Utilization Read Ninety Percent While the Crew Was Idle

Why this matters

A capacity number that is wrong in the flattering direction does not sit there quietly. It gets used. It talks a shop out of hiring, out of chasing work, out of asking why the phone is slower than last year, and it does all of that with the authority of a measurement. The shop below spent most of a quarter believing its crew was at the edge of its capacity while a third of its paid field days carried no work at all, and nothing in the report was arithmetically wrong.

The signal

Two in the afternoon on a Tuesday in what should have been a busy month. The owner walks through the yard and finds two technicians at the bench with nothing on, not waiting for a part, not writing anything up. The quarterly utilization report on his desk read 90 percent.

Both readings were about the same crew in the same quarter, and they could not both be true. Somebody is either standing around or at the top of their capacity. That contradiction, and not the low mood in the yard, is what makes this worth chasing: a number that disagrees with what you can see is either wrong or is measuring something other than what its name says.

The quarter: 7 field technicians on payroll, 13 weeks, 65 working days, a nominal 8-hour field day. Logged field hours for the quarter, 2,184.

Candidate one: the hours are padded

If technicians were logging more time than they worked, a real denominator would still produce a high ratio. This is the first place anyone looks and it is worth killing properly rather than on instinct, because killing it wrongly poisons everything that follows.

Two pieces of evidence, both from records the shop already had.

Payroll paid 3,198 hours over the quarter, counting attendance, shop time, meetings and PTO taken. Logged field hours were 2,184, so logged hours came to 68.3 percent of hours actually paid. Padding runs the other way: a crew inflating its time logs pushes logged hours up toward paid hours and sometimes past them. A third of paid time never reaching a job is the normal shape, not the suspicious one.

Then the job-level check: across the shop's three most common job types, the median logged on-site time sat within 0.2 hours of the job template's own estimate on all three. Padding does not produce that. It produces a median that runs over estimate on the job types a technician runs most often, because those are the ones where the extra half hour is safest to add.

Candidate one out. And the 68.3 percent figure, which nobody had ever computed, turned out to be the first useful thing anyone found.

Candidate two: the jobs are not real

The other way to inflate a ratio is to inflate what counts as work. Double-counted jobs, a job split across two records, administrative entries closing as field work.

Reconcile the job count against billing. The quarter completed 612 jobs. Of those, 566 were invoiced and 46 were warranty or return visits correctly carrying no invoice, and 566 plus 46 is 612 with nothing unaccounted for. The unbilled 46 are 7.5 percent of completed jobs, which matched the shop's own trailing three quarters, so the free-visit load had not moved either. Nothing was being counted twice and nothing phantom was being counted at all.

Candidate two out. The work in the numerator was real work.

Candidate three: the report is describing a busier month

A quarterly figure can be carried by a strong first month and still read healthy while the last few weeks are dead. That would reconcile both observations without anything being wrong with the ratio at all.

Rebuild the same reported figure week by week on the same basis. It sat between 88 and 93 percent in every one of the 13 weeks, including the week the owner walked the yard. The report was not stale and it was not an average hiding a collapse. It was claiming, about that specific week, that the crew was at 90 percent.

Candidate three out, and that is the point where attention has to move from what the number is describing to how it is built.

The question nobody had asked

What is the denominator made of?

The capacity figure was assembled week by week from the technicians who appeared on that week's schedule, at 8 hours for each day they were scheduled. It is an entirely reasonable-sounding construction and every shop that builds capacity from its own schedule arrives at it.

Count it: 303 technician-days across the quarter carried at least one assigned job. The denominator was therefore 303 x 8 = 2,424 hours.

The payroll field roster says something different. Seven technicians across 65 working days is 455 technician-days available, or 3,640 hours. The gap is 152 technician-days, which is 33.4 percent of the quarter's available field days, and 1,216 hours of paid capacity that the report never mentioned in either the numerator or the denominator.

The two technicians the owner saw at the bench accounted for 61 of those 152 unassigned days out of the 130 available between them, which is 46.9 percent of their own days. They were the worst of it and they were not an anomaly: the other five carried 91 unassigned days out of 325, or 28.0 percent.

Three ratios, three denominators

What is being asked Denominator Hours Figure
How full were the days we had already committed? Assigned technician-days 2,424 90.1%
How much of the field capacity we pay for reached customer work? Payroll field roster 3,640 60.0%
How much of what payroll actually paid reached a job? Hours paid, attendance 3,198 68.3%

Read the third row as a diagnostic and never as a target: the paid time that does not reach a job is largely shop time, meetings, required training and mid-day travel between jobs, which are generally compensable hours worked under 29 CFR Part 785 and cannot be scheduled away, and several states require paid rest breaks on top. The figure tells you how much paid capacity is available to sell, not how much of it is waste.

