Labour is a market, not a wage

The wrong number, asked confidently

Every origin conversation reaches the wage inside four minutes. On its own, the wage is close to useless. Course 26.6 shows why in cost terms. What a buyer pays for is a standard minute — the time one operator should need for one piece of work. The wage is only about half of what an hour costs, and efficiency sits in the denominator. So a cheaper operator in a slower plant is a dearer garment.

This lesson is the other half, and it is not about cost at all. A labour market is described by three numbers, and none of them is a price.

1. Availability — how many operators you can actually add in a month, at whatever you are willing to pay.
2. Turnover — what share of the people you have leave every month.
3. Depth — how much of the work the town already knows how to do, which sets how long a new operator takes to reach standard.

Two of the three are cluster facts, not country facts. A capital city and a district four hours away, under one law and one minimum wage, will give you completely different answers. And the three together produce a number a factory cannot manage its way out of.

Turnover multiplied by training time

Here is the arithmetic. It is short enough to do on a napkin.

If a plant's headcount is steady, then every month it hires exactly as many people as leave. Anybody hired within the last training period is still below standard. So at any moment:

operators still learning = hires a month x months to reach standard
                         = headcount x monthly turnover x months to standard

share of the line still learning = monthly turnover x months to standard

Two numbers, multiplied. Neither of them is a wage. And the answer is a permanent property of the plant's town, not an event.

Here are Norforth's four candidate origins, each measured at the factory it quoted.

A learner averages 48% of a trained operator's output across the training period. That is Norforth's own figure, from its four factories' payroll and output records. So a line that is 35.84% learners produces 1 - 0.3584 x 0.52 = 0.8136 of what the same line would produce fully trained.

Ferrand's line produces 19.4% more standard minutes per operator-hour than Redmire's, and not one point of that difference is management. Same machines, same method, same supervision, same effort. The difference is that one town returns 61 experienced applicants in a hundred and turns over 1.7% a month. The other returns 9 and turns over 6.4%.

The correction this makes to a rule you already have

Course 26.6 sorts every line of a factory's position into granted, given and built, and puts efficiency squarely in built — the column a factory owns. That sort is right, and this is the exception it needs.

A measurable share of what any plant calls its efficiency is a cluster fact. Redmire could match Ferrand's method, layout, mechanics and supervision exactly, and still be 19.4% behind on minutes out of an operator-hour. A fifth of its line is always somebody who started this month. So the honest form of the sort is this: efficiency is built on top of a floor the town sets, and the floor is given.

This matters in both directions, and it is easy to get backwards.

• A buyer comparing two factories' efficiency across two labour markets is comparing two different quantities. Ask for turnover and time-to-standard beside it, or the comparison flatters whoever is standing in the easier town.
• A factory benchmarking itself against a plant in a low-turnover market will conclude that its management is bad. Part of the gap is not available to management at all. The part that is available sits above the floor, and you can only see it once the floor is measured.

The mistake nobody made

In March, Nadia Ferrell of Almiraz won a programme that needed 180 more trained operators by 1 August. Every step that followed was correct.

The machines were specified, ordered and delivered in eleven weeks, then installed and running on time. The building next door was leased. Two supervisors were promoted and trained. The costing used Almiraz's real cost per standard minute, its real efficiency and its real overhead. Nobody inflated anything.

The number nobody put on the page was the one on Almiraz's own payroll.

headcount today                     620
leavers a month at 3.1%            19.22
new starters a month, measured        34
net addition a month               14.78

180 net additions                  12.18 months
available                              5 months

Five months of hiring at Almiraz's own best measured rate delivers 73.9 operators — 41.06% of what the order needed. The machines were never the constraint. They were simply the only part of the ramp that could be bought.

And there is a second cost, which arrived as two separate complaints. To hire 34 a month, Almiraz took people with no machine background. So by August the learners on the plant were 153.0 against a headcount of 693.9. That is a 22.05% learner share, against the 13.95% it started from. Output per operator-hour fell from 0.9275 to 0.8853, a fall of 4.54%.

In September the buyer's inline audit reported a rise in defects. In October the costing desk reported that efficiency was down and asked what had gone wrong on the floor. Two different people investigated, and neither found a cause, because there was no fault to find. Both tickets are the hiring rate, arriving two months apart wearing different clothes.

Depth is a different constraint from availability, and it binds differently

Availability sets how fast you can grow in the work you already do. Depth sets what work is possible at all, and no amount of hiring substitutes for it.

Oskavia's pool is deep in knits and denim. An operator hired there has almost certainly run an overlock and a flatlock — the two machines that join and cover a knit seam. The four-hour trial that Almiraz runs on every applicant sorts them in an afternoon. Nobody in that town has made a tailored jacket. That is not a recruitment problem with a longer timescale. It is a training-from-zero problem, and the people who would run the training do not live there either.

The two markets are good at opposite things.

• Abundant and shallow — Tamsett. You can staff a very large simple programme quickly. You will also carry a permanent learner share, and be poor at anything with many operations.
• Scarce and deep — Belrone and Marendal. You can make a complicated garment well. You cannot scale a basic programme fast, because there is nobody to scale it with.

Neither is better. They win different orders. And the reason a place is good at a category is more often its labour depth than its wage, which is where lesson 5 starts.

Act as an industrial engineer and do only arithmetic I can check. My figures: average headcount of sewing operators [NUMBER], leavers last month [NUMBER], operators who actually started last month [NUMBER], average months from joining to reaching the standard rate on my operations [NUMBER], the share of my applicants who arrive with at least a year on an industrial machine [PERCENTAGE OR UNKNOWN], and a learner's average output as a share of a trained operator's [PERCENTAGE, OR say UNKNOWN and use nothing]. The commitment I am considering: additional trained operators needed [NUMBER] by [DATE], starting [DATE]. Now compute five things and show every step. One, my monthly turnover as a percentage. Two, my steady-state learner share, which is turnover multiplied by months to standard, and say plainly that this is the permanent floor under my efficiency rather than a problem to be fixed this quarter. Three, my net additions a month, which is starters minus leavers, and the number of months the commitment actually needs at that rate. Four, if that is longer than the time available, the largest headcount I can honestly promise by the date, and what the learner share will be on that date if I hire at my maximum rate. Five, what my output per operator-hour would be at that learner share against today's. Then tell me, in two sentences, what I should say to the buyer. Do not assume any figure I marked UNKNOWN and do not substitute an industry average for it.