Vinted's homefeed isn't a single ranked feed shared across all buyers — it's personalised per buyer based on their recent search vocabulary, which means the same listing surfaces in some buyers' feeds and is invisible to others, and the title is the lever that decides which cohort your listing lands in. The mechanic. Vinted's algorithm logs each buyer's last 50 search queries — both the queries the buyer typed and the auto-complete terms they tapped — and uses that vocabulary to weight which listings surface in that buyer's homefeed scroll.
A buyer whose last 50 searches included 'y2k', 'vintage', '90s', 'retro', 'noughties' has a homefeed weighted toward listings whose titles contain those words. A buyer whose last 50 searches were 'size 12 jeans', 'UK 8 boots', 'M dress', 'XS coat' has a homefeed weighted toward listings that lead with size. The same Levi's denim jacket — same item, same condition, same price — surfaces in the first buyer's homefeed if you title it 'vintage 90s Levi trucker denim jacket', and surfaces in the second buyer's feed if you title it 'Levi's denim jacket size M trucker'.
Title it as 'used Levi's jacket good condition' and it surfaces in neither feed because the vocabulary doesn't match either cohort's recent searches. Three rules for matching title vocabulary to cohort. One — pick the cohort before you title, not after.
Look at the item and ask which buyer is most likely to be searching for it tonight. A statement vintage piece is a vintage-cohort item — lead the title with the era and the vintage-vocabulary words. A practical size-led wardrobe staple is a size-cohort item — lead the title with the size and the brand.
Look at the item and ask which buyer is most likely to be searching for it tonight.
A category piece like kids' wear sits in its own cohort — lead with age and gender. Two — the lead-word weight is highest, the second-word second-highest, and so on; the title's first three words do 70% of the cohort-matching work. 'Size 12' as words one and two attracts a different feed than 'vintage 90s' as words one and two on the same item. Three — don't try to attract both cohorts at once.
Titles that try to cover both — 'vintage 90s Levi denim jacket size M trucker UK 8' — fail to match either cohort cleanly because the vocabulary signals are diluted. Pick the cohort with the larger buyer pool for that specific item in the current week and write the title for that cohort only. The four-week test that proves the cohort effect.
Take ten of your existing listings, split them into two groups of five, rewrite group A's titles for vintage-cohort vocabulary and group B's for size-cohort vocabulary, leave the photos and prices identical, and measure the favourite-rate over four weeks. The split is typically 2-3x favourites on the cohort-matched titles versus the cohort-neutral titles. VintSnap's title generation isn't cohort-blind — the prompt asks you to specify whether the item is vintage, current-season, or practical-staple, and the title vocabulary the model produces reflects that cohort framing rather than producing generic catch-all titles that match no cohort cleanly.
Part of our Reseller economics series — field notes from building VintSnap.