One-task subcontractor

Why does a capable crew pay an outsider for work they could learn in an afternoon?

A site crew who can frame a roof, hang doors and lay a floor will stop dead for a 90-minute job — coring a hole through a concrete slab, commissioning a boiler, terminating a fibre — and pay someone to come and do it. Ask them why and you often get "we're not certified", which is true but circular. Certification is itself a choice somebody made.

The sharper version of the puzzle is this: the procedure really is learnable in an afternoon. Watch the corer once and you could describe every step. So if the knowledge is that cheap, why does the money still move?

Let us take the claim seriously and ask what an afternoon actually buys. It buys the steps in the normal case. Does it buy the abnormal case? The corer's afternoon-worth of steps assumes he does not hit rebar, does not hit a post-tensioned tendon, does not find the slab thinner than the drawing says. What separates him from you is not the sequence but the ability to notice, in the first ten seconds of a cut going wrong, which of those it is. That comes from having done it several hundred times, not once.

So the first thing specialisation concentrates is not knowledge but repetitions. And here is the awkward arithmetic: a task the crew meets four times a year can never accumulate them. Four goes a year is roughly one relearning event each time, because the skill decays between uses. The specialist meets it four times a day. The gap does not close with effort; it is set by the frequency of the work, and frequency is precisely what a generalist cannot have.

Now ask about the kit. The rig, the vacuum base, the slurry control, the calibrated test set — suppose it costs as much as a small van. For the specialist that cost is spread over hundreds of jobs a year, so it lands as a few pounds per job. For the crew it lands over four. The same machine is cheap for one and absurd for the other, and nothing about the crew's competence changes that. This is the second thing specialisation concentrates: fixed costs meeting enough volume to be amortised.

Then the paperwork, which is really liability. The certificate is not a gate on knowing; it is a way of making one identifiable party answerable if the boiler leaks in three years. Buying the specialist buys a named person carrying professional indemnity and a warranty that outlives the job. The crew could learn the task, but they cannot cheaply learn to be insurable for it.

There is one more step, and it is the one that surprises people: even if the crew could do the job exactly as well and as cheaply, subcontracting can still be right. Their hours are not free. An hour spent coring is an hour not spent framing, which is the work they are unusually good at. The comparison that governs the decision is not "can we do this?" but "what does doing it cost us in the work only we can do?" — the comparative-advantage argument, and the reason absolute competence is the wrong test.

Does that mean subcontract everything? No, and the boundary is visible in the same terms. When the crew's volume of a task crosses the point where the tool pays for itself and the repetitions arrive often enough to hold the skill, the work comes in-house — which is exactly what firms do as they grow. The line moves; it does not vanish.

Two clarifications keep this honest.

First, "they could learn it in an afternoon" is usually measuring the wrong thing. The afternoon transfers the explicit part of the skill — the part that can be said out loud. What stays with the specialist is the tacit part: which resistances in the handle mean stop, which smell means a bad joint. That part is not withheld; it simply cannot be transmitted at the speed of a demonstration.

Second, this is not an argument that specialists are better people at the task in some general sense. Take the specialist off their one job and onto the crew's daily work and they will be the slow one. Specialisation is an allocation of practice, not a ranking of ability — which is why the same crew is somebody else's specialist on the job they do every day.

• How the make-or-buy boundary shifts as a firm's volume of a task grows.
• Why coordination cost sets an upper limit on how finely work can be divided.
• Licensing as a mix of genuine safety gate and incumbent protection, and how to tell them apart.

Every term the collection defines is gathered in the glossary.