Nobody is average: sizing a workforce

The situation

Tunca has won the contract. On 3 April a fitting team starts work at Torvale's six depots, and the merchandiser is handed something a retail merchandiser never sees: the actual measurements of the actual people who will wear the garments. Four hundred and twelve volunteers, measured over eleven days.

In a range you work out a size curve from last season's sales. The people no size fitted simply did not buy. They are invisible, and they cost you nothing you can see. Course 16.5 covers what that does to a retail size curve: a size that sold out was never measured, so next year under-buys it. None of that applies here. Here the population is known, it is bounded, it is yours for three years, and every person the size matrix does not reach is a person your contract obliges you to clothe anyway.

A size is a matrix, not a point

The polo shirt runs in six sizes, exactly like a retail polo. The two protective garments do not.

A field jacket is worn over a thermal base layer, and on overhead-line work it is worn under a harness. It has to close over the mid-layer with the specified overlap, and it has to leave the harness attachment points clear. So the jacket is graded on chest and back length, each one separately: eight chest sizes by three body lengths. The trouser is graded on waist and leg: ten waists by three leg lengths.

Torvale runs two colours — orange for distribution crews, yellow for metering crews — so the contract is 120 stock-keeping units (SKUs) against a retail equivalent of twelve. Every one of them has to be available to a new starter inside ten working days, for three years.

That number collides immediately with the cloth. Course 8.2 covers the mill minimum, and why it is set by the dye vessel and not by your order. The arithmetic there is unchanged. What is different here is that you are not choosing between a minimum and a smaller buy. The tender has already committed you to make all 120. The minimum stops being a purchasing decision and becomes a fixed cost you have to spread.

Covering ninety per cent of one dimension is not fitting ninety per cent of people

Here a retail instinct produces a confident, expensive mistake.

Tunca's technologist looks at the 412 measured people and sets the jacket grade so that the eight chest sizes cover 94.0% of measured chests, and the three back lengths cover 91.5% of measured back lengths. Both are generous. Both are defensible. The obvious conclusion — that the matrix therefore fits about nine people in ten — is wrong, and it is wrong in a direction that costs money.

To be fitted, a person needs both dimensions inside the matrix. If the two were unrelated, the joint coverage would be 0.940 times 0.915, which is 86.0% — four points worse than either figure on its own. They are not unrelated: taller people are usually broader, so the two failures overlap, and the real figure is better than independence would give. Measured on the 412, it is 89.3%.

So the matrix reaches 89.3% of the workforce. 447 of Torvale's 4,180 people are outside it — one in nine, and not one of them was in anybody's cost sheet.

What the tail costs, and where the cost lands

A person outside the matrix gets a garment made to their measurements. Same certified construction, same cloth, same tape, same thread — but a single-garment marker, a single-ply lay, its own paperwork and its own fitting. Tunca's made cost for one is USD 118.40 against USD 52.90, a premium of USD 65.50.

Now spread that premium, because the tender was quoted as one price per garment.

Over three years the contract consumes 8,025.6 jackets, of which the tail accounts for 858.24. The tail's extra cost is USD 56,214.72. Spread across every jacket in the contract, that is USD 7.00 a jacket — 13.2% of the price that was quoted.

There is a shortcut worth keeping, because it needs no volumes at all: the tail's cost per garment is the tail's share of the workforce multiplied by the premium. One in nine people at USD 65.50 extra is USD 7.00 spread over everybody, whether the contract runs three years or ten. If you know the share and the premium on the day of the tender, you know the number.

Nobody made a mistake. The costing team costed the garment in the specification. The technologist graded to a defensible coverage. The fitting team measured properly. The tail exists because in a retail range there is no tail — a customer the matrix does not fit walks out, and that absence never reaches a cost sheet. The habit that is correct in one category is silently, precisely wrong in the other.

Three more things a workforce does that a shopper does not

The size is assigned, not chosen. A fitter measures and issues. That is not an administrative detail: a protective jacket two sizes too large has cuffs that catch on things, and one size too small does not close over the specified overlap. In a shop, a bad fit is a preference. Here it is a safety decision made by somebody who is not the wearer. That is why the fitting record has to be kept, and why a wearer who insists on a different size needs that recorded too.

The garment is fitted in the state it is used in. Course 4.4 is the reference for measurement method — the point, the state and the tolerance — and it is worth re-reading before a size set. The workwear part is only this: the state is over the mid-layer, with the harness on. A size set fitted over a t-shirt in a warm sample room will pass, and will be wrong for every person wearing it in February.

The matrix does not move for three years. Torvale's population drifts — the average chest measurement of a workforce moves with hiring and with age. A retail buyer re-cuts the curve every season from sales. Here you are holding a grade you set once, against a population you measured once, for the length of the contract. Lesson 6 prices what that drift does.

What to take to your own contract

1. Get the measured population if it exists, and ask for it in your clarification questions if the tender does not offer it. It is the difference between a costed tail and a tail you discover too late.
2. Never multiply per-dimension coverage, and never quote one dimension on its own. Count the people inside every dimension at once.
3. Price the tail on day one as share times premium, and put it on the cost sheet as its own line. Then a later argument about it is an argument about a number, not about whether it exists.
4. Fit the size set in the state the garment is worn in, over whatever else the wearer is contractually required to have on.