Six weeks to a forecast that outperforms yours See how
Back to planning notes

The same line ran at 37 percent and at 84 percent last week

Nov 13, 2026, 8 min read

In short

Capacity utilisation measurement has four defensible denominators, and one line reads 37 percent against calendar hours or 84 percent against available hours in the same week. The numerator carries the same ambiguity, because run hours, earned standard hours and good output hours describe different things about the same shift. Reclassifying a changeover overrun from unplanned loss into planned downtime raises utilisation by five points without producing a single extra unit. A utilisation target pushes a line towards long runs of the easy product, and the stock that creates shows up two quarters later.

A capital request for a second filling line reaches the investment committee with a chart showing utilisation at 84 percent and a note saying the site is running out of headroom. A finance analyst pulls a different report, which shows the same line at 37 percent, and asks why a business would buy a second one.

Both figures came from the same week and the same time records. One divides by the hours the line was available to run. The other divides by the hours in a week.

Capacity utilisation is the only planning metric where the denominator is more contested than the numerator, and it is routinely quoted with neither one stated.

Four denominators, one week

The line was scheduled for two shifts, five days, eight hours a shift, which is 80 hours. Six of those hours were set aside for planned maintenance and planned changeovers, leaving 74 hours available. The line actually ran for 62 hours, losing five to a breakdown, four to a changeover that overran, and three waiting for material.

Against the calendar, 168 hours a week, utilisation is 62 divided by 168, which is 37 percent. Against scheduled hours, 80, it is 78 percent. Against available hours, 74, it is 84 percent. Against demonstrated capacity, meaning what this line has actually sustained in its best recent weeks, it is a different figure again, covered below.

Each denominator answers a real question. Calendar utilisation answers whether the asset could be worked harder by adding shifts, which is the question the finance analyst was asking. Scheduled utilisation answers how well the shifts you paid for were used. Available utilisation answers whether the line ran when it was meant to, which is the maintenance and materials question.

Businesses get into trouble when the capacity plan uses one denominator and the capital case uses another. A site genuinely full at 84 percent of available hours on two shifts has an obvious option that does not involve capital, and a chart quoting only that number never surfaces it.

The shift pattern is the part of this that gets treated as fixed and rarely is. The line above is full at 84 percent of available hours on two shifts. Adding a third shift takes scheduled hours from 80 to 120, and if the same availability held, run hours would go to roughly 93 and weekly output to around 12,400 units. That is a 50 percent capacity increase for the cost of a crew, against a capital project with a lead time measured in quarters.

Whether the third shift is a good idea depends on crewing cost, on whether maintenance windows can be found inside a seven-day pattern, and on whether the demand justifies it. The point about measurement is narrower: a utilisation report that only ever divides by available hours will never surface the option, because the denominator moves with the shift pattern and the metric looks identical at 80 scheduled hours and at 120.

The numerator has the same problem

Run hours count time when the line was moving. Earned standard hours count the output at its standard rate, so a line running slowly earns fewer hours than it consumes. Good output hours strip out what was scrapped.

Take the same week. In 62 hours of running, the line produced 8,300 units where the standard rate implies 9,300, and 97 percent of those passed quality. Earned hours against available hours give 78 percent rather than 84.

The decomposition that keeps these separate is overall equipment effectiveness, set out by Seiichi Nakajima in his 1988 introduction to total productive maintenance: availability multiplied by performance multiplied by quality. Availability here is 62 over 74, or 83.8 percent. Performance is 8,300 over 9,300, or 89.2 percent. Quality is 97 percent. The product is 72.5 percent.

The value of OEE in this context is that it makes the losses additive and attributable. A utilisation figure of 84 percent contains no information about whether the gap is maintenance, rate loss or scrap, and the three have different owners and different remedies.

Capacity in units moves when the sales mix moves

Everything above is expressed in hours. Planning conversations happen in units, and the translation is where nameplate capacity earns its reputation.

The line's nameplate is 10,000 units a week, a figure computed at the standard rate across the scheduled hours, at the product mix that was current when the line was commissioned. This week 40 percent of the volume was a variant with a 30 percent longer cycle, which costs about 12 percent of the rate, so the scheduled hours could have produced roughly 8,800 units at best.

Against nameplate, the week's 8,300 units read as 83 percent. Against what the week's mix allowed, they read as 94 percent.

Demonstrated capacity gives a third view. Take the best four of the last 26 weeks and average them: suppose that comes to 9,100 units. This week's output is 91 percent of it.

Nameplate is the number in the asset register and the least useful of the three for planning, because it describes a mix nobody sells any more. Demonstrated capacity is the honest one for a supply plan, and it needs a stated rule: how many weeks, over what window, whether weeks containing a breakdown are eligible. Without that rule it drifts upward every time somebody has a good week and quotes it.

The edit that raises utilisation without producing anything

Move the four hours of changeover overrun out of unplanned loss and into planned downtime, on the grounds that the changeover was scheduled and simply took longer than expected. Available hours fall from 74 to 70. Run hours stay at 62. Utilisation rises from 84 to 89 percent.

