics (MSP)) and Dr Neil Thorpe (School of Engineering) developed a Fully Bayesian (FB) approach that incorporated multivariate traffic flow predictors, flexible prior distributions, temporal trends, more realistic uncertainty quantification and sensitivity analyses [P1]. The methods can be applied alongside linked data from police collision and NHS casualty records to allow the financial consequences of proposed safety measures to be explored, which is of substantial interest to NHS secondary healthcare providers [P1, P2]. Since 2014, Mr Joe Matthews (School of MSP) has contributed to the work, first in his role as a PhD student (jointly supervised by Fawcett and Thorpe) and more recently in his lecturer role. Various grants [G1-G6] have funded the research.
The modelling has subsequently been extended and refined, with more focus on predictive
capability and uncertainty quantification in collaboration with the German traffic and logistics global software company PTV Group [P3]. Most recently, further substantial improvements over EB methods have been demonstrated in collaboration with transport researchers in Texas and Shanghai [P4].
Interdisciplinary research
The mathematical sciences research has enabled advancements within the field of transport engineering and has attracted widespread interest amongst the international traffic management community. Subsequent to the first publications, Fawcett, Matthews and Thorpe have given invited talks to the annual US Transportation Research Board Conference (every year since 2016), the London Transport Practitioners' Meeting (2017), and the Road Safety GB Conference in Birmingham (2018). In 2017 they led a Latin American Knowledge Transfer Workshop in Mexico City on Statistical Methods and Software for Predicting Road Traffic Collisions. They have given invited presentations to, and established memorandums of understanding with, the Abu Dhabi Police Force and the Dubai Road Transport Authority (both 2017), and to the Deputy Mayor for Mobility, Safety, Economy & Innovation and his team in Lisbon (2018). In May 2018 they delivered an invited workshop on Software Tools for Road Safety Data Analysis at the New York City Department of Transportation. Within the UK they have been invited to visit several national organisations, local authorities and police forces including Highways England, Gateshead Council and North Yorkshire Police.
Software development
NU’s Bayesian statistics research is the bedrock of RAPTOR, a software application for predicting collision hotspots and evaluating site-based road safety interventions [P5, P6]. Before RAPTOR, no such software existed for road safety practitioners. The mathematical methods are embedded in PTV’s VISUM Safety, a commercial software product that allows planners and road safety analysts to assess the most critical points in a transport network.
3. References to the research
[P1] Fawcett, L. and Thorpe, N. (2013) Mobile safety cameras: estimating casualty reductions and the demand for secondary healthcare. Journal of Applied Statistics, DOI: 10.1080/02664763.2013.817547. ( http://www.tandfonline.com/doi/abs/10.1080/02664763.2013.817547)
[P2] Thorpe N, Fawcett L. (2012) Linking road casualty and clinical data to assess the effectiveness of mobile safety enforcement cameras: a before and after study. BMJ Open, 2(6), e001304. ( http://bmjopen.bmj.com/content/2/6/e001304?ct)
[P3] Fawcett, L., Thorpe, N., Matthews, J. and Kremer, K. (2017). A novel Bayesian hierarchical model for road safety hotspot prediction. Accident Analysis & Prevention, 99, pp.262-271. ( http://www.sciencedirect.com/science/article/pii/S0001457516304341)
[P4] Guo, X; Wu, L; Zou, Y; Fawcett, L. (2019) Comparative Analysis of Empirical Bayes and Bayesian Hierarchical Models in Hotspot Identification. Journal of the Transportation Research Board. DOI: 10.1177/0361198119849899. ( https://doi.org/10.1177/0361198119849899)
[P5] Matthews, J.T., Newman, K., Green A.C., Fawcett, L., Thorpe, N., and Kremer, K. (2019). A decision support toolkit to inform road safety investment decisions. Municipal Engineer, 172, 1, pp. 53-67. (winner of the ICE Publishing James Hill Prize) ( https://www.icevirtuallibrary.com/doi/full/10.1680/jmuen.16.00057)
[P6] Road Safety Analysis. Software application for predicting collision hotspots and evaluating site-based road safety interventions. http://roadsafetyanalysis.org/raptor/
Grants
[G1] Two grants from Northumbria Safer Roads Initiative/Gateshead Council to develop RAPTOR (£40K)
[G2] Three internal University Research Council pump prime grants to enable impact activity (£15K)
[G3] Funding over 5 years from the Northumbria Safety Camera Partnership (NSRI) (£75K)
[G4] Health impacts study funding from the NSRI, collaborative with Northumbria NHS Trust (£25K).
