Website Personalization
Showing different visitors different versions of your website based on who they are, where they came from, or what they're trying to do, so each person sees the experience most likely to convert them.
Website personalization is showing different visitors different versions of the same page, based on what you know about them. A static website shows everyone the same thing. Personalization breaks that: a returning customer might see a loyalty offer, a visitor from a Google Ads click sees a headline matched to the keyword they searched, a mobile shopper sees a stripped-down hero. Same URL, different experience.
The signals personalization uses
Personalization decisions are driven by context, the things you can observe about a visitor before they act. The common signals are:
- Traffic source: which ad campaign, search query, or referrer brought them in
- Device and screen size
- New versus returning visitor
- Geography or language
- Past behavior, such as pages viewed or items already in the cart
Each signal is a clue about intent. Someone arriving from a "cheap running shoes" ad wants something different from someone who typed your brand name directly, and personalization is what lets the page meet each of them where they are.
Why it matters for conversions
Done well, personalization eliminates the "average winner" problem in A/B testing. Two variants might tie overall, but one wins on mobile while the other wins on desktop. A standard test forces you to pick one and ship it to everyone. Personalization ships both, each to the audience it actually works for, so you stop sacrificing one group's experience to lift another's.
Where it usually goes wrong
Traditional personalization is a manual, multi-team project. Marketers hand-build the segments and rules, designers produce a variation for each one, and developers wire it all up. The work doesn't scale: every new segment is another round of the same effort, so most teams get one or two segments live, declare victory, and never touch it again. The promise of "the right experience for every visitor" quietly becomes "two slightly different homepages."
Our take
At Dalton, we think personalization shouldn't be a separate project bolted onto your optimization work. It should fall out of it.
Our system is a contextual bandit, which means it isn't just learning which variant converts best overall, it's learning which variant converts best in which context. The same engine that optimizes your site is already paying attention to the signals above, so as it gathers data it can tell that one variant pulls ahead on mobile while another wins for returning visitors, and shift allocation accordingly. The "average winner" problem and the personalization problem turn out to be the same problem, solved by the same algorithm.
The practical upshot is less manual work. Fewer hand-built segments, fewer engineering tickets, and a site that grows more relevant per visitor as it runs, rather than freezing at the one or two segments a team had time to set up. You still approve what goes live; the system handles the matching.
That's the version of personalization we think most brands actually wanted: not a quarter-long project that stalls at two segments, but a natural result of optimization that never stops.