Customers are identified considering health but also factors such as rising living costs, exclusion from digital technologies, or lack of a smart meter, among others.

The network pointed out in its Network Innovation Alliance Summary of 2024 that “current identification standards in the industry are not as advanced as in retail and telecommunications”. Project Spotlight is helping to change that. It identifies PSR, FP, and LNB customers at a household level using new data sources from sectors including telecommunications and finance and the power of AI.

Just as important as identification of vulnerability is ensuring that customers make use of the support that’s available (such as inclusion on the PSR). So the project emphasises the value of effective engagement, using data insights to choose the most effective channels for various needs across vulnerability categories.

CKDelta’s Ayuso backs up what many in the utilities industry suggest when it comes to vulnerability. “With recent increases in gas and fuel prices, more people are pushed over the fuel poverty threshold. There are increasing numbers of storms and power outages, and that affects vulnerable people as well.” For some customers, having the power back on is a matter of life or death.

The △Priority AI application that powers Project Spotlight takes in a wide range of data to identify vulnerability. This includes existing PSR data, energy consumption data from UK Power Networks, medical data, information about household income (which helps to define customers who have slipped into fuel poverty), government data (such as from the census), and commercial data specifically purchased for the project.

All this information had to be cleaned up before it was ingested. “The data is at different geographical granularities,” explains Ayuso. “We have to merge the datasets at a household or postcode level, distinguishing businesses from people’s homes.”

A supervised machine learning model is then used to predict which customers are vulnerable. The training data for the AI was the existing PSR. That data helps the model to learn about households with similar demographics.

Following optimisation of the machine learning model, the most promising parameters were defined for the model to make predictions. The model is continuously changing because new households are added to the PSR all the time. “We refresh the list every day. And then the model predicts if a household is PSR-eligible or not PSR. Meanwhile our fuel poverty model works out the household income versus the expenditure of that household on bills,” says Ayuso.

Privacy is vital. UK Power Networks carried out data protection assessments and put robust measures in place to ensure all personal data is securely managed. CKDelta AI strategist Aarthi Kumar emphasises the importance of discretion: “Because we are handling personally identifiable information, it is important we are doing that correctly for compliance with GDPR, keeping general privacy and ethics to the forefront.”

Before Project Spotlight, it was difficult for UK Power Networks to know which households to target. Marketing campaigns targeting vulnerability might be quite general in character. With △Priority, the network can identify vulnerability at the level of specific households or postcodes and target individual PSR groupings. This specificity has helped UK Power Networks to develop targeted marketing campaigns.

A number of campaigns and contact methods were used, including messaging strategies using SMS and email communications. This data was carefully analysed to see which campaign methods were most successful. “We looked at the sign-up rate. Did they register for the PSR after the first SMS or email, or did it take a second or third one?” explains Ayuso.

Some customers the team identified were less well-versed in certain technology and channels. They might be concerned about scams and handing over information. Others said they were of pensionable age but didn’t feel they needed to be on the PSR, and that others would benefit from having the resource.

“We looked at which campaign performed best,” says Ayuso. “The future will involve using LLMs or new technology to enhance this analysis, modelling how we reach out to customers, and making recommendations about the best way to contact them. We are gathering data about contact methods and campaign success to build a model for reach-out effectiveness.”

Kumar explains that “the data is just part of the puzzle”. “We also need the campaign to let people know the PSR exists. Data can amplify the effort and make it more pointed. Some may be hesitant to register, or there may be a mental block. In the past, you might have been sending out emails as part of a standard campaign which went into the customer’s spam folder or were otherwise unread.”

Get the targeting of communications right, and there is a societal benefit, she adds. Kumar hopes the △Priority AI application will end up being used by many utilities. “Now we have built it, we are seeing that this a product that can be used by all utilities.”

“This project is a crucial step towards building a more inclusive energy system,” confirms Jo Lomax of UK Power Networks, “and we are sharing our knowledge across the utilities industry to help them adopt similar methods.”