Why do ads look like they are not working when they are?
Because in most accounts the numbers describing the ads are wrong before anybody looks at them. When a client tells us the ads are not performing, the first thing we check is not the creative and not the targeting. It is whether the measurement is telling the truth. In the majority of audits we run, it is not.
This matters more than any optimisation, because every optimisation is a comparison. If the two numbers being compared are both wrong, you are not optimising. You are choosing at random and feeling productive about it.
The five ways tracking usually breaks
These are the faults we find most often, roughly in order of how frequently they turn up. None of them announce themselves in the dashboard, which is why an account can run for a year on numbers nobody has ever verified.
- A duplicate pixel. The tag ends up on the page twice, usually because it was added once in the theme and once through a tag manager, and every conversion is counted twice.
- Conversions API sending events the pixel already sent, with no deduplication key, which produces the same double count by a different route.
- Attribution windows nobody chose. The default is a decision, and if you have never opened that setting then the platform picked your reporting model for you.
- Offline conversions never imported. If deals close on the phone, the channel that actually closes them looks like the channel that wastes money.
- Untagged links. Any traffic arriving without campaign parameters lands in direct or referral, so the best performing campaign quietly credits itself to nothing.
How to check your own account this week
You do not need an agency to do the first pass. Three checks will tell you whether your reporting is trustworthy, and they take an afternoon.
Start with volume: pick one week, count the conversions the platform claims, and count the orders or enquiries that actually arrived in your order system or inbox. If those two numbers are more than a little apart, stop optimising until you know why. Then check the tag fires once, not twice, using the browser tools your platform provides. Then check that every link you publish carries campaign parameters, which the UTM builder will generate consistently so two people tagging the same campaign do not produce two different labels.
- Reconcile one week of platform conversions against your own records.
- Confirm the pixel and the server side events are deduplicated, not additive.
- Write down your attribution window, deliberately, and keep it fixed while you test.
- Tag every outbound link with the UTM builder so channel reporting stops guessing.
Getting the unit economics right before you scale
Once the events are trustworthy, the next failure is scaling on the wrong threshold. Plenty of accounts are paused at a return on ad spend that was actually profitable, and plenty more are scaled at one that was not, because nobody worked out the break even point against real margin.
That calculation is arithmetic, not opinion. The ROAS and break even calculator gives you the number your campaigns have to clear before they make money rather than revenue. The CPC and CPM calculator and the click through rate benchmark tell you whether a rising cost per acquisition is an auction problem or a creative problem, which are two very different fixes. If you are setting a budget from scratch, the budget planner will at least stop you funding a test too thinly to produce a readable result.
Only now is it fair to judge the creative
With events verified and the break even point known, creative testing finally means something. Before that point, a creative test is a coin flip you have paid for.
What changes is the standard of evidence. You stop asking which ad has the better cost per acquisition this week and start asking whether the difference is large enough and stable enough to act on. Most creative differences that look decisive in a dashboard on a Tuesday are noise, and the accounts that improve fastest are the ones that stopped reacting to noise.
What skipping this actually costs
Two things, and the second is worse. The first is wasted spend, which is visible and recoverable. The second is a year of decisions built on the wrong numbers: creative killed for underperforming when it did not, budget moved to a channel that was double counting, a supplier blamed or replaced on the strength of a report nobody validated.
That second cost does not appear in any dashboard, which is precisely why it accumulates. Fixing measurement is unglamorous and it is the highest return work available in most accounts. It is where our performance marketing work starts, before any new creative is made.
Common questions
How do I know if my conversion tracking is double counting?
Compare one week of platform reported conversions against the orders or enquiries that actually reached your own system. A platform figure close to exactly double your real figure is the classic signature of a duplicate pixel or of server side events firing alongside browser events without a shared deduplication key.
Should I use the Conversions API as well as the pixel?
Yes, but only with deduplication configured, so the same event sent by both routes is counted once. Sending both without a shared event ID is the most common way a well intentioned tracking upgrade makes reporting worse than it was before.
What attribution window should I use?
The one you choose deliberately and then leave alone while you are testing. The specific window matters less than keeping it fixed, because changing it mid test changes every number you are comparing and makes the test unreadable.
How long should I run a creative test before deciding?
Long enough for the difference to be larger than the week to week noise in the account, which for most small budgets means weeks rather than days. If you cannot state what result would make you keep the losing ad, the test is not ready to be judged.