Reconciling ad platform numbers: why they never match and what to trust

Every shop eventually puts three numbers side by side and finds that the platforms between them claim more conversions than the business had orders. This is not fraud and it is not a broken integration. It is what happens when three systems answer slightly different questions and each reports its own answer confidently.

Once you can name the five differences, the numbers stop being alarming and start being useful.

The five reasons

1. View-through. Platforms count conversions from people who saw an ad without clicking. Your analytics counts nothing of the sort. This is usually the single largest component of the gap.
2. Different windows. A platform may use a long click window and a shorter view window; yours may be seven days. The same order falls inside one and outside another.
3. Last touch versus self-attribution. Each platform attributes to itself whenever it can. Meta and Google will both claim the same order, so their totals overlap and must never be added.
4. Modelled conversions. Where consent or tracking prevents observation, platforms estimate. Those estimates are in the reported figure and are not separated out.
5. Different definitions of a conversion. A purchase event fired on a thank-you page is not the same as a completed order in your database — the page can be reloaded, and the order can later be cancelled or refunded.

Which number to trust for what

• Your shop's orders — the truth about revenue. Complete regardless of consent, because it is not tracking.
• Platform numbers — useful for comparing campaigns within that platform, and for its own bidding. That is genuinely what they are for.
• Your analytics — useful for behaviour and funnels, incomplete by consent, and never the source for a revenue figure.

Plan budget with the first. Optimise campaigns with the second. Diagnose with the third.

Track the ratio, not the gap

Chasing the gap to zero is a waste of a quarter. What is worth having is the ratio between each platform's claim and your own attributed number, per channel, tracked over time.

1. A stable ratio is fine, whatever its value. It means both systems are behaving consistently and you can mentally discount.
2. A ratio that moves suddenly is the signal. Something changed — tracking, consent rate, a checkout change, the platform's modelling.
3. Investigate movement, not magnitude. The absolute difference is structural; the change is an event.

A monthly reconciliation worth doing

Fifteen minutes, once a month, with four figures:

• Total orders and revenue from the shop, net of refunds.
• Attributed revenue per channel from your own data, on one window.
• Each platform's claimed conversions and revenue.
• The unattributed remainder — orders your own attribution could not assign to any channel.

That last figure is the interesting one and it is usually large. It is direct traffic, word of mouth, and everything consent rules made invisible. A shop that treats it as a failure of measurement will over-invest in whatever is measurable; a shop that treats it as a real and permanent share will make better decisions.

The only way to settle an argument

When two channels both claim the same revenue and you have to decide which to fund, attribution cannot answer it. Only a holdout can.

Turn one channel off in a region, or for a random slice of the audience, for long enough to read a difference. It is uncomfortable, it costs some revenue, and it produces the one number that is not an accounting convention. Shops that do this once a year make materially better budget decisions than shops that argue about dashboards.