Predictive Analytics to Identify At-Risk Households

Tackling fuel poverty with Cadent Gas

Issues and Impacts



Fuel poverty, the condition of living on a lower income in a home which cannot be kept warm at a reasonable cost, presents a big issue for local government. Currently, over 2.5 million households have to choose between heating their house and other essential costs, which leads to a multitude of health and societal issues. Identifying households that are in fuel poverty is difficult and expensive. This often means local government and other organisations (utility providers, charities, third sector) spend vital money in both identifying which households are experiencing fuel poverty and inaccurately target services and interventions.


We worked with Cadent Gas on developing a cost-effective and accurate targeting mechanism. In order to do this, we built a machine-learning algorithm that uses publicly available data to identify households at risk of being in fuel poverty.

The Results

The resulting product is a web-based platform that can be used by local authorities to identify households in fuel poverty and target interventions aimed at improving energy efficiency and keeping houses warm. With the resulting software, local authorities can identify, with 75 to 80% accuracy, whether a household is experiencing fuel poverty. Using the data from the platform can be a key asset for local authorities in their policy-making process, as it can help them understand the distribution and concentration of households with high health and wellbeing risks.