Nothing was made. Nothing was fixed. A category on a downtime code list moved.

This is the single most common distortion in utilisation reporting, and it happens without intent, because coding downtime is done at shift level by people who reasonably classify a known event as planned. The protection is to fix the denominator by policy: available hours equals scheduled hours minus a maintenance allowance agreed once a quarter, with all other losses recorded in the numerator's loss categories rather than deducted from the base.

What a utilisation target makes people do

Long runs of the easy product. Changeovers consume available hours and produce nothing, so a line measured on utilisation will lengthen campaigns and defer the small, awkward orders. The stock this creates is real and shows up two quarters later.

Building to stock in quiet weeks. A target that must be hit every week converts a demand trough into inventory, since the alternative is a visible dip in the metric.

Deferring maintenance. Planned maintenance reduces available hours in the week it is taken and increases breakdown hours later. The metric charges the first and hides the second inside an event nobody planned.

Running the non-constraint. Output from a machine that feeds a slower one becomes work in progress. Goldratt's argument in The Goal, published in 1984, was that an hour gained at a non-constraint buys nothing, and a site that measures every asset on utilisation will generate exactly that hour.

There is one more consequence, and it is the reason a high number is not automatically good news. Waiting time in any process with variable arrivals and variable service times grows faster than linearly as utilisation approaches one, so the last few points of utilisation buy queue and lead time rather than throughput. N18 works that relationship through in detail on a marine terminal, and the shape of it applies to a filling line, a paint booth and a goods-in dock without modification.

What to pair it with

Output in units of what you sell, on the same page. Utilisation is a means, and a week where utilisation rose and saleable output did not is worth a question.

The loss breakdown behind the number. Availability, rate and quality separately, in hours, so the gap has an owner.

Lead time or work in progress at the same asset. These are what tell you whether high utilisation is being paid for downstream, and they move before the service metric does.

Schedule adherence for the same period. A line at 90 percent utilisation making the wrong products is a specific and common failure, and BB9 covers how that gets measured without flattering the plant.

The mix assumption written next to any capacity figure in units. One sentence: this capacity assumes the following share of long-cycle variants. It stops the annual argument about whether the line has slowed down.

Where this stops

Utilisation on a non-constraint asset carries almost no information about the system's ability to deliver. Most sites have one or two constraints at any given time, and the constraint moves with the mix, so a fixed set of utilisation reports across twenty assets will spend most of its attention on machines whose spare capacity is irrelevant. Identify the constraint for the current mix first, then measure it closely and the rest loosely.

Capacity is also a distribution rather than a number. It depends on mix, on crewing, on which operators are on shift, on ambient conditions in some processes, and on the yield of the material being run. Quoting a single figure to three significant digits in a supply plan implies a precision the underlying process does not have, and the sensible practice is to carry a range and to plan against the lower end for commitments.

And a labour-constrained operation needs a different measurement entirely. Where the equipment is idle because nobody is available to run it, equipment utilisation describes the wrong resource, and the useful denominator is available labour hours by skill. That distinction gets lost in businesses that historically were equipment-constrained and have quietly become labour-constrained without changing their reports.

The test for which one you have is quick. Take the twelve hours the line lost last week and attribute each one to the resource that was missing at the time: equipment, material, or a qualified operator. Do it for a quarter. If a third of the lost hours trace to crewing, the capital case in the opening is answering a question the site does not have, and the money would be better spent on the training matrix that decides how many people can run the line on a Tuesday night.

That attribution also settles a recurring argument about whether utilisation should be measured on the asset or on the cell. Where one operator runs three machines, the machines cannot be treated as three independent denominators, and summing their individual utilisations produces a number with no physical meaning. Measure the resource that binds, and measure the others only closely enough to know when the binding constraint has moved.

Write your denominator on the face of next week's utilisation report, in hours, with the maintenance allowance stated: if two people cannot reproduce the same figure from that definition, the number in the capital case has not been agreed yet.

More on this in metrics on the DemandLabs platform.

See what DemandLabs could do for your plan

Request demo

More planning notes

View all

Nov 12, 2026

Your bias number is near zero and the warehouse still filled up

The quarterly pack has two numbers on the forecast slide. Accuracy at 78 percent, flat on last quarter. Bias at 0.2 percent, with a footnote saying it sits inside a tolerance of plus or minus two. Nobody asks about either one, because both look fine.

Nov 11, 2026

Service went up after go-live and so did four other things

The quarter three review has the chart on it. Line fill was 94.1 percent in the six months before go-live and 97.3 percent in the six months after. That is 3.2 points, roughly a million and a half of recovered margin on volume that would otherwise have been missed, and the programme is declared a success on the second slide.

Nov 10, 2026

Half your benchmark gap is a definition and a product mix

The slide has two bars on it. Yours at 2.1 inventory turns, a peer group median at 3.8, and a gap converted into something like eleven million of released cash. By the time the deck reaches the operating review, the number has been written into next year's plan and given an owner.