Advertising Waste Audit · System 01 · the first check before any spend decision
Check the Numbers Before You Change the Ads
Every ad account reports its own success. Before you cut, move, or add a dollar, find out how many of the results it’s counting are customers you can find in your own records. That one comparison tells you whether anything else in the account can be trusted.
The idea in one line: an ad account’s results are its claim to credit, made under its own rules, by the company you’re paying. Your business runs on a different number: customers it can find in its own records. Measure the distance between the two first, because every decision made inside the ad account carries it.
What the ad account is counting
A conversion is whatever the ad account counts as a result: a form filled in, a call, a purchase. Each platform has its own rule for which ad gets the credit. Three reports, three different rules. At the time of writing, the standard settings are:
Each one is asking the same thing: did this person come across one of our ads, within our time limit, before they acted? None of them asks whether the customer would have come anyway, and none of them knows what the others counted. Someone who clicks a Facebook ad on Tuesday and a Google ad on Thursday, then buys, shows up as a result in both reports. Add the two together and you’ve counted that customer twice.
Each report is consistent by its own rules. They can’t be added to each other or matched to your books, and that’s by design: each platform is sharing out its own credit, and none of them is trying to be your business’s record of what happened.
Why the count runs high
Most of these are ordinary side effects of how the platforms count, and nobody did anything wrong. They matter because the count gets read as if it were a list of customers.
Two platforms, one customer
- Google and Facebook each saw the person within their own time limit, and each claims the sale. One customer, two results, and both totals get added together in a spreadsheet somewhere.
Counted twice by the website
- Two pieces of tracking code recording the same action, or a thank-you page that counts again when it reloads. One action, two or three results.
Small actions counted as leads
- Starting a form, clicking a button, tapping a phone number, scrolling, downloading a brochure. Useful signs of interest, and none of them is a person you can call. If one of them counts as a main result, the numbers look better than they are and the automatic bidding chases the wrong thing.
People who only saw the ad
- Facebook’s standard setting gives an ad credit when someone saw it yesterday and never clicked. Some of that is real influence. All of it counts the same.
Leads that aren’t people
- Bots, fake email addresses, the same person twice, people who never answer the phone. The form worked and the result was recorded. There’s still no customer.
Better tracking mistaken for more demand
- Improved tracking links more of the sales you were already making to the ads people saw first. The reported numbers go up. Nothing extra was sold.
Ads in places you didn’t choose
- Settings that show your search ads on other websites, switched on by default. Results from there look the same in the totals and often behave nothing like the rest of the account.
How to match the two sets of numbers
Pick one month: recent enough that you can still identify the customers, and far enough back that slow deals have closed. Then get two things and match them.
Notice which step the numbers drop at, because one of them has nothing to do with advertising. A big drop between found in your records and a person you could reach means inquiries are coming in and nobody is getting back to them, which no change to the ads will fix. And “a good lead” needs a written, dated definition, owned by whoever lives with the consequences of getting it wrong, before anyone argues about lead quality.
Open ChatGPT, Claude, or whichever you use, and upload both files: the ad account’s month and the same month from your own records, as a spreadsheet. Then send these three messages one after another in the same chat. The matching is the slow part, and it’s what an assistant does well, as long as you hold it to the evidence.
- “Here are two files for the same month: what the ad account reported, and what our own records show. Match them as far as the data allows, and tell me which matches are certain, which are likely, and which you can’t make.”
- “For every result you can’t find in our records, tell me which of the causes above most likely explains it, and what evidence would confirm it.”
- “Don’t recommend any changes to the account yet. Tell me how far apart the two numbers are, where the gap is concentrated, and what I’d need to get to close it.”
What the gap tells you
You’ll end up with one of four pictures, and each sends you somewhere different. Our working rule for small and large: if you can find about four in five of the ad account’s results in your own records, the gap is small. Below that it’s large. A large gap is concentrated when one campaign, network, website, or set of ads accounts for more than half of the missing ones, and spread out when none does.
Small gap
- Four in five or more of the reported results are in your records, and most of those are people you could reach. The ad account’s number is good enough to run the ads day to day. Judge the spend on your own figures, and move on to System 02.
Large gap, spread out
- Fewer than four in five found, and the missing ones are spread across campaigns in roughly equal shares. That’s a counting problem: tracking code counting twice, a small action counted as a main result, or a tracking change on a date you can find. Fix what’s counted before changing any bids. The account is learning from the wrong numbers, and every change you make teaches it faster.
Large gap, concentrated
- Fewer than four in five found, and one campaign, network, website, or set of ads accounts for more than half of the difference. That’s a finding, and it goes to System 02 to be sorted and sized. In the example above, 44 of 140 are missing and 31 of them come from one set of ads, so it’s this picture.
Too little data to say
- With a small budget, the limit is how many results each piece of the account has: a campaign, a town, an hour of the day. A campaign with six results this month can’t tell you anything about hours, towns, or devices, and no report changes that. Below about thirty results in a piece, check one level up and stop there.
Whichever picture you get, the comparison also settles what the ad account’s number is for. Use it to run the ads day to day. Use your own records to run the business. Whether the advertising brings in customers who wouldn’t have come otherwise is a third question. That takes a switch-off test: stop the ads for one group of customers that’s similar to the rest and watch what your own records do. It’s System 03.
A worked example
A marketer took over an ad account that generated leads and ran the check above. The ad account was clear about its best performer: the setting that shows Google search ads on other companies’ websites was bringing in leads for about $188 each, against $575 on Google’s own search results. Three times better, by the account’s own sums, and the automatic bidding was steering money toward it.
Matched against the company’s own records, those ads had spent about $78,000 on 417 reported leads. Of those, 380 were junk or spam, eight were good leads, and one became a customer. On Google’s own search results, more than half the leads were good ones. The ads the account rated three times better were producing one good lead for every ten thousand dollars.
Nothing inside the ad account would have shown this. The ad account was the thing calling it a success. That’s the pattern to remember: the account’s count feeds every decision made inside the account, so when the count is wrong, working harder on the ads takes you faster in the wrong direction. The way out is a number the ad account didn’t produce.
Once you know how far the ad account’s count is from your own, you know which problem you have: a counting problem to fix first, or a gap to size. System 02 covers the sizing: four kinds of advertising waste, and how to tell which one you’re looking at.