You spent $2,000 on Google Ads last month. Here's what it actually booked.

Google Ads says your campaign booked 41 conversions last month. Your schedule says maybe 12. Both numbers are true. They’re just counting different things, and the gap between them is where most owners lose the thread on what their advertising actually did.

This is the question every “Google Ads for plumbers” guide walks you up to and then abandons. They end where the click ends. The click is not the money. The booked job is the money, and getting from one to the other is harder than any dashboard admits.

So let me say the uncomfortable part first. Attribution on field-service data is partly guesswork. Anyone selling you a precise number for what each ad dollar booked is overselling it, us included. What you can do is narrow the guess honestly, and act only on the part that’s solid.

Both numbers are true, and neither is the answer

The 41 is a conversion count. It fires when someone calls the tracked number or fills the form. It does not know whether that call became a job, whether the job got done, or whether the customer would have found you anyway.

The 12 is what your schedule can trace back to Google, and only if someone wrote down where the lead came from. If your CSR left the lead-source field blank on half the tickets, your real number isn’t 12. It’s “12 that we know of, and an unknown pile we didn’t log.” That’s not a rounding error. That’s the whole problem.

What each tool can actually see

Before you divide anything, know what your three systems can and can’t tell you.

• Google and Meta see clicks, calls, and form fills. They see the top of the funnel and nothing past it. A conversion is a hand raised, not a job won.
• Jobber or Housecall Pro see the booked job and, if someone filled the field, its lead source. This is the only place the click and the work meet, and it meets only as well as your intake habits allow.
• QuickBooks sees the cash. Not what a job was worth on paper, what actually landed in the account, and when. On a shop carrying real receivables, that “when” can be sixty days behind the work.

No single tool spans the chain. The ad platform can’t see the collected dollar; QuickBooks can’t see the campaign. The read only exists if you join them, and the join is where the uncertainty lives.

Working it on Northside

Here’s the math on Northside Comfort, our labeled sample shop, an HVAC and plumbing operation. The figures are the sample shop’s, not a real customer’s.

Northside spent $2,000 on Google Ads in June. Google reported 41 conversions. Cost per conversion: about $49. That’s the number the platform wants you to see, and it’s the one worth the least.

Now walk it down the chain. Of the jobs booked that month, 12 had a lead source pointing to Google. So the traceable cost per booked job is $2,000 ÷ 12, about $167. Already more than triple the platform’s figure, and we haven’t hit the hard part yet.

The hard part: that 12 is soft in both directions. Some of those 12 were referrals who Googled the name before clicking, and would have called anyway. Some untracked jobs, the ones with a blank lead-source field, may genuinely belong to Google. So the honest cost per booked job isn’t $167. It’s a range, somewhere around $140 to $250, and the width of that range is a direct measure of how loose your intake is.

Push it one more hop, to cash. QuickBooks shows those 12 jobs collected about $9,600 so far, which puts cost per collected dollar near $0.21. But “so far” is doing real work in that sentence: two of those invoices are still open inside Northside’s $11,700 in receivables. Collect them and the number improves; write one off and it gets worse. Even the denominator is still moving.

Three numbers, three tools, and the uncertainty compounds at every hop. That’s not a flaw in the method. That’s the honest shape of the thing.

Where attribution genuinely breaks

Two patterns break attribution no matter how clean your logging gets, and it’s worth labeling them for what they are.

The first is the referral who clicks your brand ad. A neighbor recommends you, the customer searches your company name to find the number, your own ad sits at the top, they click it, and Google charges that booked job to advertising. You’d have won it for free. That’s not a measurement error you can fix with better tracking. It’s the structure of last-click attribution.

The second is the phone-first customer who never touches a tracked path. They saw your truck, kept the magnet, called the shop line six weeks later. No click, no conversion, no attribution. Real job, invisible to the whole apparatus.

The distinction that matters: the 41 and the 12 are measurements, imperfect but countable. Which of those jobs the ad actually caused is inference, and inference is where confident software lies to you. We built more of the below-the-line reasoning, why the leak is often your phones and not your ads, into lowering cost per booked job. I won’t repeat it here.

The one move that’s actually safe

Owners want the takeaway to be “cut Meta, pour it into Google.” I’m not going to give you that, and I’d be suspicious of any tool that does. At the numbers above, the range on a single channel is wider than the difference between channels. Moving budget on that would be moving it on noise.

There is exactly one move the data supports right now, and it costs you nothing. Have your CSR log the lead source on every ticket for 30 days. One required field: “How’d you hear about us?” Ask it, write it, every job, no exceptions. Do that for a month and your traceable count stops being “12 that we know of.” The range narrows because the blank pile shrinks. Then, and only then, is there a budget conversation worth having, and it’s yours to have, not your software’s.

I’ve watched owners skip this step for years because it feels too small to matter. It’s the single most useful thing on this page.

The 15-minute monthly ritual

Once the field is filling in, the whole read takes fifteen minutes a month:

1. Google and Meta: pull spend per channel. Two numbers.
2. Jobber or Housecall Pro: count booked jobs by lead source. Divide spend by jobs. That’s your cost per booked job, stated as a range if the logging is still thin.
3. QuickBooks: check what those jobs have collected, and what’s still open. That’s your cost per collected dollar, with a note on what’s in flight.

Three tools, three numbers, one honest picture with its uncertainty attached. It won’t tell you the future. It will tell you when something moved enough to look closer.

Where this ends up

That blended read, spend joined to jobs joined to cash, each figure carrying its own caveat, is exactly what the Monday briefing does for you so you don’t run it by hand every month. Guidepost reads your Jobber or Housecall Pro and QuickBooks, shows the range instead of a falsely precise point, and names the one field that would tighten it. It won’t pretend last-click is precise, and it won’t tell you to move the budget on a noisy month. That call stays yours.

If you want to see the join on your own numbers instead of Northside’s, book a demo: 20 minutes, your real data, no migration.