How do you recommend products to a first time visitor with no history at all?

October 29, 2026

Short answer: With no history, recommend from context, not identity: the landing page, the entry product, season and geography at the coarse level, and bestsellers within the category the visitor chose, not the whole store. Then update from the first two or three clicks, which say more than any third party profile.

Bestsellers are a confession, not a strategy

The sitewide bestseller rail is what a store shows when it knows nothing, and visitors can tell. A person reading about trail shoes does not need the store top selling candle. Category level popularity is the honest version of the same data: the products people like this visitor, arriving at this page, actually bought.

Geography and season, used coarsely, are the next layer. Rain gear rises in a wet week in the north; that is not surveillance, it is shopkeeping, and it predates the internet by a century. The line to hold is precision: region and season are fine, a guessed identity profile rented from a data broker is where relevance turns into creepiness and, often, into error.

The first session is the profile

By the third product view, the visitor has told you more than any onboarding quiz would: price sensitivity from what they opened, category depth from what they compared, intent from whether they touched size and delivery information. Recommendations should re rank on that behaviour within the session, visibly enough that the shopper feels the store paying attention.

Stores underuse the strongest cold start signal they own: the search box. A typed query is a stated need. Recommendations that ignore the session searches and keep pushing the entry page category are leaving the clearest instruction on the table.

Ask, but ask like a shopkeeper

A short either or prompt can earn its place: shopping for yourself or a gift, this season or next. Two taps that change the rails beat a five step style quiz that most visitors abandon and the rest answer carelessly. Every question must visibly change something on the page in the same session, or the visitor learns the store asks and does not listen.

And know when to stop asking. A cold start visitor who has found their product and is reading delivery terms is done being profiled. The best recommendation at that point is a quiet page and a working checkout.

What good looks like by the second visit

The returning stranger is where cold start work pays compound interest. If the first session ended with views but no purchase, the second visit should open with those products and their close alternatives, in stock, at the prices seen before or honestly marked as changed. That continuity, not a rented profile, is what personalization means to the person experiencing it.

Keep reading: How should stores personalize the homepage for a shopper who arrives from a single product ad?, Search on your own store: fixing result pages that return nothing, and Bundle offers that lift basket size without training discount waiting cover the nearby cases in this series.