Don't guess whether ads beat lead fees. Run both, and count the jobs.

You typed some version of the same question: is it cheaper to pay Angi or Thumbtack for leads, or to run my own Google ads? Nobody can tell you in general — not us, not a blog, not the guy selling either one. It depends on your town, your trade, your season, and how much work your crew can take on. But you can find out for your own shop, cheaply, in about a month. One number settles it: what one closed job actually cost you on each side, counted over the same weeks, by you.

Why no one can just tell you

The sellers won't run this test for you. They charge per lead, not per closed job — the fee lands whether the work does or not.

Two rules already tell you when a monthly fee is a scam on its face: the ratio rule and the absorption rule. Those live on the price map, with the going bands, in what web work actually costs. This page is a different tool. Not "is this fee a rip-off," but "which channel actually turned money into booked work."

The one number that decides it

Here's the yardstick, in one line: the money you spent, divided by the jobs you actually booked and did. That's cost per closed job. "Loaded" means you count your own time and fuel in it too — not just the fee or the ad bill — the same way you'd never quote a job off parts alone. The full version of this math, and why it beats every other number a seller shows you, is chapter 4.1.

It's easy to compare the wrong thing. The lead app shows cost per lead. The ad guy shows clicks. Both are cheap to rack up, and neither pays your mortgage. A lead is a maybe. A click is a maybe. A closed job is money. So this test scores closed jobs only — the phone ringing and turning into work is the whole point.

You already run tests like this

You do this in the field without thinking. You don't trust a valve because the box says it holds — you pressure-test the line before you close the wall. You don't take a rep's word that the new pump runs quieter — you run it and listen.

A lead channel is the same kind of claim. Angi says its leads close. Google says its clicks call. Don't argue with either one. Put them on the same job, under the same conditions, and count what came back.

Set it up so it's a fair test

A test only means something if both sides ran under the same conditions. Set it up like this.

• Same weeks — and both sides running before the clock starts. Run the fee platform and your own ads over the exact same stretch — 30 to 60 days. The fee platform bills as soon as a lead lands. Your own ads have a gate in front of them: Google says most ads are reviewed within one business day, and that some take longer. Local Services is heavier — license checks, plus insurance and background checks required for select users, covering the business, the owner, and the field worker roster. So set the ads side up first, and don't switch the fee platform on until you've seen your own ad live on a real search. Checkable result: one start date and one end date, written down, covering both. A month where only one side ran isn't a tie — it's a no-test. Reset the dates.
• Same service area, same crew. Don't let one side cover the whole county while the other covers your town. Checkable result: the same map and the same people answering the phone, both sides.
• A money cap on each side — and a hand on the switch. Pick a number you can lose without it hurting, say a few hundred dollars each. Then read what a cap actually is: on Google Ads it's an average, not a stop-switch. In Google's own words a campaign "might spend up to twice your average daily budget" on a given day, though the month lands at "no more than 30.4 times" it. The month has a ceiling. So set the cap where each platform lets you, check both balances once a week on the same day, and pause a side the day it hits your number. Checkable result: two caps written down before a dollar goes out, plus four weekly balance checks initialed.

It's time-boxed, capped, and checked weekly. You're buying an answer, not betting the year.

Know which call came from where

Here's the step that makes or breaks the test. You have to know which call came from which channel. That's called attribution — a big word for a simple thing: tagging each call so you can tell them apart. A call you can't trace is a number you can't use.

Two ways to do it. Pick one.

• Ask, every time. Whoever answers the phone asks one plain question and writes the answer on the job ticket.

  "Before I get your address — mind if I ask how you found us? Google, a neighbor, or one of those lead apps?"

  Copy these words⧉

• Or put a separate phone number on the ads. The lead platforms already report their own calls, so a second number on your ads keeps the two from blurring together. The how-to for each channel lives in its own chapter — Local Services Ads is 4.4, Google Ads is 4.5.

Test the tagging before you trust it: find your own ad on a real search and call the number on it. On Local Services that call lands in your account as a lead, so do it once, and mark it a test so nobody counts it later as a real job.

