Word-of-mouth trade pipeline

Why can a tradesman's diary go from empty to booked for years without any advertising?

A self-employed plasterer spends two years scratching for work, taking whatever comes, wondering whether the trade will keep him. Then, without changing anything he can name — no advertising, no van livery, no price cut — he is turning jobs away and quoting dates eight months out.

Nothing in the story has a step change in it. His skill improved gradually, his customers arrived one at a time. So where does a sudden transition come from, when every input to it moved smoothly?

Start by asking what a finished job actually leaves behind. Not just a wall: a customer who will, at some point, mention him to somebody who needs a plasterer. Some of those mentions become jobs.

So define a single number — call it the referral rate — as the average number of new jobs one completed job eventually produces through the people who saw it. Not the number of compliments; the number of jobs. If the number is below one, each job produces less than a full replacement, and the work he has generates a shrinking succession of descendants. He must keep feeding the pipeline from outside: adverts, an agency, a builder who subcontracts to him. If it is above one, each job more than replaces itself, and the descendants of a single job grow without bound until something stops them.

Now notice what that implies about smoothness. The referral rate is built from ordinary, continuous things: how pleased people are, how sociable the area is, whether he turns up when he said. Nudge any of those and the rate drifts smoothly — but the behaviour of the pipeline does not, because there is a threshold at one. Below it, decay; above it, compounding. A gradual improvement in workmanship can walk the rate from 0.9 to 1.1 and flip the sign of the whole system.

Then ask why the flip is not visible at once. Two reasons. The first is that early exponential growth is small in absolute terms and looks exactly like stagnation — two jobs becoming three is not a story anybody notices. The second is the lag. A referral is not collected when it is made; it waits until the neighbour's own ceiling cracks. That may be years. So the doubling time of this pipeline is set by how often people need the trade, not by how fast he works — which is why it takes so long, and why the trades whose customers need them yearly build books faster than those whose customers need them once a decade.

Finally, ask what stops the growth, because a diary booked for years is not unbounded growth — it is growth that has hit a wall. One man can only work so many hours. Past that point the excess demand cannot become more work, so it becomes queue. The exponential does not stop; it just stops expressing itself as output and starts expressing itself as lead time.

That is where price enters, and it is the part most tradesmen leave alone. A waiting list is unpriced excess demand. He can hold his price and grow the queue, or raise it and convert the same demand into income while the queue shortens. Raising it also feeds back on the referral rate, since price affects who says yes — so the two levers are not independent. I should be honest that nobody can hand you a number here: the referral rate is not measured in practice, and its size varies enormously by trade, locality and how mobile the customers are.

Two refinements keep this from becoming folklore.

The first is that reputation here is a rate, not a stock. It is tempting to picture a reservoir of goodwill filling up until it spills. But a reservoir would produce a smooth, saturating curve, not a takeoff — the takeoff only appears if each job generates further jobs, which is a multiplication, not an accumulation. That distinction is testable: if the pipeline were a reservoir, a man who did fifty small jobs would be as well placed as one who did fifty jobs that fifty different households saw. The referral account says the second man is far better placed, because what compounds is the number of independent witnesses, not the volume of work.

The second is that the same arithmetic runs in reverse, and faster. Dissatisfied customers are generally readier to volunteer their experience unprompted than satisfied ones, which puts a heavier coefficient on bad jobs than good ones. The direction of that asymmetry is well attested in service research; the ratios people quote for it are not reliable, so treat it as a sign, not a size. It explains a rule that otherwise sounds sentimental — going back to fix a small fault for free is not generosity but arithmetic, because a job that ends badly does not merely fail to reproduce, it actively suppresses.

Here is what would refute the whole account. If diaries filled in proportion to years in the trade rather than to the number of separate households served, the referral loop is not the driver and something else — ageing into a network, say — is doing the work.

• Why some trades reach the threshold in months and others never do, and what the need-arrival interval has to do with it.
• What the right response to a years-long waiting list is: raise the price, take an apprentice, or narrow the work — and how each one changes the referral rate itself.

Every term the collection defines is gathered in the glossary.