Dispatch · Call-Fit · Van intelligence · §4–5
The right Van for the call — not the nearest Van, and not the highest-revenue tech.
Every eligible Van receives a dynamic multi-objective Call-Fit Score. It is not a single sales metric and it is not distance-first. Design / stub in this portal.
Call-Fit
Call-Fit = quality + service competence + technical fit + client-tech fit + geolocation + truck inventory + estimated runtime + expected fixed-dollar contribution + expected margin + schedule impact + exploration opportunity.
Geographic / ZIP intelligence
Clusters, ZIP patterns, branch boundaries, traffic, and corridors can consolidate travel. A farther technician can still win when technical resolution, inventory, quality, or contribution materially outweighs the route savings. Fuel is never the final deciding factor.
Dynamic route drift
An 8–12 hour field day will not stay on the morning drawing. Drift is normal evidence. After every meaningful event: actual position/state → rebase → next event. Day-level measures: morning forecast quality, adaptation quality, final-day outcome.
Exploration / curveball
Pure exploitation poisons the model: technicians who historically receive a job type gain evidence; others look weak because they were never tested. Exploration is controlled and competence-gated, not random. Observed capability is tracked separately from uncertainty / lack of opportunity.
Van-first identity
Operational analytics should primarily reference Van identifiers rather than technician names. Identity remains available to authorized roles; optimization and reporting can be pseudonymized where practical.
| Internal role tag | Meaning |
|---|---|
| Service 1 / 2 / 3 | Entry, intermediate, advanced service tiers |
| Warranty | Approved for warranty work |
| Selling Tech | Equipment-sales capable field technician |
| Repair Tech | Complex repair execution |
| Comfort Advisor | Dedicated sales / comfort consultant |
| Manager | Authorized management / escalation |
Tags are fully modifiable by authorized humans. Trades can recommend for or against a change; it cannot supersede it. Every modification lands in Chain C. Maintenance Tech Levels 1–3 appear in §26 alongside the same human-wins rule.
Output vs performance
Output reports what happened. Performance estimates how well it happened after context normalization. Revenue is not skill. Windows: 1 week, 30 / 60 / 90 days, 6 months — so a spike is not the employee narrative.