CASCase study
Making a pricing decision defensible in a multi-branch trader
A multi-branch building materials trader, where every salesperson set prices from experience. Years of sales data existed, and none of it helped anyone decide.
The constraint No external market price available: everything had to come from internal data alone. And a salesperson on the floor will never open a tool that asks them to understand a statistical model before answering a customer.
How it ran
- Diagnosis before remedy: relative margins by branch, family and item, dispersion, share of revenue below the median.
- A reference price derived from the internal market, with a high tariff and a floor rather than one imposed price.
- The decision runs through an explicit tree, readable by a salesperson, not a black box.
- The calculation engine is isolated from client context: the mechanics stay frozen, thresholds and taxonomies adapt to the company.
- A what-if simulator to arbitrate before applying, with the expected gain waterfall and an export.
- Whatever the tool cannot compute yet is labelled “pending” instead of producing a plausible price.
Outcome
- Live at the client, on their real data.
- A pure calculation engine, isolated and test-covered, separated from everything company-specific.
- Self-hosted database, private access.
- No invented price: the strictest design constraint in the product.
What I take from itA pricing tool almost never runs into the maths, it runs into trust. As long as the salesperson cannot explain where the number comes from to their customer, they will not use it, however good the model behind it.