[G5] EPSRC PhD award (£70K)
[G6] EPSRC NPIF PhD award (£78K)
4. Details of the impact
Health, economic and commercial benefits have occurred at a regional, national, and global level. This case study details examples with corroborated impact from the UK, Europe, South America and within the International Transport Forum (ITF).
Impact in North Yorkshire (NY)
Following NU RAPTOR research in 2016, North Yorkshire Police (NYP) expanded their fleet of safety camera vans to further reduce the number of collisions, deaths, and serious injuries on the region’s roads [E1]. *“…independent research by academics at Newcastle University shows an estimated 20% reduction in casualties owing specifically to the presence of mobile safety camera vans…In 2017 six new, more agile vehicles were introduced by the police…” - *NY Police & Crime Commissioner [E2].
Three-year “before and after” analyses either side of the 2017 £107K fleet expansion, demonstrates the following benefits [E3]:
a 24% reduction in casualties at 22 specific locations;
a significant contribution to estimated accident prevention savings of £22.5M p/annum across the whole North Yorkshire region.
The Police, Fire and Crime Commissioner states: “Since 2017, the augmented fleet has enabled greater presence on more high-risk routes to influence speeding and anti-social road use. It has also allowed us to provide visibility and reassurance to many communities which had not previously had a safety camera presence.” [E3].
“The Newcastle research was significant in clarifying the rationale for the expansion of SCVs (Safety Camera Vehicles) and therefore in contributing to a reduction in the annual number of KSI [Killed or Seriously Injured] incidents. This has reduced on average from 514 casualties per annum to 436, bringing an estimated accident prevention saving of £22.5M p/annum and significant societal health and economic benefits.” [E3].
Impact in Northumberland and Tyne & Wear
NU’s RAPTOR software is used by the Northumbria Safer Roads Initiative (NSRI).
*“Since 2018 we have applied RAPTOR to identify traffic collision hotspots in the Northumbria Police Force area (Northumberland and Tyne & Wear) and guide the allocation of enforcement resources… RAPTOR is a key component of traffic strategy and operations and is now an established part of our year-on-year decision-making process”* **– **Senior Transport Planner, NSRI [E4].
All 130 mobile safety camera locations within the region have been assessed by RAPTOR for continuing enforcement. There is an estimated collision event reduction of 38 over a two-year period for the whole of the region since RAPTOR analysis began. Newcastle research has made a significant contribution to estimated collision reduction savings of approximately £4M for the years 2018 and 2019 [E4].
NSRI also use RAPTOR to evaluate the effect of road safety interventions. Pre- and post-analyses of three-year durations either side of the RAPTOR-driven safety camera redeployment demonstrates a significant decrease in casualties of 33% at four sample locations, resulting in estimated accident prevention savings of £235K over a three-year period for these sample sites [E4].
Impact in Lisbon, Portugal
In 2018, the Municipality of Lisbon, Portugal and NU agreed a memorandum of understanding to develop and improve road safety in Lisbon. NU researchers undertook an analysis to assess the effects of road safety cameras on casualty reduction within the city. Subsequently, NU researchers used their statistical methodology embedded in PTV’s VISUM Safety software to predict the safety impacts (“what-if scenarios”) of a new low emission zone scheme in the Baixa-Chiado district of Lisbon. *“The results from the assessment impacted upon the design of the scheme. The Low Emission Zone was presented by the Mayor in January 2020, but unfortunately had to be temporarily suspended due to the (Covid-19) pandemic situation. The Newcastle research methodology has been playing a role in the design of a safer Lisbon…” – *Deputy Mayor, City of Lisbon [E5]. It is expected that benefits from the scheme would now have been evident without the outbreak of Covid-19.
Impact in Bolivia
In 2016, NU researchers delivered a series of road safety workshops about their methods for analysing collision data, the rationale behind their research and a demonstration of the software solutions in Santa Cruz, Bolivia. The work highlighted the benefits of collision prediction and scheme evaluation from the appropriate interrogation of collision data. One of the significant outputs from the event was a Road Safety Charter signed by participants calling for an increased investment in road safety in Bolivia. Based on this Road Safety Charter, which included specific NU research, the Global Road Safety Facility (GRSF) provided the CAF bank a grant of US$200,000 to enhance the development of road safety strategies in four cities in Bolivia: Santa Cruz, Tarija, La Paz and El Alto [E6].