The same 2,184 logged hours, three questions, three answers 30 points apart. The reported 90.1 percent was not false. It was a correct answer to the first question, printed under a heading that made everyone read it as an answer to the second.

Two more figures from the same numbers finish the picture. On the days somebody did have work, the crew logged 2,184 / 303 = 7.2 hours of an 8-hour day, which is genuinely busy. Across all available days it was 2,184 / 455 = 4.8 hours. That pair is why both observations were true at once: the busy days were very busy, the technicians on them felt stretched, and a third of the days were not busy at all.

Why this construction can never report idleness

An unassigned day contributes nothing to the numerator, because no hours are logged on it, and nothing to the denominator, because it never appears on a schedule. It leaves both sides of the fraction.

So a ratio built from assigned days is structurally incapable of reporting slack. Cancel every job in the shop except one, and the figure describes that one technician's day and reads beautifully. The worse the quarter, the fewer days enter the calculation, and the more the survivors resemble a busy shop.

That is the property to look for in any capacity ratio somebody hands you: does an idle unit of capacity appear in the denominator, or does it simply vanish? If it vanishes, the number cannot fall when the thing it measures gets worse, and a number that can only move one way is not a measurement.

It is worth saying that the usual denominator errors run the other way. Counting a part-timer as a full week, carrying a leaver for the whole quarter, leaving a coordinator on the technician list: all of those inflate assumed capacity and make the figure read pessimistically low. The reference on what the utilization denominator assumes works through that set. This shop was in the opposite failure, which is exactly why nobody went looking there. Their number was too good to audit.

The correction, and the habit that keeps it true

The denominator now comes from the payroll field roster, not from the schedule, with a written rule for the three cases that are otherwise argued about every quarter: a part-time technician contributes their standing hours, not a full week; somebody who joins or leaves mid-quarter contributes the weeks they were present; and a name that logs no field hours by design is not field capacity and comes off the list entirely.

The roster is reconciled monthly against payroll, by name and not by count. Timing it to the payroll run matters, because that is when a headcount change has already been made somewhere in the building. Matching by name matters because a count match can hide a leaver and a joiner cancelling each other out, and that is precisely the month where the two sides quietly stop describing the same people.

Assigned-day coverage now sits beside the ratio, every period: technician-days carrying at least one assignment over technician-days available, here 303 / 455 = 66.6 percent. It is one division, it uses data the shop already had, and it would have shown the problem in the first week rather than in the twelfth. It is also the same move the neighbouring coverage figures need, for the reason derived in the card named in the references.

The decision it changed

The corrected figure was 60.0 percent against a crew that was being paid for 455 technician-days and working 303 of them.

Note that 60.0 percent sits at the bottom edge of the healthy band for a payroll-roster denominator, where 60 to 70 percent of paid field capacity reaching customer work is a common starting point to tune against your own trailing four quarters, rather than below that band, so the level on its own would have prompted a shrug. The decisive evidence is the assigned-day count, not the ratio: 152 of 455 field days carried no work. That is the cut that separates a sales problem from a scheduling one, and it landed unambiguously. Scheduling can move work between days and it cannot invent work that never came in.

The 1,216 unbilled-capacity hours are worth stating in work rather than in money. At this shop's own 2,184 / 612 = 3.57 logged hours per completed job, 1,216 hours is roughly 341 jobs' worth of capacity, about 56 percent more than the 612 the quarter actually completed. Read that as a ceiling on what the existing payroll could have carried, not as a forecast of what it would have sold, because the constraint that turned out to be binding was demand.

Which is the real finding, and it took a corrected denominator to see it. The shop had been treating a sales problem as a capacity success for most of a year. The response was on the sales side, not the scheduling side, and the thing that had been preventing it was a number reading 90 percent.

References

  • See related: Technician Utilization and What the Denominator Assumes - the three denominator errors that run the opposite way, and the band this figure should be read against
  • See related: SLA Compliance Only Counts Jobs That Carry a Deadline - coverage as the companion figure to any rate built on rows somebody had to create
  • See related: Jobs per Day per Technician: Why More Is Not Better
  • See related: Zero-Revenue Jobs and Which Numbers They Touch