Every lead machine shows you one number.
It isn't the one that counts.

Here's the whole page in one line, before the rest earns your time. Judge every paid lead source by one number: what a real, closed job actually cost you, figured the same way for all of them. Not the price on the ad. Not the leads that came in. The job you finished and got paid for.

You already know the feeling that brings you here. Marketing is the necessary evil you put off, and when you finally spend, you spend on hope — the money goes out the door and the working part is the thing you're crossing your fingers on. This page swaps the hope for arithmetic you already do every day.

Because you do do this math. On a real job you build the price up from what it costs you to do it — the copper, the water heater, the hours, the fuel, your margin. You'd never bid off a number with nothing behind it. A lead machine is the one "job" the whole trade buys blind, on a promise your phone will run off the hook.

Four machines, four numbers, none of them the one that counts

Every lead source hands you a sticker price. The trouble is that four different machines quote you four different numbers, and none of them is the number that decides whether you made money.

Here's the shape of each one, from the sellers' own pages.

The machine	What it charges you for	The number it shows
Angi / HomeAdvisor	A fee every time it hands you a match. Its whole U.S. lead business is fees pros pay for matches — $587.1 million of them last year, called "a significant volume of low-dollar transactions."	Price per lead
Thumbtack	A charge when a customer contacts you. You set the price, and that set price is what you pay per lead.	Price per contact
Local Services Ads (Google's pay-per-lead listings)	A fee for each valid lead — a call or a message — not for the click.	Price per lead
Google Ads / Yelp Ads	A charge each time someone clicks, at a price that floats with supply and demand.	Price per click

One word in that table is doing hidden work: valid. Google decides what counts. Its help page, as of July 2026, says it no longer supports credits for "job type not serviced" and "geo not serviced" leads — wrong trade, wrong town. Chapter 4.4 prints the walls. For this math, the rule: a bad lead still counts as money you paid until the credit shows on your balance.

Per lead. Per contact. Per click. A flat monthly fee on top for some. You can't stack those numbers next to each other and compare them — they don't measure the same thing.

So put them all through one gate instead.

The one math, in your own arithmetic

Start with the instinct every tradesman already has. One countertop installer ran an Angi job through the same arithmetic he'd run on any bid — his hourly rate, his truck's fuel, the gross the job would actually bring — and the numbers didn't clear. He put it as a question about whether the guys signing up ever do the math at all. That worked example, run all the way to the dollar, lives in the Angi and HomeAdvisor chapter.

That's the math. Take it one step further than the sticker and it judges anything.

Cost per closed job = what you paid the machine over a stretch of time ÷ the jobs you actually closed from it in that same stretch.

Not per lead. Per job you finished and got paid for. The gap between those two is where the money hides, and here's why the gap is big: one lead is not one job. A shared lead goes to more than one pro. It can be a price-shopper, a wrong-town call, a number that never picks up. So it takes a stack of leads to make one booked job — and only you can count how tall that stack is.

Run it in order. Reading this is the quick part. Step 2 is an evening at the kitchen table, and that's the work, not you being slow.

1. What one lead costs. The sticker — per lead, per contact, or per click. If you're not paying yet, per-trade ballparks live on the price map.
2. How many leads it took to close one job. This is the number you count yourself, off your own phone: last month, how many of those leads turned into paid work? Five leads for one job means five times the sticker before you've turned a wrench.
3. Cost per closed job. Multiply. That's the real price of the job the machine got you — the number the sticker never shows.
4. What's left. Subtract that from the job's gross, then subtract your loaded cost to actually do it — the truck, the fuel, the hours. What remains is what the machine truly made you. On a small-ticket job, after a lead stack and a day of loaded labor, "what remains" can be a number you don't like.

Open the call log and step 2 may stop you cold — nobody ever wrote down where a caller came from. That's not you being sloppy; the tracking was never set up, so there's nothing to look up. Two ways at it. Backward: pull the platform's own lead list for last month and check each name against your paid invoices. Forward, if backward won't work: start tonight, one line per call, "where did you find us." Your first real verdict is thirty days out — so hold the machine at its current spend until then, not a bigger one.

And notice the job itself matters. A small repair lead and a big replacement lead can cost about the same to buy — put your own two job sizes in, a $____ repair and a $____ replacement — but only one of them can carry that lead cost and still leave money on the table. The machine doesn't sort them for you. The math does.

