The short answer
If you add up what each platform claims, you will get more conversions than you had. Each is reporting what it touched, by its own rules, and none of them can see the others.
The only way to know what a channel actually caused is to vary it and watch total enquiries, not platform-reported conversions.
Why the numbers exceed reality
- Each platform counts a conversion it was involved in
- View-through counting credits impressions nobody clicked
- Attribution windows differ, sometimes by weeks
- Cross-device journeys get counted twice
- Your own analytics uses different rules again
None of this is dishonest. Each platform is answering a narrower question than the one you are asking, which is what caused the enquiry.
What attribution models do
| Model | Credits | Bias |
|---|---|---|
| Last click | The final touch | Favours search and brand terms |
| First click | The initial touch | Favours awareness channels |
| Linear | All touches equally | Treats trivial and decisive alike |
| Time decay | Recent touches more | Favours the end of the funnel |
| Data driven | Modelled from your data | Needs volume, still a model |
All of these are allocation rules. None of them measures causation, and switching between them changes which channel looks best without changing what actually happened.
The test that settles it
Turn a channel off, or reduce it substantially, in a region or for a period, and watch total enquiries. If they fall, the channel was producing them. If they do not, it was claiming credit for demand that existed anyway.
- Pick a period long enough to cover your sales cycle.
- Change one thing, not several.
- Watch total enquiries from all sources, not the platform's number.
- Account for seasonality by comparing against a similar prior period.
- Accept that the answer may be uncomfortable.
Brand terms are the classic case
Ads on your own brand name frequently show excellent cost per conversion, because the people clicking were already looking for you specifically.
Whether those conversions are incremental is a separate question, and it is testable by pausing brand bidding and watching total enquiries. Some businesses find a real effect from competitors bidding on their name; others find they were paying for clicks they would have had free.