*“An understanding of the Newcastle research, specifically i) the methodology for the accurate analysis of collision and casualty data to identify collision hotspots and ii) the evaluation of road safety schemes such as speed cameras, has enabled the cities to develop a system of self-funding road safety strategies.” – Road Safety Advisor, GRSF/CAF [E6].
Global Commercial Impact
Corroborated International impact is demonstrated through collaboration with PTV Group, a global German company specialising in software solutions for traffic and mobility. Statistics-based algorithms, used within RAPTOR, have been embedded within VISUM Safety software [E7] [E8]. “Algorithms developed by Newcastle University have been integrated within our VISUM safety software from 2017 through to the latest edition VISUM Safety 21 , http://vision-traffic.ptvgroup.com/en-uk/products/ptvvisum-safety/ , which is licensed currently to 140 organizations in 40 different countries, with a total of 3700 licenses so far.” “…Newcastle University research has helped us to achieve estimated license sales of our VISUM Safety Module at an estimated of License value of approximately €1,1M”– Vice-President Business Development & New Mobility, PTV Group [E9].
Global Policy Impact
The International Transport Forum (ITF), within the Organisation for Economic Cooperation and Development, is an intergovernmental organisation with 61 member countries. The ITF acts as a platform for discussion and pre-negotiation of policy issues across all transport modes at a global level. NU’s novel collision prediction research is detailed in the section on ‘proactive network management’ within the ITF policy document, “New Directions for Data Driven Transport Safety”, published in 2019 [E10]. Subsequently, the research is used by global member countries of ITF to influence future transport policies and the effective use of data to drive safety decision making.
5. Sources to corroborate the impact
[E1] Online media article . https://www.yorkpress.co.uk/news/ryedale/15223085.six-new-mobile-speed-cameras-launched-in-north-yorkshire/ Provides evidence of the research collaboration and NU research leading to the expansion of the safety camera fleet in North Yorkshire.
[E2] Public report by the North Yorkshire Police & Crime Commissioner: Making North Yorkshire’s Roads Safer https://www.northyorkshire-pfcc.gov.uk/content/uploads/2018/09/Report-Making-North-Yorkshires-Roads-Safer.pdf e.g. page 4 (foreword by the Commissioner) and pages 30-31. Provides evidence of the research collaboration and NU research leading to the expansion of the safety camera fleet.
[E3] Testimonial from the North Yorkshire Police, Fire & Crime Commissioner. Provides evidence of the investment in safety cameras, casualty reductions and estimated accident prevention savings.
[E4] Testimonial from the NSRI. Provides evidence of the casualty and collision reductions, and estimated accident prevention savings in Northumberland and Tyne & Wear.
[E5] Testimonial from the Deputy Mayor of Mobility, Safety, Economy, and Innovation in Lisbon. Provides evidence of the impact on the design of the low emission zone scheme.
[E6] Testimonial from the GRSF/CAF Road Safety Advisor. Provides evidence of the grant for the development of road safety strategies and influence on Bolivian cities.
[E7] PTV Group and VISUM Safety https://discover.ptvgroup.com/road-safety-evaluation-prediction Provides evidence of the research collaboration and the integration of the statistics research within VISUM Safety.
[E8] Media Article from PTV Group. https://www.iamigniting.com/neilandteam/ Provides evidence of the NU research team and collaboration with PTV Group.
[E9] Testimonial from PTV Group. Provides evidence of the research collaboration, the integration of the statistics research within VISUM Safety and the increase in revenue.
[E10] International Transport Forum (ITF) policy document. https://www.itf-oecd.org/sites/default/files/docs/new-directions-data-driven-transport-safety_0.pdf pages 28 and 30. Provides evidence of the research influencing international transport policies.
- Submitting institution
- University of Newcastle upon Tyne
- Unit of assessment
- 10 - Mathematical Sciences
- Summary impact type
- Technological
- Is this case study continued from a case study submitted in 2014?
- No
1. Summary of the impact
A School of Mathematics, Statistics and Physics research team has delivered significant impacts within Small and Medium-sized Enterprises (SMEs) through the innovative design and application of statistical methods and data analytics.
Economic and commercial, as well as societal and environmental impacts, have occurred within the marine, utilities, automotive, social housing, and service sectors, corroborated by evidence. Total benefits include:
increases in annual sales of greater than £4.5M;
increases in annual exports of greater than £4M;
increases in annual profits of greater than £1M;
the creation of intellectual property;
the creation of at least 13 new jobs.
