Trade demarcation
A Socratic walk-through of trade demarcation — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why does a building site split the work among a dozen trades that each stand idle waiting for the others?
Walk onto a building site on a Tuesday afternoon and count the people actually working. It is a small fraction of the trades the job needs. The plasterer is not there; he is waiting on the first-fix electrician, who is waiting on the carpenter. The usual verdict is restrictive practice — a dozen guarded jurisdictions producing a dozen queues. But notice what the complaint assumes: that the waiting is a cost, and that somebody on the site is paying it.
Reasoning it through
REASONING #Ask first who the idle plasterer works for. Not the site. He works for a plastering firm serving many sites, and the contractor has bought a window of his presence, not his year. "Idle on this job" and "idle" are different statements, and an observer standing in one place sees only the first. The waiting is absorbed by a pool of demand the site cannot see.
That reframing has a testable consequence. If demarcation were mainly restrictive practice, it should track the power of the trades. If instead it is a way of sharing lumpy specialists across many jobs, it should track the repeatability of demand: wherever one producer can keep every trade continuously fed, the reason to buy them by the window disappears. Which happens? The second. Volume housebuilders producing near-identical units, shipyards and offsite modular factories all integrate trades and blur the boundaries between them, while demarcation is sharpest on one-off bespoke building — the opposite of what the restrictive-practice story predicts.
So why not a single multi-skilled builder who never waits for himself? Partly the obvious: each trade takes years, and the generalist does none well. But there is a sharper reason with nothing to do with skill. Several of these boundaries are boundaries of legal answerability. Electrical and gas work must be certified by someone qualified to sign for it, and the signature is the product: a bricklayer who could competently run a cable still cannot, because no inspector will certify it and no insurer stand behind it. Add the tools and plant each trade owns, and a trade is a bundle of skills, capital and liability sellable as a unit.
Now the sequence. Structure before first fix, first fix before plaster, plaster before second fix, second fix before decoration. These are not scheduling preferences; work in place physically blocks and enables other work, and drying and curing take the time they take. A site is a job shop with hard precedence constraints and a shared, congested workspace — and wherever precedence meets variability, queues form. That is arithmetic, not a management failing.
Which brings the real question into view. In any such system you can keep the resources busy or keep the work flowing, not both: variability is absorbed either by things waiting for people or by people waiting for things. A factory chooses the first, keeping machines and workers loaded while part-finished work queues in front of them. A site chooses the second, and the choice is not arbitrary.
Look at what is waiting in each case. On a factory floor the work-in-progress is a pallet of components, and storing it costs a little space and cash. On a site the work-in-progress is the building — financed, insured, weather-exposed, on land producing no revenue, with a completion date and penalties attached. Its holding cost per week is enormous. So the site keeps the expensive thing moving and lets the cheap thing wait. The cheap thing is the trade, and it is cheap to leave waiting precisely because it is pooled across other sites and is not, in aggregate, waiting at all.
That is the whole inversion. A factory makes the product wait to keep the worker busy. A site makes the worker wait to keep the product moving. Neither is confused; they simply face opposite holding costs.
The analogy
THE ANALOGY #Think of a hospital operating theatre. The surgeon does not stay in the room from admission to discharge; anaesthetist, surgeon, radiographer and physiotherapist each appear for their interval and leave. The theatre and the occupied bed are the expensive things and are kept moving; the specialists are shared across the hospital, so their gaps in any one case are somebody else's list.
the hospital owns all its specialists and reschedules them centrally, whereas a site's trades belong to separate firms with commitments elsewhere, so the site cannot pull one forward when the sequence slips — which is why site delays cascade as theatre lists usually do not.
Clarifying the model
THE MODEL #Two refinements, a boundary, and a genuine dispute.
The idle time is not free, even when it is somebody else's. Every gap is a mobilisation and a demobilisation: travel, unloading, setting out, learning this particular building. That fixed cost per visit is why trades want long continuous runs, and why programmes minimise returns to site rather than anyone's downtime. It also makes the arrangement self-locking — because each trade attends in a window, work must be packaged into trade-sized blocks, and because it is packaged that way, the trade attends in a window.
A boundary worth stating: this is not the ordinary case for specialisation, which explains why people divide work at all. It is the narrower question of why site divisions are hard-edged and the resulting waiting tolerated — an answer about pooled external labour and lopsided holding costs, not about relative productivity.
The dispute: a substantial body of construction-management work argues much of this waiting is genuine avoidable waste, and that reliable-flow methods can compress programmes considerably. Against that, historical demarcation disputes were sometimes plainly about protecting work rather than organising it. How much of the idleness is structural and how much is rent is honestly contested.
A picture of it
THE PICTURE #How to readthis is a task board repurposed to show trades rather than tickets, and it is a snapshot of one week on one building, not a schedule. Each column is a state a trade occupies with respect to this site, and the crucial reading is that only the "on site" column costs the site anything — the cards in the other three are not idle men, they are firms working elsewhere. Cards move leftward into that column as the precedence they wait on is satisfied, and the column that must never be empty is the working one, because an empty working column means the building itself is waiting.
What became clearer
WHAT CLEARED #The idleness on a building site is largely an artefact of where you are standing. Trades are bought by the window from firms serving many jobs, so their gaps on your site are somebody else's working week — and the boundaries between them are drawn by certification and liability as much as by skill. What remains is a real choice about who absorbs variability, and the site makes it the opposite way from a factory for a good reason: here the work-in-progress is the building, and the building is far more expensive to keep waiting than the plasterer.
Where to go next
ONWARD #- How offsite manufacture changes the calculation by turning parts of the building back into factory work-in-progress.
Key terms
TERMS #| Term | What it means |
|---|---|
| Demarcation | the boundary defining which trade may perform which work, often reinforced by certification and insurance rather than skill alone. |
| Work-in-progress holding cost | what it costs to have partly finished work sitting still, which on a site includes finance, insurance, weather exposure and delayed handover. |
Every term the collection defines is gathered in the glossary.