Predictive AI churn modelling delivers £13m in revenue uplift for TradePoint
Challenge
TradePoint raised concerns about its potential level of customer churn. It was unclear which customers had truly lapsed and which were just following their normal customer lifecycle, thus making it difficult to identify customers who should be targeted with a customer win-back campaign.
Solution
It was important to quantify natural variations in the purchasing cycle of different groups of customers. Some buy every three days whereas others may only make a purchase every 20 days. A one-size-fits-all solution would not be appropriate to reveal who had truly lapsed and who would come back if you wait a little bit longer.
Consequently we determined the need for separate prediction models for each customer group to understand where in their expected purchase cycle they were.
Profusion created an AI model to identify natural rates of return for each customer and thus more clearly identify where in their customer lifecycle they are. It gave every customer a label of ‘neutral’, ‘active’, ‘at risk’ or ’churned’. This allowed Profusion to identify 475,000 out of 2 million customers at high risk of churn. These customers were targeted with a win-back programme that included 5 different sets of incentives.
Result
A predictive churn model accuracy level of 91%.
53% of reactivated customers were still active after six months.
Over £4 million of additional revenue was generated from previously inactive customers in the first month.
£13 million of incremental revenue was generated in the first six months.
Sound good?