2. Underpinning research
The SME challenge
Developments in statistical and data analytics technologies are relatively easily disseminated through academic communities and large, well-resourced organisations. It is more difficult for a SME to become aware of and subsequently exploit the best available techniques. SMEs contribute 50% to the UK economy and employ 82% of the UK workforce. Within Europe, SMEs employ around 65 million people.
A mathematical sciences research team at Newcastle University (NU), led by Dr Shirley Coleman and including Dr Jian Shi, has developed and transferred statistics-based data analytics tools into a wide range of industrial and business SMEs [P1-P4]. Initially supported by €750,000 from the European Regional Development Fund [G1] and €500,000 from the European Social Fund [G2], and subsequently through five Knowledge Transfer Partnerships within the REF period [G3-G7], the group has, inter alia, embedded generic (e.g. [P5]) and targeted (e.g. [P6, P7, P8]) Newcastle research within more than 100 SMEs. The following case studies provide examples.
Marine sector
In collaboration with Royston Diesel Power, using research reported in [P1] and [P3], location data from company systems was combined with open access meteorological data on tide and wind to analyse fuel consumption and emissions for active marine engines. Summed rank cumulative sum techniques [P6] and other original statistical process control methods [P7] were applied to identify process step changes that led to negative environmental impacts and to build a predictive model to allow efficient scheduling of services.
Utilities sector
In collaboration with Advanced Engineering Solutions (AES), Gaussian process regression methodology [P5] was extended for detection and sizing of pipeline defects. AES detection equipment records changes in magnetic flux as eight external sensors traverse regions of interest. The flexible Gaussian process method automatically maps the signals into measures of remaining wall thickness, allowing defects to be identified without strong parametric assumptions. Empirical Bayesian methods are used for hyperparameter estimation and principal components for multivariate summaries.
Automotive sector
In collaboration with Rain Data, a portfolio of data science reliability methods for the automotive industry was developed and applied in [P8], based on big data such as 14 million part return records and 6 million replacement part records. The portfolio includes a novel funnel plot control chart and normal mixture survival analysis. The methods allow the SME to provide the sector with near real time reliability information for parts and vehicles, for use in product matching, stock control, pricing and planning.
Social housing sector
Social housing accounts for 50% of rental properties in the UK. In collaboration with Orchard Information Systems, Coleman and colleagues [P3] developed methods to exploit the large reservoir of data on rent balances, property repairs and empty properties to identify tenants in danger of falling into arrears. The methods blend machine learning and data visualization techniques, specifically random forests, T-distributed stochastic neighbor embedding and SHAP plots.
Service sector
In collaboration with Enzen Global, official statistics and other publicly available open data were integrated with company data [P1] to quantify the effect of improved localized weather measurements for predicting gas consumption. Structural equation modelling and Bayesian network analysis evaluated the relationships between socio-economic factors and gas consumption.
3. References to the research
[P1] Coleman, S (2016). Data mining opportunities for small to medium enterprises from official statistics. Journal of Official Statistics 32, 849-866. doi.org/10.1515/jos-2016-0044
[P2] Coleman, S.Y., Gob, R., Manco, G., Pievatolo, A., Tort-Martorell, X. and Reis, M. (2016). How can SMEs benefit from big data? Challenges and a path forward. Journal of Quality and Reliability Engineering International 32, 2151–2164. doi.org/10.1002/qre.2008
[P3] Vicario, G. and Coleman, S.Y. (2020). A review of data science in business and industry and a future view. Applied Stochastic Models in Business and Industry 36, 6-18. doi.org/10.1002/asmb.2488
[P4] Coleman S.Y. (2019) Data science in Industry 4.0. Mathematics in Industry 30, 559-566. doi.org/10.1007/978-3-030-27550-1_71
[P5] Shi JQ, Murray-Smith R, Titterington DM. (2005) Hierarchical Gaussian process mixtures for regression. Statistics and Computing 15(1), 31-41. doi.org/10.1007/s11222-005-4787-7
[P6] Stewardson, D.J. and Coleman, S.Y. (2001). Using the summed rank cusum for monitoring environmental data from industrial processes. Journal of Applied Statistics 28, 469-484. doi.org/10.1080/02664760120034180
[P7] Coleman, S. Y., Arunakumar, G., Foldvary, F. and Feltham, R. (2001). SPC as a tool for creating a successful business measurement framework. Journal of Applied Statistics 28, 325-334. doi.org/10.1080/02664760120034063
[P8] Smith, W.S., Coleman, S.Y., Bacardit, J. and Coxon, S. (2019). Insight from data analytics with an automotive aftermarket SME. Journal of Quality and Reliability Engineering International 35, 1396– 1407. doi.org/10.1002/qre.2529
Grants
[G1] European Regional Development Fund to engage and transfer research into SMEs (€750k).
