Privacy, consent and signal loss

Ad platforms know less about individual users than they did, and the trend is one-way. Understanding why matters, because most of the "measurement is broken" complaints in the industry are really this.

What changed

Third-party cookies, cross-site identifiers and unrestricted client-side tracking have been progressively limited by browsers, operating systems and regulation. The result is that a meaningful and market-dependent share of conversions is never directly observed.

Consent, and where it bites hardest

In the EEA, consent is a legal precondition for much measurement. Consent mode communicates a user's choices to Google's tags, which then adjust behaviour and model the unobserved portion. The practical effects:

• reported conversions include modelled conversions
• platform and backend numbers will not reconcile, and the gap varies by market
• markets with high refusal rates have structurally less signal, so automated bidding has less to work with there

Regulation is fragmenting the platforms

The Digital Markets Act imposes obligations on designated gatekeepers, reshaping search, app stores and ad data access in the EU. This is no longer abstract: August 2026 produced a concrete precedent when, under European Commission mandate, Google stopped applying site reputation abuse enforcement inside the EEA from 30 August — the first time its enforcement split by user location.

That was an organic-search policy, but the shape is what matters. Platform behaviour is becoming regional, and "how does it work" increasingly needs "where?" attached. Assume your home market's behaviour is not universal.

Data protection regimes are also spreading beyond Europe — Indonesia's new rules, signed July 2026 and applying from January 2027, carry fines of 2% of revenue with a 72-hour response window for access and erasure requests.

What actually helps

• First-party data. Consented email and phone, sent as enhanced conversions, recovers matches that would otherwise be lost.
• Offline conversion import. Bringing CRM outcomes back closes the loop that client-side tracking cannot.
• Server-side collection, where you have the engineering to do it properly.
• Holdout tests. When observation degrades, experiments become the reliable way to know whether spend works.

What trips people up

• Treating modelled conversions as observed ones in board reporting.
• Comparing markets with different consent rates as if the difference were performance.
• Assuming a US-observed platform behaviour applies in the EEA.

You have got this when

You can explain to a finance stakeholder why platform conversions exceed backend conversions, without either number being wrong.