Performance is not a channel. It is a standard of proof.

Somewhere along the way 'performance' stopped describing a way of working and started describing a media list: paid search, paid social, affiliates, the bottom of the funnel. Everything above it became 'brand', and the two got separate budgets, separate teams, and separate standards of evidence: one held to a ROAS target to two decimal places, the other to a tracking study once a year.

This split is convenient and wrong. It is convenient because it lets each side avoid the other's hardest question. It is wrong because the customer does not experience a funnel, and because the channels sitting in the 'performance' column are frequently the ones harvesting demand that something else created.

A channel is not high-performing because it is cheap to attribute. It is high-performing because turning it off costs you money.

The only definition that survives contact

Performance means every meaningful spend decision carries a claim about what would have happened otherwise, and that claim is testable. That is the whole definition. It says nothing about which formats you buy. A television campaign with a clean geo holdout is performance marketing. A paid search account reporting last-click ROAS on branded terms is not, whatever the line item says.

What it costs to work this way

• You have to run holdouts, which means deliberately not spending money in places where you believe spending works. This is unpopular and it is the price of knowing.
• You have to accept slower answers. An incrementality read takes weeks; a dashboard takes seconds and is frequently wrong.
• You have to publish results that make you look bad. An agency that has never told a client a channel was not working has not been measuring it.
• You have to hold brand investment to a standard too; a longer, coarser one, but a standard.

This is the standard we hold our own work to, including when the honest answer is not the one a client was hoping to hear.