The end of the free data lunch

For twenty years, adult sites ran on analytics stacks that would have been illegal in any other vertical. Third-party trackers everywhere, consent assumed, retention indefinite. Nobody audited it, because the regulators had other things to look at and the industry had other things to hide.

Then the regulators ran out of other things, and the adult web discovered it was holding the most expensive liability on the internet in the form of twenty years of unconsented behavioural data.

What the banners actually cost

The obvious cost is the consent banner itself — the measurable drop from a full tracking stack to one that runs for the users who click accept, which is a fraction of the traffic and a biased fraction at that. Every model trained on post-banner data is trained on a sample of the population that is, for reasons nobody fully understands, not like the whole.

The larger cost is the quiet one. The industry's entire knowledge base — the search relevance systems, the behavioural models, the ten-year trend charts — was built with data collected under rules that no longer exist. Much of it is now unsourceable. The companies that rebuilt their measurement layer after the banners have a working instrument. The ones that did not are driving on the old chart, at night, on roads that may have moved.

The unglamorous survivors

What it means for the actual product is subtle: the adult video site that still knows what its users do, in a way that would survive an audit, is the one that rebuilt the boring plumbing in the last five years and told nobody. You can feel it in the search relevance, which holds up when the traffic mix shifts. And you can feel the alternative, which is a once-good site slowly drifting away from its own audience because the feedback loop got legally severed and nobody replaced it.

The banner was not the hard part. The hard part was what the banner asked for: a way to know your users that could survive them saying no.