The Challenge
A national coffee chain needed to increase prices but wanted to do it intelligently — not blanket increases across the board. Previous price changes had been made without understanding which products or locations were most price-sensitive, leading to unpredictable volume impacts.
The Approach
Using POS transaction data, I estimated price elasticity at two levels:
- Product-level elasticity: Identified which menu items customers were most and least sensitive to
- Site-level elasticity: Measured how price sensitivity varied across locations
This allowed for targeted recommendations: - Increase prices more aggressively on low-elasticity items - Protect high-elasticity traffic drivers - Re-tier sites based on local price sensitivity
Data Challenges
The data wasn’t clean: - Outlier prices from manual overrides - Multiple price points appearing at the same site on the same day - Missing price data for many product-site combinations
These required careful cleaning and validation before modeling.
The Outcome
The model identified which products could absorb price increases with minimal volume loss, and which sites should move to different pricing tiers. This replaced gut-feel pricing with data-driven recommendations.
Key Insight
Not all price increases are equal. A 5% increase on a low-elasticity item might cost you nothing, while the same increase on a traffic driver could hurt footfall across the entire basket.