AUTOMATED EVENT DETECTION

USE CASE

THE CHALLENGE

A large retail bank wanted to be more relevant for its clients and pro-actively help them solve their life situations by approaching them with right offers.

Standard propensity campaigns have low response rates, which makes them cost ineffective. In addition, customers are nowadays becoming more sensitive to impersonal direct marketing.

OUR SOLUTION

We developed a private library of various event detection algorithms together with a framework to pre-process the transactional data and evaluate multiple settings of each method with respect to predictive power of shift in needs for different banking products or services. When designing the event detection methods we cared about them being theoretically robust and techinally well implemented with speed as the main factor. We also made sure the events have a sound internal logic behind them in order to allow for a strong business story.

RESULTS

We were able to achieve 2-3 times higher response rates compared to standard propensity-based campaigns. Our solution automatically adds, updates or removes events ensuring the events are relevant. With hundreds of relevant events, thousands of customers can be contacted daily, making thevent-based campaigns a significant source of sales