Fix Your Tracking Before You Spend Another Dollar on Ads

Quick Takeaways

1. Most arguments about advertising performance in small business are arguments about numbers that were never correct to begin with.
2. A pixel firing twice, a conversion event pointed at the wrong action, or a thank-you page anyone can land on will all quietly inflate your results.
3. Wrong numbers do more damage than missing numbers, because they hand you the confidence to make a decision you should never have made.
4. Broken events also corrupt the targeting, because the platform learns who to chase from the events you send it.
5. Six checks, run properly once and then reviewed weekly, will tell you whether the dashboard can be trusted.
6. Platform figures and analytics figures will never match to the dollar. They should agree on direction and sit within a sensible margin.
7. This is the least interesting work in marketing and the highest leverage half hour you will spend this quarter.

Everyone argues about the ads, almost nobody checks the instrument

Here is a conversation I have had more times than I can count. A business owner tells me Meta does not work for them, Google Ads was a waste of money, and the agency they used last year burned through a pile of budget for nothing. They have the screenshots, the cost per lead and a spreadsheet of monthly spend, and they are certain.

Then we look at the tracking, and the figure they have been making decisions from was never measuring what they thought it was measuring.

Nobody enjoys hearing that, because it means reopening six months of settled opinion. It is far more comfortable to blame the platform, the creative, the agency or the market, and occasionally one of those is genuinely the answer. You cannot know which while the instrument reading the result is faulty.

This has quietly got harder. Browser tracking has been squeezed by privacy changes and consent requirements, server-side setups arrived to fill the gap, and most accounts now run both at once. More of a buyer's journey happens somewhere you never see: a question asked in ChatGPT, an AI Overview at the top of Google, a mate's recommendation in a group chat. The measurement stack became more fragile at exactly the moment more businesses started leaning on it, so the dashboard reads confidently and precisely and can still be wrong.

Two in the morning, fixing a pixel

I have written before about being up at two in the morning fixing a pixel because Meta was matching the wrong people. It is the least impressive story I own. No strategy session, no clever insight, no big creative idea, just a match quality problem and events carrying data that did not line up with real customers.

I stayed up because I already knew how the next fortnight would go if I did not. The campaign would spend, the numbers would look ordinary, and every decision after that would be built on sand. We would change the creative, then the audience, then the budget, and learn nothing from any of it, because the feedback coming back was describing the wrong people.

The other one that stays with me is simpler and worse. We took over an account where the lead events had never been wired up at all, meaning nothing was misconfigured or double counting, it had simply never been connected. Every conversion metric read zero while real enquiries kept arriving in the inbox. The owner had spent a year believing their advertising did not work, and the honest position was that nobody had ever been in a position to judge it.

I am not naming either business, and I never will. Broken tracking is close to universal and almost never anyone's fault in particular. It happens when a site gets rebuilt, a plugin updates, a developer leaves, or three people touch the account over two years and each assumes someone else checked.

What wrong numbers actually cost you

Broken tracking sounds like a reporting annoyance. The damage actually runs in three directions.

The first is the decisions you make. A profitable campaign gets switched off because half its conversions were never recorded, while a mediocre one gets more budget because a double-counting setup made every sale look like two. Then the money drifts towards whichever channel happens to have the most generous attribution window, and everyone congratulates themselves on being data-led.

The second is the targeting itself, and this is the one owners rarely see coming. Meta and Google optimisation feeds on the events you send. Tell a platform that a conversion happened, and it goes and finds more people who look like the person who triggered it. Point that event at a page view, a newsletter signup, or a bot that fires on every session, and you have taught the system to hunt down browsers. It will do that job brilliantly. You will get cheap conversions that never turn into money, and the account will look healthy right up until you compare it with the bank.

The third cost is trust, which is the hardest to repair. Once an owner has been burned twice, the door closes and they will tell you ads do not work for their kind of business. Sometimes that is true. More often they have never once seen a clean test.

The pre-flight checklist

This is the bit you can hand straight to your agency, your web developer or whoever holds the logins. You do not have to be technical to follow it. Somebody does have to actually look, which is the step that keeps getting skipped.

1. Confirm the pixel fires once, and only once. Walk your site the way a customer would and watch what the platform records. The Meta Pixel Helper and Google Tag Assistant both do this free in your browser, and you want one base pixel per page load. Double installs are extremely common, because the pixel goes into the theme, then again through a plugin, then again through a tag manager somebody set up and forgot. Two installs can mean everything reports twice, and a campaign that looks like 4:1 might be doing 2:1.

2. Confirm the conversion event maps to the thing that makes money. Write down in one sentence what a valuable customer action looks like in your business: a booked job, a submitted quote request, a completed checkout, a phone call longer than thirty seconds. Then check what the platform has been told counts as a conversion. If your primary conversion is a page view, a button click or a newsletter signup, your cost per conversion is a vanity figure and the optimisation is being trained on the wrong behaviour.

