The first is to develop a corporate memory and understanding of who each customer is, based on historical information. Databases of historical service and account data organized by customer identity provide a basis for analyzing customer value (profitability), channel preferences product affinities, geographic and demographic data all of which are useful for segmenting customers and developing customer management strategy. Using historical customer information capably is table stakes in today’s relationship managed banks.
The second way to leverage customer behaviour data has a more responsive orientation. It involves parsing through transactions and service contact data in near-real time to know what customers are doing. Analyses can be automated to identify exceptions that prompt an intervention by sales and service staff. Ideally, identification of exceptional customer activity enables timely responsive customer interaction. There is certainly value in responsive behavioural analytics provided the process can work quickly and accurately enough to provide leads to sales and service staff that are credible, timely and relevant. False leads delivered to the front lines can foment resistance in the field quickly stalling responsive programs with inadequate business rules.
The third way to leverage customer data is more proactive. Models are developed to predict what customers are likely to do and allocate resources using this knowledge. Most banks already use predictive credit scores to adjudicate loans and to evaluate likelihood of default for credit loss provisioning. Similar predictive scores can be developed to identify customers at risk and those most likely to accept an offer. Predicting behaviour enables proactive customer management programs to be developed for acquisition, cross-selling and retention. If you can predict what customers are going to do, you can improve sales and service performance. The keys to program effectiveness are precision in scoring coupled with effective customer engagement by sales and service staff.
In all three cases what matters most of all is relevance. There is no point in identifying or predicting something that does not matter with a high degree of precision. Or worse, identifying / predicting the wrong thing.
Unfortunately this is exactly what happens a lot of the time in bank customer intelligence analytics. Models are created that identify “significant deposits” or predict “probability of account closure”, for example. Neither of these things is the right target behaviour of interest. Significant deposits may or may not reflect a significant source of new money to the bank. Similarly account closure may not bear any relation to the withdrawal of funds from an account.
We need to remember that the retail banking business is about flow of funds, and managing their cash flow is what customers do in real life. We need to understand the types of cash flow behaviours from a customer perspective rather than a transactional or data driven perspective. Focusing on what customers do with their money and the patterns of these behaviours offers a sharper and more effective basis for understanding historical behaviour, predicting future behaviour and reacting to current activity. The essential thing is to know what customer behaviour really is, then measure it, then model it.
- David McNab
