Conversion tracking audit

Measurement pipelines fail toward undercounting, and knowing that changes where you look first. A dead component emits nothing, so breakage silently removes conversions. Overcounting takes an active defect — a duplicated tag, a replayed event, bot traffic — so it needs something to be doing extra work rather than nothing at all.

The practical consequence: conversions collapsing while spend, clicks and sessions hold steady is tracking loss until proven otherwise, not a demand collapse. Treating it as a demand collapse is how a working campaign gets cut.

Which events fire, how many times, and whether each one corresponds to one real outcome. What each platform is counting under the word Conversions, and whether the definition has changed inside the comparison window you are looking at. How much is modelled rather than observed, which on iOS traffic after ATT is a real share rather than a rounding error. Whether the platform's number and the business's own record can be reconciled, and by how much they differ.

The output is a number and a direction, not a verdict of fine or broken. Fine is a claim about every dimension and is almost never supportable; a specific, witnessed fault is established by one observation.

Conversions arrive days or weeks after the click, so any recent window is undercounted at the moment you look at it and backfills afterwards. Read a fresh seven-day window as final and you will diagnose a collapse that is really a lag.

The same applies in reverse to long windows. A thirty-day lookback both hides real waste and flags spend that was about to work. There is no window that is correct in general — the right one depends on how fast the account's own conversions arrive, which is a thing to measure rather than assume.

A fix list in priority order, with the size of the exposure attached to each item, so it is possible to decide what is worth doing rather than doing all of it. Where the fix is in your own site or CRM rather than the ad account, it is written up for whoever owns that, in enough detail to hand over.

For clients on either plan this is where the engagement begins, because there is no point optimising against numbers nobody has checked. It is also available on its own.