Roadside Service Locator

When a truck goes down, dispatch searches maps, calls numbers and hopes. Distance is the only thing a map gives you, and distance is a weak predictor of whether a shop answers the phone at 11 p.m., has the part, or charges what it quoted.

That knowledge existed — in the heads of the dispatchers who had called before. It just was not written down anywhere.

Capture the outcome, not just the search. After every call-out, dispatch records what happened: answered or not, fixed or not, price as quoted or not. Two fields at the end of an incident turn a directory into something that learns.

Rank on distance and history, weighted. A shop 30 miles further that has never let anyone down beats a nearer unknown. Both factors are visible in the result so dispatch can override.

Cache aggressively. The set of service points in a region barely changes week to week. Hitting the maps API on every breakdown would be slow and expensive for no benefit.

Dispatch starts from a ranked list instead of a map search. The institutional knowledge that used to leave with an employee now stays in the database.

The outcome form was optional at launch, so for the first weeks it was mostly skipped and the ranking had nothing to learn from. Making it a required step to close an incident — one screen, three taps — would have got the system useful a month earlier.