For most of the history of mass media, buying advertising required a shared fiction: a single accepted number describing how many people saw something. Everyone knew the number was an estimate built on panels and models. It worked because everyone agreed to use it, which is what a currency is.
That agreement has broken down, and the industry's response has been to stop trying to restore it.
Fragmentation as the settled state
The proximate causes are familiar. Consumption moved across platforms that measure differently and have commercial reasons to prefer their own figures. Privacy changes removed the identifiers that cross-platform measurement depended on. Streaming, podcasts, retail media and creator-driven distribution each arrived with their own counting conventions and no incentive to harmonize.
The deeper cause is that the thing being measured stopped being one thing. An impression on a muted autoplaying video, a podcast host reading a code, a search result and a fifteen-second skippable pre-roll are not comparable events, and a single number describing all of them was always averaging across incommensurable experiences.
What sophisticated buyers do now is triangulate. They run incrementality tests to isolate causal effect on a subset of spend, use media mix modeling for aggregate allocation, watch platform-reported metrics as directional signals rather than truth, and increasingly accept panel and survey data for reach questions the deterministic sources can no longer answer. No single input is trusted. The portfolio is.
This has been quietly good for formats that never measured well. The niche podcast advertising market grew despite measurement that would have disqualified it under the old currency, because incrementality testing showed effects that impression counting could not see. The same logic favors newsletters, trade publications and any environment where a modest audience is unusually attentive.
It has been correspondingly hard on formats that benefited from generous counting. Inventory whose value rested on large reported impression volumes has faced buyers asking what happens when the spend stops, and a number of long-standing allocations have not survived the question.
The cost of the new regime is that it advantages scale. Running incrementality tests and maintaining mix models requires analytical capacity that large advertisers have and mid-sized ones do not, which pushes smaller buyers toward the platforms that grade their own homework. That is a real concentration effect, and it is rarely mentioned in the industry's discussion of measurement, which tends to treat the problem as technical.
Publishers have adapted by selling outcomes rather than inventory where they can, which favours those with a direct relationship to their audience. That is a meaningful advantage for trade publications and for the independent media operations built on subscription rather than reach, since both can describe who their readers are with a precision the open market lost.
The honest summary is that measurement got worse and decision-making got better. Buyers who spent decades optimizing against a precise number that measured the wrong thing are now working with imprecise numbers that measure closer to the right one, and most of them, asked privately, would not trade back.
