Over the past several years, the Canopy Labs team has interacted with a number of financial services companies — in banking and in broader services, such as insurance and management consulting. Our experiences have given us a look into just how similar most analytics capabilities are within these companies, and to think of some opportunities to grow the effectiveness of their models and predictions. With that in mind, we’d like to share a few ideas that have yet to be in widespread use in financial services and represent opportunities for a bank or credit union to differentiate themselves. Here are three growing opportunities in analytics for financial services.

Behavioral web data

Even though most companies are collecting data on their website visitor and behavior, we have yet to see a financial services institution that is using this data effectively to segment clients and optimize their outreach. This is a no-brainer and does not even require further analytics; simply sending triggered e-mail marketing or running cold calling campaigns based on the resources a customer requests is a quick win for any data-oriented financial services organization. Indeed, this is often a better approach than investing in more heavy duty modelling capabilities.

Text mining of transaction descriptions

The descriptions that are associated with transactions often contain an immense amount of useful information. Using standard “bag of words” text mining allows you to correlate this data with purchase propensity for specific products without much manual training or data set creation. There are many intuitive examples of this: for instance, individuals who subscribe to certain magazines or purchase at luxury retailers are more likely to have investable assets. Taking this further, a text mining approach would enable you to find less obvious, but equally valuable examples.

Probabilistic graphical models (PGMs) for customer decisions / events

This is less about specific outcomes and more about a new type of approach to analytics. Specifically, consumers make financial decisions based on a broad set of considerations — what assets they own, their future outlooks, and much, much more. This means their propensities to buy are probabilities that must be conditioned on these contexts, and decisions are often made hierarchically (i.e., “I will buy a house” then leads to “I will need a mortgage”, not the other way around). Many traditional analytics approaches do not take this into account; a PGM-based approach could enable this hierarchical decision-making model approach while scaling well for large datasets.

This list is not exhaustive or comprehensive, but we feel they represent very obvious and clear opportunities for analytics excellence within the financial services space. We haven’t seen many companies implement any one of these (let alone all three!). We’d love to hear if you have done this, or if you think we missed other opportunities.