analytics

Today I see many utilities tackling fraud in a very siloed way - running specific teams that focus on specific processes, backed up with one-off projects to boost revenues short term, writes Iain Stewart, a consultant at Teradata.

This typically focuses on areas such as billing and revenue assurance, identification of unbilled revenues by comparing metered consumption versus energy billed out and where possible comparison of power, gas or water metered values throughout the network versus the sum of metered consumption downstream at the customer level.

According to a Northeast Group[1] study released in May 2017, non-technical losses (AKA fraud) cost utilities $96bn a year globally. A quick internet search will bring to your attention other international and regional studies, all of them quoting hefty figures. Regardless of which studies you feel are accurate, the problem – or as I prefer to think of it the opportunity here for utilities to address, is huge.

This silo, point solution approach whilst delivering some benefits can only do so much however. It is massively hamstrung from a number of angles, including:

  • Lack of any significant ability to understand, classify and importantly prove non-technical losses in a way that follow up actions can be justified, and indeed show any given case is valid.
  • Financial benefits of acting on certain categories of fraud, or indeed individual cases are often not well understood.
  • Non-technical losses are managed in isolation from technical losses. As a simple example - if a customer premise includes complex metering arrangements of which a portion are missing from a utilities’ system and/or just not billed accurately, the customer could well understand that they are being underbilled and have just chosen not to confess. Are those losses technical or non-technical? Utilities need to think about losses as holistically as possible – this is all revenue!
  • Focus is tactical in nature - on short term fixes for short term gains, instead of longer term solutions focused on permanent prevention.

These limitations are driven by a lack of focus on how to use data – both from the perspective of what can be done with it and how to integrate it at scale from multiple sources in a way that everyone can tackle losses holistically more effectively.

The good news is there are proven examples internationally of utilities that have overcome these barriers, and of the benefits on offer. As far back as 2013, speaking at our customer led Teradata Partners conference - a Brazilian water utility talked about its integrated approach to losses that considered non-technical losses alongside other factors including water pressure and leakage to understand the gap between total losses it was incurring across the board, and what it termed “unavoidable losses”.

The gap was then fed into more targeted programmes looking at losses technical and non-technical, with the latter using various forms of detailed segmentation highlighting illegal connections and meter bypass, illegal use of hydrants, and tampering with meters themselves and/or the data readings taken.

More strategically going forwards this work also fed the company’s metering replacement programme, focusing the rollout of smarter meters and telemetry on network zones linked with higher losses to stamp out these issues more permanently.

All in this delivered a 7% increase in revenues in the first year alone - not bad I am sure you will agree. The kicker here however, was that they could re-use the integrated data view again in other innovative ways.

They had integrated masses of data from sources such as the CRM, billing and finance estate, GIS, metering systems and more – that could be used to underpin other initiatives at minimal cost, such as improving customer water efficiency to meet new regulatory goals for example.

Other utilities have adopted a similar holistic approach to losses, and indeed data integration. One of our US customers in the space of just two months uncovered $2m of revenues not billed from the previous month, $2.5m of revenue stuck in exception BAU processes with no path to resolution, and $300k of revenue based on metered consumptions that had never been billed.

Another focused-on debt recovery, and what to prioritize – for example, if a customer was committing fraud willingly or unwillingly based on its financial position the utility wanted to understand this, and to be able to potentially de-prioritize those cases given that cash collection could be costly, or indeed there might be no cash available! This had the associated benefit of enabling it to better manage bad debt reserves for example, all based on the power of integrated data.

Building on this concept of using data to fight fraud – there are multiple pre-built segmentation tools with baseline analytic capabilities that segment fraud cases, and importantly identify cases up front on the market.

These tools can be very powerful - but note they are only as good as the data you integrate and feed them with. In the case of one major US utility, we enable such a tool with an integrated data view similar to that quoted in the Brazilian case, which is used to produce live dashboards that can be updated in near real-time if required, providing a focus and extra intelligence for fraud detection teams.

This end to end solution focuses on theft of energy in the field via meter tamper and bypass for example, not only identifying individual customers as being potential fraud cases - but also showing time period during which suspected fraud is taking place as extra evidence to pursue cases.

Again, the kicker here is that the integrated view of data was not just used for fraud detection – it was also used to underpin energy efficiency drives, demand response programmes and the creation of advanced time of use tariffs for example, and to underpin rollout and monetization of smart metering, and more – all of which indirectly of course contribute to lower losses period. A holistic approach to energy provision more widely leads to a more holistic approach to tackling losses.

Refocusing specifically on fraud, it is important utilities look to other industries for new ideas on how to tackle this problem, and indeed for ideas around using data. Modern technology and analytics make data much more accessible to people as information in ways that are both easily understood, and indeed even quite visually appealing! Fraud Invaders[i] is a good non-utilities example of how pretty fraud analytics can look, and the types of analytics that can be considered.

Looking forwards, one thing we can be sure about is that those committing fraud don’t stand still! Artificial Intelligence has the potential to really step up here, giving us ever more intelligent insight on what those fraudsters will look like, and indeed look to do in the future. Other sectors are in my view slightly ahead in this area – look at our work in financial services in Denmark[2] for example.

To summarise - the best way to resolve non-technical losses is to take a strategic approach that also links to your work on technical losses, all of which needs to be underpinned by a solid data strategy in terms of those areas you want to focus on, and indeed how to integrate your data. Integrated data can be used both directly by the utility to run its own analysis, and indirectly to feed third party tools – and as importantly, remember the same integrated data can be monetized in lots of other ways in areas totally separate from losses!

There are examples in utilities - proven over many years that you can learn from, and indeed looking forwards ourselves and others are already looking at the power of integrated data for artificial intelligence as the next frontier in tackling the fraudsters! They never stand still – so why are you in your attempts to beat them?

[1] http://www.northeast-group.com/

 [i] https://www.teradata.co.uk/Resources/Videos/Fraud-Invaders

[i] https://www.teradata.co.uk/Resources/Case-Studies/Danske-Bank-Fight-Fraud-With-Deep-Learning-and-AI

This article first appeared in ESI-Africa Edition 3, 2018. You can read the full digital magazine here or subscribe here to receive a print copy.