A pricing engine that learns the showroom.
A multi-brand automotive group was pricing on gut feel and monthly committee reviews. We built a dynamic pricing copilot that lifted conversion 9 points while protecting margin through cycle.
The situation on arrival.
Pricing decisions sat with individual showroom managers, informed by a monthly committee and a spreadsheet. Discounting was inconsistent across brands and regions.
The group had no visibility into which discounts actually closed deals versus which were simply given away.
How we structured the work.
We modelled three years of transaction data against inventory age, regional demand, and competitor listings to identify where discount was buying conversion and where it was pure margin leakage.
Critically, we built the system as a recommendation engine, not an autopilot — showroom managers retained final authority, which was essential to adoption.
The override button was the most important feature we built. It was how the model learned the showroom.
Systems now in production.
- A demand forecasting model at SKU-region level.
- A recommendation engine surfacing a suggested price band per unit with reasoning.
- A manager override interface capturing why a recommendation was rejected — which became the highest-value training signal.
- A group-level margin dashboard for the commercial committee.
What changed for the client.
Conversion improved 9 points across the pilot brands. Margin held 340bps above the prior-year comparable despite a softer market.
Override data revealed two systematic model blind spots that were corrected in the second iteration.
Considering a similar programme?
We take on a limited number of engagements each quarter. If the shape of this work is familiar, we should talk.
Start a conversation →