← All work Next caseAn AI bill cut by $4,050 a month →
A price for every item, set by math instead of a meeting
9.67Brows of sales, priced
- Scale
- Every item, every location
- Methods
- Price elasticity per item, competitor matching, a constraint solver
The problem
Prices were set by gut feel. A/B tests took a quarter to read. One elasticity number was used for every item, so every item got the same advice. Nobody believed it, so nobody used it.
What I did
I measured price sensitivity for each item from real sales and labeled every estimate by how much to trust it. Competitor prices were matched to our items even when the names didn't match. Then a solver picks a price for every item at once, under the real rules: margin floors, contract caps, competitor ceilings, and how items steal sales from each other.
FIG.01 ยท pick an item, drag the price: demand, profit, and the rulestry it · sample data
What changed
- Every recommendation is a dollar amount, a projected lift, and the one rule holding it back.
- It runs inside the warehouse, on demand.
- Six other tools read from it.
What this looks like for you
"The model says raise it" doesn't survive a pricing meeting. "Raise it to $9.49; the contract cap is the limit" does.
Let's talk data →