Let me start this article with a case study. A challenger bank decided it would use an existing, rich data source to segment its customers into around eight groups.
This had the potential to be a powerful tool. Using the snapshot they took from the data, they could view customers groups based on income, age, likelihood to fall into a collections cycle and a range of other measures. They wanted to use commercial segmentation to understand which customers were most likely to buy their products.
In the first month of implementing this segmentation model, the bank was able to implement the appropriate strategies for each group to develop their business and minimise losses.
But this is where the story of success ends. Subsequent months showed the model continued to segment customers neatly into groups as planned.
The drawback was that these groups had no stability and large numbers of customers changed from one segment to another each month. As a result, it was impossible to use the model to make strategic decisions relevant in the next quarter, let alone the following year.
Test, learn and review
It shows that testing, learning from the outcomes, and building a proof of concept are good ways of doing things. However, you also need to understand your data, how your models are constructed, and how to implement them. The consequence of failure is lost business and a damaged reputation.
This example is taken from some years ago, before Open Banking and other alternative data sources, which offer businesses more avenues to improve their strategies.
Using more than one data source provides more certainty that you are seeing an accurate, reliable picture. But picking the most appropriate sources can be a challenge without wide expertise of how advanced analytics affect data and modelling results and strategies.
There is no better alternative to testing, learning and analysing the results. However, this is a time-consuming process, and there are plenty of opportunities to make decisions that lead you no closer to your goal.
At Experian, one of the ways we are most valuable to our clients is by lending them our experience. Our data scientists have tried and tested many strategies and are well placed to advise you on what has worked in the past – and what hasn’t.
Data on its own can only tell so much of a story. Real success lies in using high-quality analysis to discover the most relevant sources and how they can be harnessed to get relevant insights.