≡ Menu

Optimizing Customer Lifecycle Management

Share

David Griffith from CUNA Mutual Group talked about predictive analytics in a B2B environment. CUNA targets credit unions and cooperatives and their members with software and insurance products. CUNA Mutual needed to acquire new credit union accounts for direct insurance products – credit unions who sign up can offer a full range of insurance products in return for access to the credit union customers. The program was mature and already had 50% market share and incremental sales were a challenge. Plan was to target the 50% of credit unions analytically and focus on those credit unions likely to sign up and those whose members were likely to participate.

The approach was to develop look-alike targets using analytics by analyzing product performance and sales rates/pipeline. Used Angoss’ KnowledeSeeker tool to develop decision trees – the power of the tool for exploratory work and the ease of communication of decision trees were key drivers.

First step was to find credit unions who performed well – whose members signed up. 80 variables that described the credit union (asset size, branches, state, type….) and 20 variables that described the membership of the credit union (ages, income etc). Focused on credit unions that had become members of the program recently to see which ones would be good. The tree quickly identified those with higher rates of married members as doing better than average, for instance. Credit unions with average numbers of married members and high rates of auto loans coming from auto dealers were very below average. And so on – what kinds of credit unions will be more successful. Second step was to find targets based on sales activity. First thing was found that credit unions with multiple products like 401K were more likely to join this program.

To make this work they had to focus it on the actual decision, which was going to be executed by the sales team. Developed a little 3×2 grid mapping likely/unlikely to buy and above average/average/below average memberships. Did the scoring and then mapped scores on to the grid and loaded the grid into the CRM system – essentially the action being recommended – so that the sales people could use it. Models worked. There was a 20% difference in actual performance between above and below average performance accounts and almost twice the response rate in more likely to buy v less likely to buy.

Conclusions:

  • You can use predictive models even in B2B.
  • Make the information easy to understand for sales, front-line
  • Translate into actionable information
  • Engage users in the process for more adoption
  • Position analytics as enhancement

Biggest challenge was overcoming the cultural barriers to a change in the sales targeting approach.

Share

Comments on this entry are closed.