Using powerful data science to prevent customer churn
TUI came to us with a churn problem and asked for a model to predict which customers were not likely to purchase another holiday within two years.
The first challenge was to understand whether there is a churn problem or a definition problem – package holidays are expensive so expecting customers to re-book within two years may not be realistic. That meant we needed to re-frame the problem – how can we ensure that we are having relevant conversations with customers when they are ready to make a booking?
To do this, we developed proactive retention – a suite of models that predicts (monthly and automatically) when customers are likely to book, when they are likely to travel, the destinations they are most likely to go to and the type of holiday they are likely to book. This means that every month there is an automated decision engine that tells the marketing team what action they should take with every customer and prospect in the Nordics. It even gives the best alternative destinations, as well as their top pick, to enable multiple options automatically. And when customers are unlikely to book, the client could decide to engage with them to build a relationship rather than always push for a sale.
To date, we've created and automated 43 models. This has resulted in a global rollout by the client and over 1m new customers acquired.