If it doesn't land on the ads side, stop — don't start the month. And it isn't your fix: hand it to whoever set up the ads or runs your phones. If nobody can make the tagging work, fall back to asking every caller and writing it on the job ticket. That costs nothing and works on a flip phone. Checkable result: your own test call showed up tagged to the right column, marked as a test.

Add up each side honestly

Now the money. Count it the way it actually left your account, not the way anyone hoped.

Fee side. On the pay-per-lead platforms you're charged for the lead — the contact — whether or not it ever becomes a job. Don't quietly cross off the duds. Thumbtack's terms start at "all sales on the Platform and Fees paid by You to Thumbtack are final and non-refundable," and the complaint Thumbtack itself names — "Often, many pros request refunds because they weren't hired" — is written into the policy as not qualifying. You can request one within 45 days, decided case by case. And a granted refund goes to your Thumbtack balance by default, not to your card; chapter 4.3 walks that. So two rules for this test: your fee-side total is every charge in the window, closed or not, and you only subtract a refund if it actually hit your card. Credit sitting in a platform balance is still money spent on this channel.

Ads side. On Google Ads you pay per click — you're charged the moment someone clicks, call or no call. Local Services Ads works differently: there you pay per lead, a call or message, not per click. Google can credit a lead, but that's Google's call and not yours, and an approved credit lands on your account balance — in most cases within 30 days, with the original charge still showing on the invoice. Chapter 4.4 has the walls around it. So score the month on what the account actually billed you. If a credit lands after you've done the arithmetic, redo that one line.

One line both sides owe. Count your own time. Break-even on the invoice can be a loss once you count the truck and the phone. Add a rough dollar figure for your own hours to each side's total.

Do the arithmetic

Now divide. For each side: total spent ÷ jobs that closed = cost per closed job.

Here's the shape, with made-up numbers to show the machine. Yours go in the blanks.

• Fee side: say it billed $600 over the month and turned into 3 closed jobs. $600 ÷ 3 = $200 a closed job. (Example — replace with your own.)
• Ads side: say it billed $900 and closed 3 jobs. $900 ÷ 3 = $300 a closed job. (Example — replace with your own.)

In that made-up month, the fee side won: $200 beats $300 per closed job. Your real numbers will land somewhere else — that's the whole point. You're reading your shop, not this example.

One note on timing. A channel can start slow and pay off later, and we won't put a number on how long — we don't have one. If the two sides land close, nothing's decided: run another 30 days and score it on the same sheet.

Read the winner — and what it can't tell you

The smaller cost-per-closed-job wins the test — for your shop, this season. Circle it, and write the date range and the job count next to it. A read built on a handful of jobs is thinner than one built on a season of them.

What the test can't tell you, said straight:

• It can't promise next season looks like this one. Demand moves.
• It can't tell you whether you should be running ads at all. That depends on your shop's size, season, and trade — and it has its own chapter, when ads make sense.
• It can't guarantee a return. What it gives you is what already happened with your own money. That's the only thing worth trusting.

Run it: the one-page scorecard

Print this, tape it by the phone, and fill it in as the month runs.

Print the scorecard

The split-test scorecardone month

Window and rules: Start ______ (the day BOTH sides were live) · End ______. Same weeks, same area, same crew — yes or no. Caps set before spending: Fee-side cap $______ · Ads-side cap $______. Weekly balance checks: __ __ __ __ Tagging works: I placed a test call, it showed up on the right side, and it's marked as a test — yes or no. Fee side: $______ actually billed ÷ ______ closed jobs = $______ per closed job. (Refunds only subtracted if they hit the card.) Ads side: $______ actually billed ÷ ______ closed jobs = $______ per closed job. (A credit that lands later: redo this line.) My own time: added about $______ for my hours to each side. Verdict: smaller number is the ______ side, at $______ per closed job. Job counts: fee ______ / ads ______. Owner sign-off: ______

That's the whole test. You ran both channels over the same weeks and read the one number that settles it: what a closed job actually cost you on each side. Keep the scorecard. Next season, run it again — the answer is allowed to change.

Where to go from here. The yardstick itself is the one math. The two fee platforms are Angi and HomeAdvisor and Thumbtack. Your own ad channels are Local Services Ads and Google Ads. Whether to run ads at all is chapter 4.7. The whole map is the lead-machines hub.