The number no dashboard shows you

Step 2 — leads per closed job — is the number that decides everything. The dashboard glows with leads delivered, clicks served, impressions up. The one figure that would let you judge them isn't on the screen.

That's not you failing to find it. Two of these machines have a specific way of running that stack up — Angi and HomeAdvisor by the shared-lead model, one request sold to several pros at once, and Yelp by a billing meter you can't read from the outside. Each gets its own chapter:

• Angi and HomeAdvisor — the shared-lead math. The math judges every shared lead you're handed. Walk it at the Angi and HomeAdvisor chapter.
• Yelp — the meter you can't read. Billed by the click on a monthly budget, with filtered reviews left out of your star rating and review count. Walk it at the Yelp chapter.

There's a second number worth naming once. When you pay someone to manage ad money for you, divide the fee by the money it steers. Owners' rule of thumb: when the fee outweighs the spend it steers, the number has stopped being a price. It's mapped across every web price on what web work actually costs. Run it any time a quote wraps a management fee around a smaller ad budget.

What the math can't tell you

This is the part the sellers skip, so it goes here in plain type. The math tells you what a channel cost. It cannot promise you a closed job, and no machine can — anyone who guarantees you one is the tell you're already scanning for.

Two stories keep the math honest, one on each side.

One owner did everything right on a lead platform — textbook speed on every callback — and still closed nothing and took a loss, because speed can't rescue a lead that was never real. That story, with the receipt, is the honest-limit beat of the missed-call text-back chapter.

And the other side, because a fair page carries both. One story that gets passed around is an owner who came back from a week off to twenty leads sitting in the inbox. Here it is with the wrapper off: that's a peer's post in a contractor forum, it's his own count of his own year and not audited books, and in the same thread the spend behind it is search work on his own site — not a lead machine at all. So take it for what it is — money spent online is not automatically money burned — and take it no further than that. Judge his win the way you judge the other man's loss: closed jobs and dollars. Your count on your own jobs beats anybody's story, ours included.

And sometimes the honest answer the math gives is none of them. If you've been booked weeks out for years on word of mouth, and no customer has ever once mentioned finding you online, the math may tell you to spend nothing here at all. That's a real answer, not a behind-the-curve one. If that's you, your question isn't which machine — it's whether you have a phone problem instead.

Run your own number

You don't need to trust any of this. You need to run it once, on whatever you're paying or being pitched right now. It ends in a number and a verdict you write down.

Run your own number, then circle the verdict

Machine: . It charges me $ per (lead / contact / click / month). Last month it brought leads, and of them became paid jobs. Cost per closed job: $ ÷ = $. This kind of job grosses about $; after my loaded cost — truck, fuel, hours — of $, what's left is $. Verdict (circle one): worth it · not yet · no.

Here's the honest shape of the ask, before you make it. We went looking and didn't find that number published anywhere — leads per closed job at a shop your size. A seller may genuinely stop counting at the handoff, so one who doesn't have it isn't necessarily hiding it. Don't score him on whether he has it. Score him on what he does when you ask. Here's the one sentence that asks, on the phone or in an email:

"Before we talk price — tell me, out of the leads you send, how many turn into a booked job for a shop like mine, and how do I check that number myself?"

Copy these words⧉

A straight answer sounds like "we don't track that past the handoff — here's your raw lead export, count the closes yourself." A fog answer is leads delivered, clicks served, or a promise your phone will ring. Ask, then go quiet and let him fill the silence. Ask twice if you have to. If the second answer is still fog, write "wouldn't say" in the verdict box and get off the call. Nothing has to be decided while he's on the line.

Print the scorecard

Print the one-math scorecard

The one mathone page

The rule: judge every lead machine by one number — what a real, closed job cost you, figured the same way for all of them. The formula: cost per closed job = what you paid the machine ÷ jobs you actually closed from it. Then subtract that, plus your loaded cost, from the job's gross. The four billing shapes: Angi/HomeAdvisor = per lead · Thumbtack = per contact · Local Services Ads = per lead · Google/Yelp Ads = per click. Four different measures — they don't stack. The missing number: leads per closed job. Count it yourself off your own phone. The second number: a management fee bigger than the ad spend it steers has stopped being a price. My number: Machine ____ · Cost per closed job $____ · What's left after loaded cost $____ · Verdict ____ · Sign-off ____

Whatever the machine, whatever the sticker, it comes down to the one number: what a real, closed job cost you, figured the same way for all of them. Run that, and no seller's dashboard can tell you a story your own count won't check.