Field-Service Advisory

The utilisation trap: why a fully booked operation can deliver less

Utilisation tells you a technician's time was occupied. It does not tell you the operation produced more. Past a point set by variability, a fuller calendar buys churn, not throughput.

By Paula Navarro · Cosmicalley  |  July 2026  |  7 min read

Skilled-labour shortages have run at historically high levels across advanced economies for most of the past decade, and structural forces — ageing workforces, the green and digital transitions — are keeping them tight (OECD). For a service operation, that pressure has an obvious-looking answer: get more out of the technicians you already have. Fill the white space in the calendar. Push utilisation up.

It is the right instinct pointed at the wrong number. Because the fuller the calendar gets, the less capacity the operation keeps to absorb the ordinary variation that field work runs on — and past a certain point, raising utilisation lowers the work that actually gets finished.

Utilisation measures occupancy, not output

Technician utilisation answers one question: was the time booked? It is silent on every question that decides whether the day succeeded. It doesn't know whether the time completed useful work, prevented a repeat visit, protected an SLA, created overtime somewhere downstream, or forced a reschedule that broke someone else's route. A technician can be 95% utilised and finish less than one who is 80% utilised — if the 95% is spent absorbing the knock-on effects of a day scheduled too tight to recover.

A calendar can be full while the operating system underneath it is losing work.

Why the last few points of load cost the most

This isn't a motivational point about "breathing room." It's a structural property of any system where demand meets limited capacity, and it's well understood. Queueing theory — Kingman's formula is the standard statement of it — shows that as a resource's utilisation approaches 100%, waiting time doesn't rise in a straight line. It rises sharply, and the rise is amplified by variability in both arrivals and how long each task takes. Two operations at the same average load can behave completely differently: the one with more variable job durations and more unplanned demand hits the wall first.

Healthcare learned this the expensive way. A widely-cited BMJ simulation of hospital bed occupancy found the risk of being unable to admit a patient stays near zero below about 85% occupancy, then climbs steeply above 90%. But the more useful lesson is the correction that followed: later work, pointedly titled "the 85% bed occupancy fallacy," warned against treating any single number as a universal target. The threshold is real; a fixed value for it is not. It moves with how variable your demand and durations are.

What the metric never sees: the cascade

In field service the mechanism is concrete. A tightly loaded day has no slack, so a single disturbance doesn't stay local — it propagates. The first job overruns by forty-five minutes. The second customer gets a late arrival. Dispatch pulls a technician off a third job to cover, and the replacement doesn't carry the right part, so that job becomes a repeat visit. The fourth job slips to tomorrow, where it consumes capacity that was already booked.

At the end of it, the timesheet shows a highly utilised team and the service report shows missed appointments and overtime — at the same time. The metric and the outcome point in opposite directions, because the metric was measuring the wrong thing.

A better lens: three kinds of capacity

The fix isn't "lower utilisation." It's to stop treating a full calendar as the goal and start distinguishing three different things a single utilisation number silently blends together.

Seen this way, the trap is obvious: raising scheduled capacity toward its ceiling consumes resilience capacity, and once resilience is gone, productive capacity falls even though scheduled capacity looks perfect. The operation is busiest exactly as it becomes least reliable.

When a full calendar is fine

None of this argues for a permanently half-empty schedule — idle time is a real cost, and in the right operation high loading is entirely correct. The dividing line is variability. An operation running predictable preventive-maintenance routes, with low job-duration variance, stable geography, few emergency insertions, dependable parts and access, and flexible SLA windows, can safely plan close to its ceiling: it has little variation to absorb, so it needs little slack. A reactive, geographically dispersed network full of same-day priority changes cannot — its buffer is its productivity. The same utilisation target that is prudent for the first operation is reckless for the second. That is exactly why a single benchmark number is the wrong tool.

Before you raise the target or hire

So when productivity pressure arrives, the useful move isn't to reach straight for the utilisation dial or the headcount request. It's to find out where capacity is actually going. A few questions separate a genuine shortage from self-inflicted churn:

• How much same-day demand enters the system, and how much variance sits between planned and actual job durations?
• How often does one delay propagate into later appointments?
• What share of schedule changes happen after a job was prepared?
• How much overtime coexists with unfinished planned work — the signature of the trap?
• Which service centres reliably convert scheduled time into completed work, and which don't?

If overtime and unfinished work are climbing together, adding technicians into an unstable schedule mostly buys more expensive churn. The constraint isn't people. It's the resilience the plan gave away.