[G2] European Social Fund to engage and embed statistical skills within SMEs (€500k).
[G3] KTP programme with Royston (£134k). 2014.
[G4] KTP programme with Enzen Global (£134k). 2014.
[G5] KTP programme with Orchard (£140k). 2016.
[G6] KTP programme with Rain Data (£133k). 2016.
[G7] KTP programme with Royston (£137k). 2017.
4. Details of the impact
Economic, commercial, societal and environmental impacts have occurred within the marine, utilities, automotive, social housing, and service sectors.
Impact within the Marine sector
Royston Diesel Power are specialists in diesel power, sales and repairs, with a significant customer-base within the marine and offshore industry. The research work has led to the following benefits since 2016 [E1]:
an increase in annual sales turnover of £4M;
an increase in annual exports of £4M;
an increase in annual profit of approximately £1M;
the creation of 5 new jobs within the business;
an investment of greater than £1M in new software and hardware implementation and support.
Statistical data analytics research was used to develop modules for Royston’s fuel monitoring product Engine i. “The research work enabled us to develop new opportunities via our existing customer base and global shipping databases. The data analytical capability has added significant value to customer data through the creation of high-value operational intelligence for the customer.” [E1]. As a result of the research collaboration with Newcastle University, Royston began a 3- year collaborative R&D programme with Innovate UK with a focus on “whole vessel energy monitoring”. The value of the programme was £1.2m.
Impact within the Utilities sector
Advanced Engineering Solutions (AES) are specialist pipeline and pipeline equipment engineers. The research has generated annual turnover of £275,000 [E2] for AES through the capturing of new service contracts. Modern statistical methodology has become an integral part of AES’s analysis processes.
The well-recognised AES pipeline condition assessment process is used on a global basis.
The statistical analysis method for pipeline defect identification, detection and sizing has been implemented and utilised within our processes, used by water companies across the UK, and is becoming a key component within technological services provided to our collaborative organisations in France (Suez) and Australasia (Asset Integrity Australasia).” [E2].
Impact within the Automotive sector
Rain Data specialise in data and cataloguing services for the automotive sales after-market. The research work achieved the following benefits [E3a, E3b]:
an increase in annual sales turnover of £60k (estimated to be £180k within 3 years);
an increase in annual profit of £35k (estimated to be £95k within 3 years);
the creation of 2 new jobs.
Newcastle University research has improved Rain Data services to their customers through establishing data analytics capacity within the business. The implementation of data analytics tools has increased levels of data automation and accuracy; avoided the outsourcing of data cleansing and analytics activities; enhanced the profiling of stock; enabled better prediction of car trends and likely mechanical failures; reduced customer returns thus increasing client satisfaction; and improved stock level efficiency [E3a].
Impact within the Social Housing sector
Orchard Information Systems provide property management solutions and services to housing organisations and councils throughout the UK. Social housing research impact benefits include [E4a, E4b]:
the development of Intellectual Property (IP) related to predictive models for rent arrears and void properties. This IP has been embedded into Orchard products and services. The prediction of voids is a sector first;
the creation of 2 new jobs;
supporting the development of Income Analytics, Orchard’s flagship product.
Data analytics has become a key component of new solutions and business insight for Orchard clients. The generation of new products has strengthened Orchard’s position in the sector and provided competitive advantage.
Impact within the Service sector
Enzen Global provide consulting, technology, engineering and innovation services to the energy and water industries. Core benefits, as a result of the creation of an improved data analytics capability through the research, are [E5, E6a, E6b]:
an increase in annual sales turnover of £292,000
an increase in profit of £70,000
the creation of 4 new jobs
The subsequent year-on-year increase in sales turnover and profit could be up to 10 times higher, although Enzen have not undertaken a full cost-benefit analysis.
The research has been embedded within Enzen’s Knowledge Centre and enabled the company to expand their range of business products and services. The Newcastle University data analytics research has enabled applications to diverse challenges within the utilities service sectors across the UK [E5]. Work with Northern Gas Networks established improved gas demand estimations by using local rather than regional weather measurements. A collaboration with Scotia Gas Networks identified key relationships between socio-economic factors and residential demand. A project with Wales and West Utilities resolved public reported escapes (PREs) of gas yielding financial optimisation benefits.
