The short answer
Count leads at the point they arrive, not in a dashboard, because one caller often registers as a Google Ads conversion, a GA4 event and a form email at the same time. Then read the number over ninety days against booked jobs. If you still cannot tell, turn the channel off for three weeks and watch what happens.
How one caller becomes three conversions
Somebody clicks your ad, fills in the form, does not hear back fast enough, comes back the next day through a search for your name, and calls the number on the site. That person is one lead. Google Ads counts a form conversion. GA4 counts two key events. Your call tracking counts a call. Your inbox has an email. The monthly report says five.
Three settings drive most of the inflation. Conversion counting set to every instead of one, which is correct for ecommerce and wrong for lead generation. A conversion action firing on a thank-you page that people reload or bookmark. And overlapping definitions, where a call, a form and a click-to-email all count as separate wins for the same person.
Fixing this usually cuts the reported number by a third or more. Owners hate that meeting and then run a better business afterward, because every decision that follows is made on a number that means something.
Count leads where they land
The only count that has never lied to us is the one kept by the person who answers the phone. One line per inquiry: date, name, how they came in, what they wanted, whether it turned into work. A spreadsheet is fine. Most trades in Monmouth County could run their entire marketing decision-making off that sheet and a bank statement.
Support it with a separate tracked phone number per source, so the dashboard and the sheet can be reconciled rather than argued about. What you should not rely on is asking callers how they heard about you. People say they Googled you when they clicked an ad, and they say a friend recommended you when the friend recommended you and then they Googled you.
What number should you actually watch?
Cost per booked job, by source. Everything else is diagnostic. Impressions, clicks, click-through rate, engagement, followers and rankings are all useful for telling you why the real number moved, and none of them are the real number.
To get there you need three inputs you probably already have: what you spent, how many jobs closed from that source, and what those jobs were worth. Divide spend by booked jobs. Compare that to your average job value. If a booked job from a channel costs you less than what the job earns you, and there is room to buy more of them, that is a working channel.
Keep one supporting number next to it: the share of leads that turn into scheduled work. When that figure falls while lead volume holds steady, the marketing is usually fine and something after the lead has broken. Nine times out of ten it is response time. A lead called back within ten minutes and a lead called back the next afternoon are not the same lead, and no ad account adjustment will close that gap.
Read it over ninety days, not a month
A month of a small account is mostly weather. Twelve leads instead of eighteen can be a holiday week, a competitor's sale, one broken form, or nothing at all. Judging a channel on thirty days is how good campaigns get killed and lucky ones get expanded.
Compare ninety days to the same ninety days last year rather than to last quarter, especially in a seasonal trade. And write down what you changed and when. Most accounts we inherit have no record of changes, which means nobody can say whether June was better because of the new landing page or because it rained.
The test that settles the argument
When attribution will not resolve it, run a holdout. Turn the channel off for two or three weeks in a period you can afford, keep everything else constant, and count the leads that still arrive. If the phone rings the same, you have your answer. If it drops by a third, you also have your answer, and it is more trustworthy than any attribution model.
Do it in a shoulder season, not in your peak, and do not test two things at once. This is blunt and slightly uncomfortable, which is why almost nobody does it, and why the people who do stop arguing about which platform deserves credit.
What you will never know
Attribution below a certain volume does not resolve cleanly, and no tool fixes that. Somebody sees your truck for two years, reads a review, searches your name, and clicks an ad. Every system will credit the ad. The truck and the review did the work.
So treat the numbers as directional and hold them to a simple standard: is the cost per booked job stable or improving, and is total booked revenue growing. If both are true, the details of which channel gets credit matter less than the argument about them suggests. If neither is true, no amount of dashboard tuning will change it, and the problem is in the offer, the follow-up speed, or the pages people land on rather than in the measurement.
Follow-up questions
Take one week and match every reported conversion to a real person in your call log or inbox. If the report shows twenty and you can name twelve, you know your multiplier. Most small accounts we check land somewhere between 1.3 and 2 times the real count.
Paid search gives a readable signal in about four to six weeks once tracking is clean. SEO and organic content take a couple of seasons. Judge anything on thirty days and you are mostly measuring the calendar.
If phone calls are how you get hired, yes. It is inexpensive and it turns arguments into arithmetic. Use one number per source and keep your real number on the site as well, so nothing looks unfamiliar to a repeat customer.
Let’s build something.
Twenty minutes on the phone and you’ll know whether we’re the right fit. Either way you’ll leave with something useful.