The measurement gap nobody is talking about openly
For over two decades, digital marketing ran on third-party cookie-based attribution. The model seemed almost magical in its precision: you could trace a customer's journey from the first display impression to the retargeted social ad to the organic search click that finally converted, across devices, sessions, and weeks. Marketing investments could be justified, channels could be compared, and the CFO could be shown a clear picture of returns.
That model is breaking down. It is not breaking down all at once in a single dramatic event. It is eroding gradually, through browser privacy changes, iOS tracking restrictions, regulatory requirements, and the steady decline of the third-party cookie as a reliable tracking mechanism. By the time most brands fully recognise the extent of the gap, they have already been making decisions on unreliable data for months or years.
What is actually breaking
Last-click attribution
Last-click was always a flawed model; it gave all credit to the final touchpoint before conversion and ignored everything that came before. But it was consistent enough to be useful for channel comparisons, even if the absolute numbers were wrong. As cross-session tracking degrades, even the relative comparisons are becoming unreliable. Channels that appear at the beginning of the customer journey, display, social awareness campaigns, content, are being systematically undervalued by any model that depends on tracking individual users across time.
Retargeting pools
As users opt out of tracking at increasing rates, retargeting audiences are shrinking. The users who remain in retargeting pools are increasingly the ones who have not opted out, which introduces a systematic bias toward certain demographics and behaviours. CPMs for retargeting are rising as the available pools shrink. The return on retargeting investment is declining in most markets.
Paid social reporting
Meta's advertising platform in particular has seen significant reporting gaps since the iOS 14 changes in 2021. The modelled conversions that Meta reports to fill the gaps created by limited tracking are, by Meta's own admission, estimates. For many advertisers, reported ROAS is materially higher than actual ROAS once you account for the modelling uncertainty. Decisions made on the basis of these reported numbers without validation are likely to be suboptimal.
What is replacing it
First-party data
The most robust measurement asset any brand owns is its own first-party data, email subscriber behaviour, CRM engagement records, on-site logged-in user data, and direct customer surveys. These data sources do not depend on third-party tracking infrastructure and are not subject to the same erosion. Investing in first-party data collection and activation, building email lists, encouraging account creation, linking customer records across touchpoints, is the most durable response to the attribution crisis.
Marketing mix modelling
Statistical modelling that estimates channel contribution based on aggregate spend and outcome data, without requiring individual-level tracking. MMM uses regression techniques to identify the relationship between marketing investment across channels and business outcomes over time. It has been used by large consumer goods companies for decades and is now increasingly accessible to smaller brands through open-source tools like Google's Meridian.
MMM does not give you granular session-level attribution. But it gives you a statistically valid estimate of channel contribution at the aggregate level, which is often more decision-relevant than the session-level data that privacy changes are eroding anyway.
Incrementality testing
Running controlled experiments to measure the true incremental effect of specific marketing activities. This means pausing a channel in one geographic region while maintaining it in another, or testing a message variation with a randomly selected subset of your audience. Incrementality testing is the gold standard for measuring whether a marketing activity is actually causing business outcomes or merely correlating with them, a distinction that last-click attribution was never capable of making.
The brands that invest in measurement infrastructure now will have a significant competitive advantage as tracking continues to degrade industry-wide.
What to do this quarter
Start with a first-party data audit. What customer data do you already own, and how well are you using it? Then review your attribution model and be honest about where it is likely to be overstating or understating channel performance. Finally, design one simple incrementality test, something you can run in thirty days, that will give you better information about the true contribution of one of your major channels.
Imperfect measurement, better decisions
The companies waiting for attribution to become perfect again will be waiting a long time. The shift is structural and largely irreversible. The question is not how to restore the old model; it is how to build a new one that fits the reality of privacy-first digital marketing. The brands that invest in that capability now will have a genuine competitive advantage over the ones that do not. Not in two years, sooner than that, as tracking continues to degrade and the measurement gap between prepared and unprepared brands widens.