Water & Wastewater Treatment Magazine
Issue link: https://fhpublishing.uberflip.com/i/1169034
14 | JUNE 2017 | WWT | www.wwtonline.co.uk In Focus: Customer bills 14 | OCTOBER 2019 | WWT | www.wwtonline.co.uk " Northumbrian Water Group (NWG) is dedicated to eradicating water poverty across our operating areas by 2030. This goal, as set out in our 2020-2025 business plan, supports our commitment to providing affordable and inclusive services for all by delivering an unrivalled customer experience. "Using data science for continual monitoring means we can successfully accomplish our goal. Historically we have estimated volumes of customers in water poverty, but without a reliable picture at individual account level. We recognise that realising this ambition demands ac- cess to truly actionable insight. "This led us to our Water Poverty Eradication Modelling (WPEM), which was established in 2018 with the over- arching aim of providing intelligence via innovative use of leading-edge machine learning techniques and toolsets to chan- nel our ocean of big data. "Modelling allows us to detect custom- ers currently in trouble and – significant- ly – triangulates contributing factors to anticipate when customers may become water poor. Consequently, we are able to take pre-emptive measures to support customers before they suffer the stress of falling behind on their water bill. "We set out with four key objectives: • Determine our current water poverty landscape – to eradicate this problem we must first clearly understand the scale of the problem • Identify patterns and trends – interrogat- ing our datasets to identify those factors which dictate/predict water poverty • Provide sustainable analytical mod- els 'owned' by NWG – previous work indicates 18.4 per cent of our house- holds (i.e. 370,000) spend over 3 per cent of disposable income on water and sewerage services but reproducing this information regularly is currently a costly exercise • Deliver predictive analytics to drive business decisions and increase customer engagement. This includes altering tariffs and targeting marketing campaigns based on WPEM findings – offering proactive rather than reactive support to customers who may be strug- gling to pay "Typically, the UK water industry has a reliance on credit reference agencies in terms of water poverty data and accessing insight required to make true impacts. This is inefficient and expensive. "To mitigate this, NWG has developed a unique neural network (WPEM). The neural network itself is not a single algo- rithm but rather a framework for many different machine-learning algorithms to work together and process complex data inputs. Such systems 'learn' to perform tasks by considering examples, generally without being programmed with any task- specific rules. "Whilst the modelling was led internally by the intelligence & analytics and litigation teams, we benefitted from NWG's ongoing commitment to partner with higher education. An undergraduate 'intern' from Newcastle University joined the intelligence & analytics team, supple- menting in-house capability and taking a central role in successful project design and delivery. "This arrangement is a great demon- stration of these mutually beneficial en- gagements fostered by NWG. The WPEM initiative and NWG teams have benefitted massively from the leading-edge machine- learning techniques and energy brought by our intern, while they've gained valu- able opportunities to apply data science practically in an industry setting. "By applying our own, sustainable machine-learning models against our customer big data, we are gaining a 360-degree view of our current water poverty situation, alongside the ability to forecast and take preventative rather than remedial action. "Tackling the issue of bad debt now, as Ofwat advise, will reduce our costs and make bills more affordable for all custom- ers. Bad debt currently adds an estimated £21 p.a. to each customer's bill. "Our modelling enables intelligence- led decisions, stretching our ambition to lead the sector in addressing water pov- erty. In addition, by reducing reliance on credit agencies' data services, NWG will realise sustainable savings every year. "The model is applied to over two million existing NWG accounts. Over 30 different variables are used to predict customer groups and individuals at risk and simulate various billing scenarios to calculate the impact of different tariff changes with regards to water poverty. "Testing/validation of the model against known income/expenditure data has provided high confidence levels – the model successfully identified 83.3 per cent of customers in water poverty. That said, WPEM is designed to provide continuously improving accuracy, its algorithms configured to learn over time and accommodate new data sources as these arise to refine trending and the ap- plication of optimal affordability tariffs. "WPEM also pinpoints worst-affected areas geographically, allowing better targeted marketing campaigns, increasing customer engagement regarding afford- ability/payment methods and driving efficiencies. "In meeting these objectives, WPEM provides an innovative solution, deliver- ing insight to our teams in support of our main aim – looking a£er our customers who may be most in need of our help – before any financial or emotional burden becomes too serious." "We can take pre-emptive measures to support customers before they suffer the stress of falling behind on their bill" • WATER COMPANY VIEW GRAHAM COULSON, PERFORMANCE & INSIGHT ANALYST BI LEAD IN INTELLIGENCE & ANALYTICS TEAM, NORTHUMBRIAN WATER Water Poverty Eradication Modelling was named the Data Project of the Year at the 2019 Water Industry Awards