Smart Ice Cream: How Data-Driven Assortment Boosted Profits in Retail
- We faced the challenge of optimizing not just pricing and portfolio, but also which products to place in each store with limited space.
- In our case, selling ice cream through wholesalers, we lacked direct data—so we built a store-level dataset from thousands of sales rep emails.
- We segmented stores by income levels, but realized assumptions weren’t enough—we needed analysis to find the most profitable product mix.
- Our algorithm evaluated real sales vs. profit margins, revealing unexpected winning combinations—even in wealthy areas.
- To scale the model, we kept execution simple: recommendations appeared directly on sales reps’ iPads, removing complexity.
- We balanced analytics with field experience by enforcing two products and letting reps choose the third—leading to high adoption and impact.