“…this work with Newcastle University opened up a new revenue stream for Enzen and a new way of thinking about data and the opportunities that it provides”. [E5].
5. Sources to corroborate the impact
[E1] Testimonial from the Chief Executive Officer of Royston Limited. Provides evidence of the financial impact.
[E2] Testimonial from the Technical Director of Advanced Engineering Solutions (AES). Provides evidence of the financial impact and global reach of the research.
[E3a] Rain Data KTP final report. Provides evidence of the main achievements of the research (pages 2-3), financial benefits (pages 5-6) and new jobs created (page 7). [E3b] Declaration by Partners. Provides evidence of completion of project deliverables.
[E4a] Orchard KTP final report. Provides evidence of developing Income Analytics (page 7), new jobs created (page 9) and developing IP (page 12). [E4b] Declaration by Partners. Provides evidence of completion of project deliverables.
[E5] Testimonial from the Client Partner at Enzen Global. Provides evidence of the main benefits of the research.
[E6a] Enzen Global KTP final report. Provides evidence of the financial impact (pages 7-8) and new jobs created (page 10). [E6b] Declaration by Partners. Provides evidence of completion of project deliverables.
- Submitting institution
- University of Newcastle upon Tyne
- Unit of assessment
- 10 - Mathematical Sciences
- Summary impact type
- Technological
- Is this case study continued from a case study submitted in 2014?
- No
1. Summary of the impact
Newcastle University (NU) Mathematical Sciences research has created Jumping Rivers PLC, a Newcastle-based analytics company specialising in data science and machine learning. The research has enabled commercial, economic and education impacts across diverse sectors throughout the UK and overseas. Jumping Rivers has directly met major objectives of the North East Strategic Economic Plan, such as “creating more and better jobs” and providing growth within the digital sector.
Specifically, NU research has created impact with corroborated evidence through:
establishing a new company;
enabling an annual turnover by 2019/20 of £638K;
achieving a year on year increase in operating profits from £4K in 2016/17 to £184K in 2019/20;
creating 13 jobs;
delivering solutions in sectors such as biotechnology (e.g. Fujifilm Diosynth Biotechnologies) and utilities (Northumbrian Water Limited).
2. Underpinning research
E-Science Projects
Starting in 2001, colleagues in the School of Mathematics and Statistics (now the School, of Mathematics, Statistics and Physics) collaborated with researchers in Computing Science and Biomedical Sciences on several BBSRC E-science and bioinformatics research programmes [G1-G4]. Relevant aspects of two of these projects (BASIS and CaliBayes) are summarised below. The underlying premise of all projects was a commitment to Bayesian modelling (e.g. [P1]) and a recognition of the need to provide complete packages of data science techniques. A common theme was the implementation of computationally intensive methods running on remote computer clusters. This nascent cloud computing pre-dated the rise of the current ubiquitous cloud (Amazon AWS was launched in 2006), enabling collaborative stakeholders to use this technological solution much earlier than what was commercially available at the time.
The BASIS (Biology of Ageing e-Science Integration and Simulation) project [P1] was a flexible web-service based modelling system which was developed to allow parallel and distributed approaches to simulation and inference [P2]. It was effectively a set of web services that allowed users to incorporate simulation tools into their workflows [P3]. In 2002, web-services were very much in their infancy and connecting disparate services was challenging. Today, many Jumping Rivers (JR) data science projects consume data from a variety of application programming interface (API) endpoints, e.g. weather data. The key to combining different data sources is a standard interface. The BASIS project provided the foundation for these skills by developing the SBML framework that enables the exchange of statistical models and data sets [P4].
The second project, CaliBayes, developed techniques for calibrating complex stochastic models against experimental data, with careful attention to the selection of parameter combinations at which to run expensive and time-consuming simulators [P5]. Crucially, these techniques were running statistical algorithms in the pre-cloud. Indeed, figures 1 & 2 in [P5] provide a data pipeline architecture, that includes API requests, load balancing and secure data access points, that underpin the architecture currently used by JR for data pipelines within their industrial and business client base.
Creation of a data analytics company
The primary aim of research in statistical methodology and data science is to produce new or improved techniques for use in practical applications. While academics may be able to install and experiment with the latest statistical software, commercial/enterprise users face significant challenges in implementing new techniques. These hurdles include accessing data, maintaining data security, upgrading statistical software, selecting the methods, applying model diagnostics, and moving from a testing to a production environment.
