Propensity to Fly
20%
increase in predictions
The Ask
The luxury airline asked our experts to predict which customers are most likely to fly again in the next year and then determine which of them can be persuaded with incentives.
Our Solution
We created an ensemble neural network model that incorporated previous flying behaviour, activity on the client’s website, nationality, miles flown and several other key features. This delivered metrics of 93% prediction and 97% recall when evaluated against the following year’s ticket purchases.
Impact
Predictions improved by 20%, arming strategy teams with characteristics of customers with the highest propensity to fly (no incentives needed), a medium-high propensity to fly (incentives will deliver significant improvements), or low propensity to fly (only the best incentives will work).