The challenge
A retail client with more than 90,000 SKUs in over 2,000 stores wanted an automated pricing tool that would set the best price point for each SKU to maximise revenue, units sold and margin.
Pricing decisions had been made on intuition. The client wanted to use advanced statistical methods instead.
What we did
- Designed a price optimisation framework: historical data analysis, statistical model building, price testing in one market, feedback, model fine-tuning, then country-wide rollout.
- Built price models at category, sub-category and SKU level.
- Used regression models that accounted for price elasticity, cross-price elasticity, cannibalisation, attachments, threshold price points and price gaps.
The result
The client used the model on a sub-category first and saw a significant lift in sales, units sold and margin.
How we would build it today
- Bayesian hierarchical elasticity models that borrow strength across similar SKUs.
- Optimisation with business guardrails such as price ladders and competitor gaps.
- A pricing web app with an assistant that explains each recommendation to category managers.
This describes our current toolkit for the same problem. It is not part of the original engagement.