MPM’s advanced modelling and analytics – reviving a Japanese credit card issuer’s direct insurance cross-sell strategy

1. Overview


Working with our partner, a Japanese credit card company, and a domestic insurer, the MPM Japan business applied a three-phased data modelling strategy to credit card holders who had been ‘over-marketed’ to by other marketing services providers for many years.

With access to more than 1.7 million members (holders of selected credit card products who met the very basic criteria to be included in a telemarketing program), MPM achieved significant increases in net revenue for our client during a six-month period, compared to campaigns being executed by other marketing service providers.

Significant improvements were tracked across a range of metrics, including:

  • Telephone contact rates
  • Sales per contact rates, and;
  • Sales per hour rates.

In addition to the positive revenue impact for our client, other achievements included reductions in cost per sale and an overall improvement in customer experience which resulted in significantly lower complaints and negative feedback received throughout the marketing period.


2. MPM Japan's brief


It was evident that a number of our clients prior partner insurers and intermediaries had ‘over-marketed’ to their client base.

These broadscale campaigns may have resulted in a series of temporary revenue increases; however, over several years, an overall reduction in sales and revenue from the insurance sales channel was evident. Customer complaints were increasing and profitability targets were not being achieved. The telephone sales representatives also did not see the credit card company’s sales programs as a progressive program to work on. As a result, morale was decreasing.

With more recent renewed growth in the number of cardholders and acquisition of a number of new business partnerships, MPM saw an opportunity to work with our client in introducing a revised program of activity. We incorporated a new level of sophistication to the data selection process, underpinned by MPM’s Campaign Factory functionality. Being a credit card company, there was access to rich customer demographic and financial information.

MPM was tasked with implementing a phased data modelling program that would result in improvements in the following categories:

  • increased sales per contact rate
  • improved first customer ‘paid rate’
  • increased annualised premium revenue per sale
  • reduced complaint volumes
  • a program of ‘choice’ for telesales representatives to work on
  • better planning and forecasting for future insurance cross-sell initiatives.

 
3. MPM Japan's methodology


Six credit card products from their suite of card products were selected, providing a significant customer volume to run the program.

Phase 1

The first phase of activity involved MPM engaging in an exploratory data analysis and identifying a moderate volume of card customers to run two initial campaigns. Only a basic level of data segmentation and targeting was used.

The aim was to attain a series of learnings and use these as ‘control’ cells to benchmark more sophisticated ‘modelled’ campaigns as we progressed with further phases of the program.  

Phase 2 A third campaign was then implemented, with a slightly more involved level of data segmentation based on some of the key attributes of responders from our initial campaigns.
Phase 3

MPM’s sophisticated modelling methodology was then applied to the third phase of activity, which used historic results from previous campaigns to develop predictive models to help identify the most responsive customers within the available customer universe. In summary, this involved four parts:

  1. Data functions

    • data aggregation from the first two ‘control’ campaigns that ran in Phase 1 into the bulk credit card company customer data
    • identification of data items to be used for modelling
    • creation of derived fields
    • data quality analysis
    • high level data profiling.

  2. Modelling functions (in step-by-step order)

    • model selections were made, identifying the most appropriate models to be used in the analysis
    • model calibration was applied, running historical data through the model to identify predictive fields
    • model verification was then undertaken, which involved testing of the model by running historical data through with past results, allowing MPM to measure scoring effectiveness
    • scoring was then undertaken for the fresh data set and then used during the campaign selection process.

  3. Profitability forecasting

    This involved using the model verification data to forecast sales, contractibility and logged hour estimates, which helped determine profitability cut-off scores used in campaign selections

  4. Campaign planning

Campaign plans across the coming months were then generated from the scoring and profitability forecasting.


4. Result summary


There were notable increases in several metrics when the earlier ‘non-modelled’ campaigns were compared with those developed through the data modelling initiative. Against the control cells:

Basic Segmentation

Contact rates increased by 18%
Sales per contact rates increased by 300%
Sales per hour increased by 211%

Data Modelling

Contact rates increased by 59%
Sales per contact rates increased by 395%
Sales per hour increased by 317%


5. Key learnings


  • Data modelled campaigns are more likely to produce more favourable response rates.
  • Data modelled customers within the 45–49 age band are most responsive.  
  • Applying data modelling has allowed MPM to open up profitable segments in the younger age bands which were previously neglected by other organisations.
  • It is likely that the same data modelling process applied against other products will produce similar results.
  • Campaign planning (volumes, budgeting and revenue expectations) is now easier, as the modelling process can be applied to assist with this function.