Rabobank International Direct Banking

Old data model to new data model
and advanced analytics

Rabobank IDB

Rabobank International Direct Banking (IDB) asked for interim input to (a) guide the transition from an old data model to a new data model, and (b) take steps into the field of advanced analytics within Rabobank International Direct Banking.

What did we do?

During our temporary input at Rabobank International Direct Banking (RaboDirect), Cmotions guided the transition from an old data model to a new data model from a customer intelligence perspective. The challenges that arise from a change in customer definition are profound. For example, translating the reporting and analysis into the new data model. Passing this knowledge onto various different professionals to work with the new data structure was also included as part of the task.

In addition, RaboDirect and Cmotions also worked together during the input period to take steps into predictive modelling. A number of preparatory analyses showed that an international Direct Bank had a relatively high level of Churn (losing customers). Based on customer behaviour, Cmotions made a model for this, making it possible to predict on an individual customer level how likely the customer is to withdraw his or her money. The model produces hugely interesting results, with a 30% lift in the first decile.

This graphic shows how effective the model is. If a marketeer selects 10% of customers at random, logic dictates that they would also select 10% of their churning customers (red line). If the marketeer makes a selection based on predictions from the model, a better picture emerges (blue line).  Customers are categorised into 10 equally-sized groups based on their churn probability. If you are a very high churn risk, you will fall into group 1, and if you are a very low churn risk, you will fall into group 10.

If the marketeer then only selects customers in the 1st group, they can select churning customers more precisely: the first 10% of customers that they select (decile 1) ultimately turns out to contain 30% of the actual churning customers. The marketeer is able to achieve the same result whilst making fewer offers.

Let’s imagine we’re talking about a total of 100,000 customers. The marketeer is now able to cause the same effect with just 10,000 offers as they could manage before when making the same offers to 30,000 customers. As well as no longer needing to make so many offers, you also no longer need to bother non-churning customers with irrelevant advertising.


What is the result?

The transition to the new data model has been completed. A new customer definition has been implemented, and there is an understanding of the consequences that the new definition has entailed. Furthermore, the complex reporting structures have all been reworked so that they can be used with the new data model.

The churn model is currently being utilised to run targeted retention campaigns to reduce churn. This means they can use their marketing budget effectively and they no longer have to contact customers unnecessarily with irrelevant messages. There has been positive initial feedback from the country where the model has been used.

Subscribe to our newsletter

Never miss anything in the field of advanced analytics, data science and its application within organizations!