Jumping Rivers PLC is an independent company that was formed in 2016 because of the NU research and alongside an increased demand for support in implementing data science and statistical methods. The Newcastle research [P1-P5] and programmes [G1-G4] provided early recognition that science could best leverage increased computing power through integrated approaches that cover all aspects of the data pipeline, from data entry and storage through Bayesian modelling and computationally intensive statistical inference on remote computer clusters, to visualisation and prediction.
The research focussed on applications in systems biology and the biology of ageing, but the techniques are generic. During the research and subsequently, it became clear that diverse organisations needed integrated data science approaches. With the School's core focus being research and teaching, an opportunity was identified to commercialise the research. Gillespie, who had been a research associate on one of the e-science grants and was subsequently a Lecturer then Senior Lecturer in the School of Mathematics and Statistics, was frequently consulted from around 2010 by external organisations seeking to implement their own circle-of-data systems. Recognising that demand was high and likely to increase, Gillespie reduced his University employment to part-time in 2016 and co-founded JR.
3. References to the research
Publications
[P1] Bayesian inference for a discretely observed stochastic kinetic
model: Boys RJ, Wilkinson DJ & Kirkwood TBL (2008). Statistics and Computing 18, 125-135 https://doi.org/10.1007/s11222\-007\-9043\-x
[P2] Towards an e-biology of ageing: integrating theory and data: Kirkwood TBL, Boys RJ, Gillespie CS, Proctor CJ, Shanley DP, & Wilkinson DJ (2003). Nature Reviews Molecular Cell Biology 4, 43–249. https://doi.org/10.1038/nrm1051
[P3] Tools for the SBML Community: Gillespie CS, Wilkinson DJ, Proctor CJ, Shanley DP, Boys RJ, Kirkwood TBL (2006). Bioinformatics 22, 628–629. https://doi.org/10.1093/bioinformatics/btk042
[P4] Systems Biology Markup Language (SBML) Level 3 Package: Distributions, Version 1, Release 1: Smith LP, Moodie SL, Bergmann FT, Gillespie CS, Keating SM, Konig M, Myers CJ, Swat MJ, Wilkinson DJ, Hucka M (2020). Journal of Integrative Bioinformatics 17, 20200018 https://doi.org/10.1515/jib\-2020\-0018
[P5] CaliBayes and BASIS: integrated tools for the calibration, simulation and storage of biological simulation models: Chen Y, Lawless C, Gillespie CS, Wu J, Boys RJ, Wilkinson DJ (2010). Briefings in Bioinformatics 11, 278–289. https://doi.org/10.1093/bib/bbp072
Grants
[G1] Bayesian inference for stochastic kinetic genetic regulatory networks . BBSRC, BIO14454, 2001, £120K, Boys, Kirkwood, Wilkinson.
[G2] Biology of Ageing e-Science Integration and Simulation System (BASIS). BBSRC BEP17042, 2002, £442K, Kirkwood, Boys, Proctor, Wilkinson.
[G3] Integration of Grid-based Post-genomic Data Resources Through Bayesian Calibration of Biological Simulators. BBSRC BBS/B/16550, 2005, £352K, Wilkinson, Boys, Kirkwood.
[G4] ComparaGrid - Enabling GRID Technologies for Comparative Genomics. BBSRC BBS/B/17158, 2005, £337K, Wipat, Boys, Pocock, Watson, Wilkinson.
4. Details of the impact
Creation of a data science company
Jumping Rivers is a Newcastle-based analytics company specialising in data science and machine learning [E1]. The company commenced trading in October 2016 and is co-located on the Newcastle Helix site alongside the multi-million-pound National Innovation Centre for Data. JR work impacts on businesses in fields as diverse as biotechnology, finance, DigiTech, manufacturing, chemicals, utilities, and healthcare.
“Newcastle University’s research has been fundamental in creating our company Jumping Rivers, providing the company’s core infrastructure and enabling a diverse range of services to be offered to our customers”. [E2].
Meeting North East challenges
Compared to other parts of the UK, the North East has traditionally experienced fewer and lower quality job opportunities. JR has directly met several objectives of the North East Strategic Economic Plan [E3] such as “creating more and better jobs” and providing growth and services within the digital sector. JR is a highly skilled team with individuals holding qualifications including four PhDs in statistics and six MScs in statistics/computing science. All staff are situated in the North East delivering projects on a global level. The ten technical members of staff specialise in areas such as simulation tools, optimised workflows, cloud-computing, statistical inference, and parallel and distributed approaches.
