Home / Case Studies / SodaStream
SodaStream · Dynamic flavors category
Ranked per visitor
A category that re-sorts itself: segment and recommendation models on SodaStream's flavors page
The challenge
Flavors are SodaStream's repeat-purchase catalog — including the PepsiCo drink-mix line — and a category page with one fixed sort order serves the average visitor. The average visitor doesn't exist: a first-timer browsing bestsellers, a Pepsi loyalist, and a returning customer who always buys the same three flavors each need a different shelf.
The approach
Dynamic ordering instead of a static sort: the product ranking is generated per visitor, driven by audience segment — source, owner status, purchase history — not one merchandiser-set sequence.
Recommendation models under the grid: what surfaces first is what the model predicts this visitor is most likely to buy, learning from behavior across the catalog.
Proof kept close to the decision: "new" and "best seller" badges, review counts and anchor pricing stay attached to the ranked results, so the personalized order still carries social proof.
Friction kept out: add-to-cart directly from the listing — the personalized shelf converts in place, without detours.
The result
The flavors category became a dynamic page — sorted by segment and recommendation models per visitor — turning a fixed catalog grid into a personalized, conversion-ready shelf for a repeat-purchase line.
The takeaway
Sort order is a ranking model, not a setting. On repeat-purchase categories, the shelf should learn who's looking at it.
Your turn
Same method. Your funnel.
The audit finds your version of these leaks — priced in recoverable revenue, with the test plan attached.