How service dispatch works

Dispatch is where a service business either makes or loses its margin. Utilisation — the share of a paid day spent on billable work — is set almost entirely by dispatch quality, and first-time fix rates depend on sending someone with the right skills and the right parts.

This guide covers how the matching decision works, the sequence for moving off a whiteboard, and what to measure. For the definition, see what is field service dispatch.

1. The matching decision

Every assignment resolves five constraints that routinely conflict:

Skill — does this person hold the competence or certification the job requires?
Capacity — what is already on them today, and does this fit?
Location — how far is the job from where they will be when it starts, not from where they are now?
Commitment — what was promised, and when does it breach?
Parts — is what the job needs on the van, or does it require a depot stop?

The nearest person often lacks the skill; the most qualified is often committed. Dispatch is a sequence of trade-offs under time pressure, which is why it benefits from being seen rather than listed.

2. Get the inputs right before automating anything

Automated assignment is only as good as three data sets, and all three are usually weaker than assumed:

Job durations. Almost universally optimistic. Measure actuals for a month before letting durations drive a schedule — a day optimised against wrong durations collapses by mid-morning.
Skills matrix. Goes stale silently as people train and move.
Van stock. Rarely accurate to the item, which is what turns a first-time fix into a second visit.

Fixing these is unglamorous and is the actual project. Software cannot compensate for a skills matrix that is a year old.

3. Choose how much to automate

Three postures, and most complex service work should stop at the second:

Manual. A dispatcher assigns using local knowledge. Effective at small scale; the first thing to break as volume grows.
Assisted. The system proposes and fills gaps; a human accepts or overrides. The right default for varied work.
Automated. Rules assign without intervention. Appropriate for high-volume standardised jobs, and a poor fit where site knowledge matters.

The reason to stop at assisted is not caution. Dispatchers hold context that is nowhere in the data — which customer tolerates a delay, which technician suits a difficult site, that the access road is shut this week. Automating that away means being right about things the system cannot see.

4. Plan for the day falling apart

The plan degrades from the moment it is made: jobs overrun, parts are wrong, emergencies arrive, people call in sick. Dispatch quality is mostly re-planning quality.

What that requires operationally: overrun has to be visible while there is still time to act, not at the point of breach; reassignment has to be quick; and technicians need a supported way to decline work that is genuinely impossible, or the information arrives as a missed commitment instead.

5. Sequence for rolling it out

Digitise the work list. Every job in one place with a status. Nothing else works until this does.
Capture actuals. Real start, travel and completion times from the field — which needs a mobile app that works without connectivity, because service sites are basements and plant rooms.
Correct durations from those actuals.
Introduce assisted dispatch for unassigned work only, leaving human decisions untouched.
Add skills and capacity routing once the underlying data is trustworthy.
Automate narrow, standardised job types last, if at all.

What to measure

Four numbers tell you whether dispatch is improving: utilisation, first-time fix rate, commitment attainment, and travel time as a share of the working day. Watch the interaction — utilisation can be pushed up by packing days tighter, which shows up as falling commitment attainment two weeks later. Improvements that move one number and quietly damage another are the norm, not the exception.

How this works in xMatix

xMatix Field Service provides an allocation console with three views of the same day: a map with live positions and travelled paths, an intraday timeline where dropping a job onto a lane schedules it, and a board where unassigned work is dragged onto a person. An optimize-day action fills the unassigned backlog into free slots, and deliberately leaves existing human assignments alone — so pressing it is never a risk.

Underneath, omni-channel routing matches work by queue, skill, capacity and presence, with accept, decline, transfer and reassign as first-class operations and a supervisor view of backlog and per-person load. Appointment scheduling adds slot capacity and holds, and an availability engine finds eligible, free bookable resources — people, equipment or a service bay.

Field execution is offline-first: job cards, checklists, parts consumption, labour hours, photos and geofenced attendance are captured without connectivity and reconciled through a durable outbox. Service commitments are measured against business-hours milestones, so a breach is visible before it happens.