Signal loss

Signal loss is the umbrella over several unrelated mechanisms that happen to have the same effect on your reports: tracking prevention shortening or blocking identifiers, visitors declining consent, platforms restricting what they return, and identifiers being stripped in transit.

They have different causes and different fixes, which is why the umbrella term is convenient and occasionally lazy. “Signal loss” as an explanation for a bad quarter should always be followed by which mechanism, because the remedies do not overlap at all.

The distinction that matters most: signal loss is about what you can see, not about what happened. Customers did not stop buying because a browser expired a cookie. The sales are still there; the attribution is not. Confusing the two leads to cutting channels that are working but are no longer traceable — which is the expensive version of this mistake, and a common one.

What fills the gap. Platforms respond with modelled conversions, estimating what they can no longer observe. That keeps totals plausible and means your reported figures are increasingly part measurement and part inference, in a ratio you cannot see.

What actually helps. Identifiers that do not depend on browser storage — a click ID carried into your own database, or a match on something the customer gave you. Aggregate methods that never needed user-level data, like marketing mix modelling. And holdout tests, which measure effect rather than tracking and are therefore indifferent to all of this.