Word of mouth that compounds

Two doors down from a deck you rebuilt last April, someone is standing in a kitchen asking her neighbor who did it. You are not in that room. You did not ask for the mention. Whether your name comes up at all — and whether it comes up as "he was great, here's his number" or "he was fine, took longer than he said" — was settled months ago by how that job ran and how you left the site on the last day.

That conversation is the whole channel. You cannot attend it, script it or buy it. What you can change is how often it happens, and how often turns out to be arithmetic.

The one number that decides whether it compounds

Word of mouth either adds a little or multiplies. The difference is a single figure: how many new inquiries the average finished job eventually produces. Call it your referral rate.

Jobs per year at steady state = base jobs / (1 - referral rate x close rate)

Base jobs are the ones that arrive regardless — search, your Google profile, a lead platform, signage, repeat customers. The referral rate is inquiries per finished job. Close rate is the share of those you win, and it usually runs higher on a referred call than a cold one, because the trust part happened before the phone rang.

Every number below is invented to show the shape. Take a shop that would land 20 jobs a year with no word of mouth at all, closing 70% of referred inquiries.

That is not a straight line. Moving from 0.15 to 0.30 buys you three jobs a year. Moving from 0.80 to 1.00 buys you twenty-two. Under about half an inquiry per finished job, word of mouth is a pleasant supplement to whatever you pay for leads. Over one, it is the business and the rest is decoration.

Your own rate is knowable: for the next fifty inquiries, ask how they got your name and write down the actual sentence, not a category.

Why the first three years look like nothing

The multiplier is a steady state, and steady states take time to arrive. Same shop, referral rate 0.80, referrals landing roughly a year after the work:

Four years to get most of the way, and the biggest jump lands in year two, long after the effort that caused it. This is why word of mouth loses every head-to-head comparison with paid leads in month three and wins it in year five. Paid leads stop the day you stop paying; a finished job keeps working, and its cost already went out the door as labor you were paid for.

The lag depends on how often your work comes up in a household. A handyman's neighbor needs something in six months. A re-roof, a sewer line or a full remodel — that neighbor's turn arrives in a decade, so the ramp stretches and the early years look flatter still. It is why the roofer who has worked one county for twenty years cannot be beaten on ad spend.

Variance, not excellence, holds the number down

Most contractors try to raise the referral rate by being better on the good jobs. The lever is the other end: the rate is an average across everything you finished, and a bad job subtracts.

A bad story also travels further, for a structural reason rather than a psychological one. It has a plot. "He did the bathroom, it looks nice" ends the conversation. "He disappeared for nine days and I couldn't get him on the phone" gets retold, because it has a middle. Skip the old statistics about how many people hear a complaint; they are mostly vendor-sponsored folklore. The direction is the part you can act on.

Variance usually comes from four places, and none of them is skill:

• The squeeze job. The small favor you fit between two real jobs, at a price that never justified doing it properly.
• The day you were not there. A helper or a sub on site alone, on a job the customer thinks you are personally running.
• The last five percent. Punch list, caulk line, the trim piece on order. The customer's memory of the whole job forms in that week.
• The callback in month five. How you answer the phone when there is no money in it is the single most-repeated thing about you.

Two networks, and they behave differently

News moves through a local market in two circuits. The block is one: a truck in a driveway, a neighborhood Facebook group, an HOA thread, a school pickup line. It produces jobs the size of the one you just did, in a radius you could walk. One completed job in a subdivision of identical houses is worth more than three scattered across a county — the honest argument for taking the second job on a street at a price you might otherwise pass on.

The trade circle is the other: the plumber who needs a drywaller, the realtor with a closing in eleven days, the property manager, the GC who lost a sub. These arrive less often and land bigger, and they are earned on a different axis — showing up when you said, pricing a change order without a fight, and telling someone their job is not yours instead of taking it. Referring work out to a good tile setter is the cheapest thing you will ever do for your own pipeline.

The habits that are worth a tenth of a point

Nothing here is clever. Each is a small tax on a busy day that shows up in someone else's kitchen a year later.

• Leave the site cleaner than the trade before you did. Shop vac, no cut-offs in the shrubs, screws off the driveway.
• Send one mid-job text the customer did not have to ask for. "Rough-in inspection Thursday, still on track" prevents the nine-day story entirely.
• Be the same person on the phone as on site. If the estimate voice is warm and the day-eight voice is short, day eight is what gets described.
• Say bad news the day you learn it, not the day it becomes visible.
• Point out one thing you noticed and did not charge for.
• Answer the warranty call inside a day, in year two, when you are slammed.

Where the follow-through usually slips

The gap between the two voices is usually a paperwork gap, not a character flaw. The estimate is careful because you sat down to write it; the change on day eight gets a verbal maybe from the top of a ladder, and that is the version the customer retells. The fix is having somewhere for the second one to go. DoneQuote is where you write it down — the added scope, the revised total — so what the customer ends up holding is what you actually meant, and you are not reconstructing a promise from memory three weeks later.

What a tenth of a point is actually worth

Run the formula on a small improvement. Same shop, base 20, close rate 70%, referral rate moving from 0.40 to 0.50: steady state goes from 28 jobs a year to 31. Three jobs a year, permanently, for no ad spend and no new tool — bought with a shop vac and a text message.

That is the case for treating this as a system instead of luck. The compounding is real, the lag is real, and the input is not your best work. It is your worst.