Impacts
Since the financial year 2016-17, JR [E2, E4]:
annual turnover has increased from £49K to £638K in 2019/20
operating profits have increased from £4K to £184K in 2019/20
staff numbers have increased to a headcount of 13 (12.4 FTE) as of March 2021
has delivered projects for over 100 organisations in Europe, North America, and Africa
has provided pro-bono sponsorship for over 40 global events, for example in the Ivory Coast, Uganda, and South Africa
The techniques created through Newcastle University e-science projects have been developed so that JR now offer a complete service through ongoing support, applied research, training, and consulting. Projects include establishing effective data infrastructures, running robust computational statistical models, generating reproducible data analysis reports, and enabling statistical inference. All these approaches enhance business decision making through the improved use of data.
JR has delivered impact across many areas; two exemplars are provided below covering the biotechnology and utilities sectors:
Fujifilm Diosynth Biotechnologies (FDB) is a global-based organisation specialising in cell culture, microbial fermentation, and gene therapies. Since 2018, JR has collaborated with FDB to optimise their data collation, usage, and analysis process. FDB has multiple research groups situated in a variety of locations across the world, these teams regularly share large complex data sets. Working closely with a FDB team, JR built a cloud-based data pipeline to conduct process and analytical risk assessments. The risk assessments, built around a Failure Modes and Effect Analysis platform, enable FDB and their clients to identify, numerically score and visualise risks to manufacturing on the journey to commercialising therapeutic drugs.
The digital application is used by all FDB sites globally and has increased their proficiency, elevating client facing capabilities. There are numerous benefits to this work including; more effective prioritisation of tasks in late stage development; reduced chances of errors; enhanced communication of risk to clients; and improved demonstration of risk management systems to global regulatory health authorities including the Food and Drug Administration and the European Medicines Agency. The work has helped FDB to demonstrate their recognition of the importance of effective risk management.
“Jumping Rivers and Newcastle University have both contributed significantly to our global operations”. [E5].
Northumbrian Water Limited (NWL) serves 1.3 million properties in the North East of England, and 794,000 properties in Essex and Sussex, primarily through water and sewage services. In recent years, reducing leakage and interruptions to supply have been high priority targets for NWL, its customers, and Ofwat (the water industry regulator).
Since 2018, JR have developed and integrated within NWL data-driven solutions to address leakage and supply interruptions problems. Development of the e-science application was on the Microsoft Azure Cloud, which matched the internal cloud of NWL, and allowed constant feedback to the new application. Data was gathered via APIs from a variety of internal and external sources. Reducing the time taken for 35 engineers to access the required data from two hours to 15 minutes has resulted in annual savings, reduced leakage and reduced interruptions.
The insights provided by JR have been valuable for enhancing NWL service delivery, primarily for NWL customers, but also for NWL themselves, as high performance in these areas is linked to considerable financial incentives.
“Jumping Rivers have demonstrated the value and capabilities of new technologies which we now are looking to adopt further within NWL. Their expertise spans a range of analytics, data science and deployment technologies -including the Microsoft Azure Cloud - which they leverage to build impactful and insightful Proof of Concept models and analyses”. [E6].
NU’s research through JR has provided the bedrock of these case studies and many other projects, enabling commercial, economic, and educational impact within diverse contexts.
5. Sources to corroborate the impact
[E1] Website for Jumping Rivers. www.jumpingrivers.com Provides evidence of the
existence of the company and its services to business and industry.
[E2] Testimonial from the Director of Jumping Rivers. Provides evidence that Newcastle
research created the company, financial impacts, and impact types.
[E3] North East Strategic Economic Plan. https://www.nelep.co.uk/wp-content/uploads/2019/02/executive-summary.pdf Provides evidence of regional objectives that the establishment of Jumping Rivers has met.
[E4] Jumping Rivers Annual Accounts for the year ended 2020 (page 4). Provides evidence
of the annual turnover and operating profits financial impact.
[E5] Testimonial from a Staff Scientist (Statistics and Computational Engineering) at FDB. Provides evidence of the impact of Jumping Rivers and Newcastle University within the biotechnology sector.
[E6] Testimonial from an Intelligence Manager (Intelligence and Analytics) at NWL. Provides evidence of the impact of Jumping Rivers within the utilities sector.