GA4 vs Programmatic Attribution: Why the Two Numbers Never Match

Two systems report on the same campaign and return different numbers. One is Google Analytics. The other is the DSP. That gap is where most advertiser-and-partner arguments begin, and very few of them get settled — because the argument is usually about which number is honest, when it should be about what each system can physically see.

What each system is actually counting

Google Analytics is a site analytics tool. It records what happens on your property and attributes it to the last click that brought someone there. That is a scope decision built into the product, not a defect.

A DSP counts something else: impressions served, to which devices, in which markets, and which of those devices converted afterwards — whether or not anyone clicked. It sees the exposure. What it cannot see is anything on your site beyond the events you choose to send it.

Neither system is lying. They are counting two different populations, and the disagreement between them is structural rather than accidental.

Why GA4 shows you the smaller number

A person served an impression on connected TV on Sunday, who searches your brand on Wednesday and deposits on Thursday, appears in Google Analytics as organic or direct. The exposure that started the sequence never touched your site, so it is not in the data at all.

GA4's move to an event-based model improved cross-channel reporting and changed the default credit logic. It did not move that boundary. Post-view exposure still sits outside what a site analytics tool can observe, because observing it would require a presence where the tool has none.

What the gap looks like when someone measures it

Across six independent measurements — five operators, two continents, six years — the post-view share of conversions landed between 89% and 96.2%. Every one of those was customer-ID-level or platform-verified. None was modelled. An operator reading clicks alone is seeing roughly one conversion in ten.

That is a statement about measurement completeness and nothing more. It says nothing about what would have happened with no media at all — a different question, needing a different test. The full working, including the impression-to-click ratio underneath it, is set out in click and view attribution in programmatic advertising.

Data-driven attribution does not close it

Data-driven attribution distributes credit across the touchpoints a model can see. That is a better distribution of a bounded dataset, not a wider dataset. If the exposure never reached your property, no weighting scheme inside your analytics tool will surface it.

There is a second thing worth naming plainly. A platform grading its own contribution has an interest in the grade. That is true of Google, and it is equally true of any DSP, including the one we trade on. Which is why the number that should govern your decisions is the one reconciled against your own backend — not the one in either dashboard.

What to do, and when to do it

The fix is procedural and it is cheap, but only if it happens before the money moves. Attribution education delivered inside a performance report — after the spend, in the appendix — does not change behaviour.

• Agree which number governs, platform-attributed or backend-confirmed, and write it down before launch.

• Fix the attribution window per objective, in the contract, and make changing it require your signature.

• Reconcile monthly, in writing: event definitions, time zones, currencies, deduplication, naming and UTM logic.

• Label anything unreconciled as platform-attributed, every time, until the backend is connected.

• Ask for the raw path-to-conversion export. A partner who cannot produce full journey data is not analysing it either.

What it costs when nobody does this is documented rather than argued: The 2% Problem follows six sportsbook operators whose accounts all ended on a measurement disagreement rather than on performance.

Testing a partner on this

If you would rather test a desk than take any of it on trust, the 20-question partner scorecard runs ThumbAd's published due-diligence standard against any partner — including us — and scores it against a bar of 24 out of 30. It runs in your browser, asks for no email, and nothing leaves the page. Our own answers to all twenty are published in full, ungated, in How ThumbAd Answers the 20 Questions.

ThumbAd is an independent programmatic trading desk in Oslo, trading on The Trade Desk across Europe, Africa, North America and Latin America since 2012, with betting and gaming at the core.