Field Service Route & Schedule Planner

What you get

Give it your jobs, your crews and your customer list. It returns one report:

• A schedule per crew - which jobs go to whom, in what order, at what time, built on real drive times (OSRM / OpenStreetMap, named in the report, no key needed), respecting each crew's working hours, each job's time window and the skills each crew holds.
• Every job accounted for - anything that could not be placed is listed with the reason (no crew with that skill, no feasible slot, address not locatable).
• The empty slots - each gap of at least your minimum length, with the crew, the time and the area the van will be in.
• Who to fill them with - the customers near that gap who are due for service (overdue first, then due soonest; open quote and plan customers preferred), ranked, each with the reason it is suggested and the extra drive time it costs. Anyone on your do-not-contact list is excluded automatically.
• Ready-to-send drafts - a short message per suggestion. Nothing is ever sent: the agent has no send code path at all; you review and send.
• A drive-time comparison against the order you supplied, so you can see what the optimisation was worth.

Every suggestion is simulated before it is shown: the crew's whole day is rebuilt with that customer inserted into the gap, and it appears only if every later job still lands inside its window.

What you provide

Three sheets (CSV or JSON), the same columns your scheduling tool exports:

Addresses are geocoded through OpenStreetMap - city + street is enough, no coordinates needed. Missing last-service dates and duplicate rows are handled and reported, never silently guessed. A planning period (e.g. a 5-day week with light days) is also supported.

What you get back

A Markdown report plus the same content as JSON (schedule, gaps, suggestions, drive_time_comparison, acceptance, assumptions).

Limits, stated up front

• The optimiser is a strong heuristic (cheapest-insertion + relocation + 2-opt), not a proven optimum: when it cannot beat your own order it says so instead of inventing a saving.
• Free public sources only: if the routing service is unreachable the run still completes and every drive time is labelled an estimate with the reason.
• It does not send messages, book anything, or write to your scheduling system.
• Typical run: ~30-40 s for a 10-job / 3-crew day (cold), under a second cached.

Verified run (this build was tested before listing)

10 jobs / 3 crews / 18 customers around San Francisco Bay, live geocoding + routing: total drive 145 min vs 162 min for the same jobs in the order given (17 min saved), 4 open slots found (692 min sellable), 11 backfill suggestions, 0 jobs unassigned. 25/25 offline self-test checks pass.