What is adaptive CRO?

Why does deciding per visitor matter?

Because a population-level answer averages opposite needs. A reassurance that helps a hesitant visitor can irritate a decided one, and a single variation shown to both measures as roughly no effect — the two cancel.

This is the structural limitation of fixed-variation testing, not a flaw in how it is usually run. A test asks 'is A better than B for everyone', and when the honest answer is 'A for these people and B for those', the result is a wash and the opportunity is invisible.

The corollary is that adaptive approaches are hardest to justify where visitors are homogeneous. If everyone arriving wants the same thing, one good answer is enough and simpler.

How is adaptive CRO different from personalization?

Personalization adapts to who someone is — segment, source, history. Adaptive CRO adapts to what they are doing right now, and is judged on whether conversions rose rather than on relevance.

The distinction is not academic. Attribute-based personalization can be decided before the visitor does anything, which makes it easy to configure and blind to the session: it treats a returning customer who is about to churn identically to one about to buy again. Behaviour-based adaptation cannot be decided in advance, which is harder to build and is the entire point.

What does adaptive CRO require to work?

Three things: a live reading of the session, a model of the site so the response fits the page, and a control group — because a system that chooses whom to help cannot otherwise tell its effect from its selection.

The third is the one most often missing, and its absence is not neutral. Adaptive systems select the visitors they act on, so any comparison between treated and untreated visitors is partly a description of that selection. Without random assignment there is no way to separate the two, and the reported result flatters the system by construction.