Your measurement infrastructure is already degraded

Three platform-level changes are converging to structurally degrade measurement fidelity for digital marketing. This is not a future risk; it is current reality, and the gap between organisations that have adapted and those that have not is widening with every quarter.

Google kept cookies, but the status quo did not hold

In April 2025, Google confirmed it will not deprecate third-party cookies in Chrome. Cookies remain enabled by default, with manual user controls in settings.

This was widely reported as a reprieve, but it is not one. Safari and Firefox already block third-party cookies. Client-side ad blockers add another source of missing events. The size of that gap varies by audience and implementation; a share of people using blockers is not the same as a share of blocked sessions. Cookie-based tracking has significant gaps regardless of Chrome’s position. Treating Google’s decision as a reason to delay first-party data investment is a diagnostic error (and IMO one of the more common ones I have seen in 2025).

GA4 is now a modelled measurement system

GA4 combines observed data with modelling where the property and feature qualify. Behavioural modelling requires an advanced Consent Mode implementation and sufficient traffic; it is not available in every property or report. Conversion modelling is a related but separate feature. Consent Mode communicates choices to Google tags. It does not obtain consent or make a configuration lawful by itself.

Consent configuration does not merely affect reporting accuracy. It changes what data is available and what can be modelled. Organisations need to know which parts of their reports are observed, estimated or absent. Otherwise they risk treating a partial view as the whole market.

For organisations with material gaps in browser event collection, I regard a well-governed server-side pathway as foundational. It can improve event delivery and control over data sharing. It cannot recreate identifiers that were never available, override a consent choice or guarantee correct attribution.

Apple’s privacy controls reduce attribution signals

ATT consent varies by app and market. Adjust’s Q2 2025 dataset reports 35% opt-in among users shown the prompt; that is not a census of all iOS users. Meta’s settings also distinguish click-through from view-through attribution: a seven-day click window can be combined with a one-day view window. The original version of this note reversed those labels.

Safari 26 further restricts known fingerprinting scripts, including their access to query parameters and referrer data. Link Tracking Protection removes selected tracking parameters in the contexts where it applies. These are browser and feature-specific protections, not a guarantee that every click ID disappears everywhere on iOS. Hide My Email is a separate feature. The direction still matters: identifiers that marketers once relied on are becoming less dependable.

What to do

The intervention sequence is clear.

Implement server-side tracking where it addresses a measured gap. Reconcile browser, server and business records, deduplicate events and carry consent choices through the pipeline. Moving an inaccurate or impermissible event to a server does not fix it.

Implement Consent Mode v2 correctly. This determines what data GA4 has to work with. Poor implementation means poor data, which means poor decisions.

Invest in first-party data collection. Direct customer relationships can provide permitted first-party signals. They do not restore every lost cross-site signal. Design around the data you can legitimately collect, and measure the remaining gaps.

Adopt measurement approaches that do not depend on user-level deterministic tracking. Media mix modelling, incrementality testing, and conversion modelling are no longer supplementary methods; they are primary.

Organisations that have not made this shift are already operating on incomplete data, and the gap compounds with every quarter of inaction.

Follow-up: waiting for legislation is the wrong play

Consolidated from my Field Note of 1 December 2025.

Many CMOs are still waiting for Tranche 2 of the Privacy Act reforms before taking action. The belief is that the right time to modernise data foundations is once legal obligations become clearer. That assumption is flawed. The performance impact is happening before the regulatory impact.

In practice, the warning signs include rising CPAs, widening gaps between reported and business-recorded conversions, declining match rates and inconsistent audience performance. None proves signal loss on its own. Creative, competition, demand and implementation can produce similar symptoms. Use the pattern to investigate.

The practical priorities are the same: strengthen first-party event collection, improve delivery where browser collection fails, and treat consent design as part of the measurement system. Make the choice clear and easy to exercise. More events are only useful when they are accurate and permitted.

These are foundational steps that protect your ability to acquire customers efficiently in a constrained signal environment. You do not need a legislative deadline to start.