Similarity Scoring

Similarity Scoring | MKT Analytics
A medical information company sought a way to provide to device manufacturers lists of surgeons similar to their existing customers as a tool for enhancing those companies’ sales prospecting efforts. Product purchase data was not available for the surgeons, so a traditional lookalike model based on purchase vs. non purchase was not an option. 

Available data for each surgeon included educational history, current practice location, primary specialty, hospital affiliations, procedures performed and manufacturer payments received. Because much of this data is categorical, common clustering algorithms were also not a feasible solution.

MKT Analytics devised a probability-based similarity scoring solution, where highly common values such as a given primary specialty received lower weighting in the scoring. Less common values, for example, a match on fellowship location, garnered higher weighting. The similarity score derived from the sum of all the weights for the available variables. 

The deliverable is an interface that allows the user upon entering a surgeon’s name to: 

  • View the profile of the entered surgeon
  • See a ranked list of the most similar surgeons, with ability to adjust the length of the list
  • See how much each category of variables (education, practice, procedures, etc.) contributed to each similar surgeon’s score
  • Adjust category weights if there is a reason to feel some are more or less important