Personalization is usually framed as a data problem. In practice, most of it fails as a design problem: layouts that jump, content that flashes the wrong version first, and pages that feel slippery because they never look the same twice.
Avoid layout shift
The classic failure: the page loads a default, then swaps in the personalized version, and everything jumps. It feels broken and it hurts your Core Web Vitals. Reserve space for variable content and render the right version up front wherever you can, instead of patching it in after the first paint.
Keep a stable skeleton
Vary the content, not the scaffolding. The visitor should still recognize the page as the message changes. Personalize the headline, the proof, the hero, the primary action, while the structure, navigation, and rhythm stay constant. A stable skeleton makes change feel intentional rather than chaotic.
Personalize the substance, not the chrome
Moving the navigation around or recoloring buttons by segment adds cognitive load with no payoff. Spend your personalization where it changes the decision: the promise, the evidence, the ask.
Design the fallback first
Every personalized slot needs a strong default, for the visitor you cannot read yet and for when data is missing. If the default is weak, personalization is just papering over a bad baseline. Design the generic version as if it were the only version, then improve from there.
Make it fast
Personalization that adds a visible delay loses more than it gains. The decision about what to show has to happen quickly enough that the visitor never waits on it. Speed is part of the design, not an afterthought bolted on later.
Respect and control
Let people understand, and where it matters adjust, what they are seeing. Personalization that feels done with the visitor beats personalization done to them. (More on customer-first personalization.)
