Predicting electricity churn for Scottish and Southern Energy

SSE is one of the largest energy companies in the UK following the merger of Scottish Hydro Electricity and Southern Electric in 1998. They serve over 6 million customers supplying gas, electricity, electrical and utility contracting and telecoms.

The Challenge

Following the deregulation of the electricity and gas markets competition from energy suppliers led to levels of customer churn running at between 10-15%. SSE wanted to take control of churn by forecasting which particular customers would defect to a competitor within a pre-defined time window. They then wanted to test out a number of different retention strategies. It was decided that the work would be carried out on electricity customers initially.

SSE needed to be able to test different time windows because although forecasting further ahead diminishes the power of the model, they needed to be able to take action to prevent churn before it happened.

They had done some initial profiling work themselves which had identified general demographic characteristics of churning and loyal customers but were not confident about using the profiles alone to test different retention strategies. They also wanted to test whether demographic data would add any uplift to the transactional models.

The Solution

CACI started off with an extensive audit of the historical data available on electricity customers. A specific type of neural model was used, which had all the power of a traditional algorithm approach, whilst retaining the transparency of a traditional model.

The customer database was then scored with the probability of churn as well as customer value and the performance of the model assessed against unseen historical data, which was then expressed as a gains curve. The model could identify 50% of churners from 30% of the file. Various different loyalty strategies, which they had at their disposal such as the loyalty points scheme, were then tested against the customer base.

The Results and Benefits

The test showed that substantial profit would result from forecasting customer churn and using the appropriate retention strategy. It was also possible by using the propensity to churn, coupled with customer value, to introduce different offers to different sections of the customer base so they could offer more expensive loyalty offers to their higher value customers. It was found that the addition of demographic data not only uplifted the performance of the model but also helped when devising the offers.

The study also allowed them to apply the profiling work carried out on the churners in order to avoid this type of person in their acquisition ampaigns. The database has been rescored with a new time window and testing is being carried out on the effectiveness of a number of different retention strategies against a control group where no attempt at retention has been made. Once this has been completed the supplier will procure a churn modelling system to enable large-scale retention campaigns to be run for both electricity and gas customers.

To find out more about our work with Scottish and Southern Energy either call 020 7602 6000 or email info@caci.co.uk.