3. Confirm server-side and browser events are not double counting. If you run the Conversions API alongside the browser pixel, which most accounts now do, both are sending the same events. They are meant to be deduplicated using a shared event ID. When that ID is missing or does not match, the platform counts both, and your results inflate by something close to double. Check the event match quality and the deduplication rate in Events Manager. This is the single most common way a mediocre campaign gets promoted to a winner.

4. Confirm the thank-you page cannot be reached without converting. If your conversion fires on a confirmation page, that page needs to be unreachable by any other route. Test it: type the URL straight into the address bar. If it loads, every bot, every internal staff visit and every curious click gets recorded as a sale. Also check that it is blocked from search indexing, because a thank-you page sitting in Google will collect traffic for years and quietly pad your numbers.

5. Confirm the platform and your analytics agree within a reasonable margin. Pull last month's conversions from the ad platform and from your analytics, then compare both with the real world: invoices, bookings in the calendar, enquiries in the inbox. They will never match exactly, and anyone who says they should is selling something. Attribution windows differ, view-through conversions are counted in one place and ignored in another, and consent settings block some sessions entirely. Ten to twenty percent variance is normal. When the platform claims forty conversions and the inbox holds nine, you have found your problem, and the inbox is the one telling the truth.

6. Confirm somebody looks at it every week. Tracking breaks silently. A theme update, a new form plugin, a site migration, a CRM swap, a cookie banner change: any one of those can take an event offline without a single error message appearing anywhere. Put fifteen minutes in the calendar every week with one job attached. Does the number of conversions in the platform still resemble the number of real enquiries you received? That question, asked weekly by someone who knows the business, catches almost everything before it costs you a month.

The bit nobody thanks you for

Nobody has ever sent me a message about a correctly configured conversion event. There is no applause for deduplication. It does not photograph well, it makes a terrible case study, and when it is done properly the only visible outcome is that everything else starts behaving sensibly.

That is exactly why it gets skipped. Creative is enjoyable to talk about and audiences are enjoyable to argue about, while checking whether your event fires twice is a job for a Tuesday morning with a coffee. It will still do more for your return this quarter than a new set of ad images, because it compounds: clean events feed accurate signals, accurate signals build better audiences, and better audiences lower what a customer costs you. Warner Pool Care up in Brendale runs at 14:1 on thirty dollars a day, which is not a heroic budget. It works because everything sitting underneath it is telling the truth.

We build the same discipline into the Billy Bots we put on client sites. A bot that answers questions and books jobs at eleven at night is doing genuine work, and if those bookings never register as conversions, the bot looks like an expensive ornament and the ads that sent those people look like a failure. Measurement follows the money or it is decoration.

Where this sits in the order of work

If you are about to start advertising, this comes before the budget conversation. If you are already advertising and unhappy, it comes before you change anything else, because a change you cannot measure is a guess with an invoice attached. And if you are about to sack an agency, run the six checks first. You may find they were doing better work than the dashboard gave them credit for, or you may find the opposite and have the evidence in hand.

We run this as part of every audit at Infinity1, before anybody talks about spending a dollar, because judging a campaign through a broken instrument wastes both our time. If you would like a second set of eyes on what your account is really recording, let's talk.

FAQ

How do I know if my tracking is broken without hiring anyone? Compare last month's platform conversions with the real enquiries you can count: emails, calls, bookings, invoices. If those two numbers are in different postcodes, something is wrong. The browser extensions from Meta and Google will then show you which events are firing on each page, free, in about ten minutes.

Is a small gap between the ad platform and my analytics a problem? No. They measure differently, with different attribution windows and different rules about who gets counted. Ten to twenty percent variance is ordinary. Consistent direction matters more than matching totals. Investigate when the gap is large, when it suddenly changes, or when the platform disagrees with your actual sales.

Can broken tracking hurt me even when I am happy with the leads coming in? Yes, because the platform optimises towards whatever you label a conversion. Faulty events teach it to find the wrong people, so your results quietly drift towards low-quality volume while the reported cost per conversion looks better every month.

Does any of this change with AI search sending traffic? The need gets sharper. More people arrive already informed, having asked an AI assistant first, so the visible click is the tail end of a longer journey. Clean conversion data is how you tell a channel that has stopped working from one whose contribution you have stopped recording.

The short version

Your advertising might be working, and it might be terrible. You cannot know which while the pixel fires twice, the conversion event points at a page view and the thank-you page is open to anyone who types the URL. Run the six checks, then judge the campaign. Judging first and checking later is how businesses talk themselves out of a channel that was fine all along.

If you want us to look at what your account is really recording before you spend another dollar, book a time and let's talk.