Predictive target audience modeling

 

Challenge:

A top 5 US insurance company was looking to create predictive audience segments for future targeting of home, automobile and life insurance policies.

Solution:

Creation of a safe harbour data environment for advanced predictive modeling using multiple data sets.

  • We created a 3rd party hosted safe harbour environment that can both digest and analyze multiple data parameters

  • We aggregated multiple data sets into this environment, including purchase, search, behavioral, census, digital cookie based, location, and survey based data

  • We anonymized all data from any Personally Identified Information and modelled relevant parts to adhere to all privacy regulations

  • We modelled behaviors against relevant objective functions to identify behavioral predictors in future behaviors

  • We created segmented audience groups based on likelihood of exhibiting future desired behaviors based on given past behaviors and data.


Results:

  • Predictive audience groups had conversions rates that were 3.6X standard audiences without compromising scale.

  • ROI on top 10% of audience group was 8X that of standard audiences

  • ROI on top 25% of audience group was 5.75X that of standard audiences

 

3.6X

increase in conversion rate

 
 

5.75X

top 25% ROI increase

8X

top 10% ROI increase