• Map of potential store selling opportunities based upon the store demographics.
  • Takes the demographic traits of top-selling stores and compares them to the demographic traits of stores for which the item is not currently traited.  
  • The map will highlight stores that have “likely high sales opportunity”, potentially high sales opportunity”, and “likely low sales opportunity”.

Likely High: >= 80% probability
Potentially High: >= 50% probability
Likely Low: < 50% probability

  • Expanded view lists store and item descriptions, probability, and the 3 most relevant demographic traits.

The particular percentage buckets are simply arbitrary, chosen by what we thought would be the good points of separation. As to how we arrive at a percentage, we take the demographics of all of the stores an item is currently in along with their sales. We train a neural network (a complex model of nonlinear equations) to fit to the existing stores. Then we take stores the item is not in, feed the demographics through the neural network, and arrive at a percentage estimated likelihood that the store would be high selling based on the trends we discovered in the training data.

Did this answer your question?