The situation
A regional freight operator running hundreds of shipments a week had grown faster than its tooling. Pricing lived in twelve interlinked spreadsheets maintained by three people; dispatch ran in a legacy tool nobody dared upgrade; invoicing was a monthly reconstruction project. Quoting a non-standard route meant finding the one person who knew the right spreadsheet — and took four hours on a good day.
The problem underneath the problem
Slow quoting wasn't a software problem at first glance — it was a knowledge problem. Pricing rules lived in people's heads and in formula cells nobody could audit. The diagnosis phase surfaced the real cost: lost bids (customers went with whoever answered first), pricing errors on complex routes, and complete dependence on three irreplaceable employees.
What we built
A quoting engine first — deliberately narrow. The pricing rules were extracted from the spreadsheets, encoded, reviewed by the pricing team rule by rule, and shipped to a pilot group in week six. Dispatch and invoicing followed as separate milestones, each replacing its predecessor only after running alongside it.
The hard part
Not the code — the trust. The pricing team had every reason to distrust a system that encoded their expertise. Two things made it work: they reviewed and approved every encoded rule (finding several long-standing spreadsheet errors in the process), and the system showed its work — every quote traceable to the rules that produced it. Within a quarter, the same team was requesting new rules instead of maintaining spreadsheets.
Results that held
Median quote time fell from four hours to six minutes — and stayed there as volume grew. Quoting stopped being a bottleneck for sales; pricing errors on complex routes dropped to near zero; and the operator stopped depending on three specific people to price its business. The engagement continues as a dedicated team, now extending the platform rather than firefighting it.