How does the Vinted algorithm work in 2026?

For most sellers, Vinted feels like a lottery. One item flies off the shelf within ten minutes, an almost identical piece sits for weeks without a single view. It’s tempting to chalk that up to luck. It isn’t. Behind every listing sits a technical system that can be understood fairly precisely, and once you know how it works, it explains a large chunk of what used to feel random.

Not a chronological list, but a search engine with a filter in front of it

Vinted doesn’t show you a simple timeline of newest items. Every time someone searches or scrolls through their feed, something happens in two steps. First the system pulls a large group of candidates, up to around 200,000 items that roughly fit that search or user profile. Then a second layer, a so-called ranker, decides which 20 to 50 items that person actually gets to see.

That means something fundamental: if your item doesn’t make that first selection of 200,000, you don’t get a few views. You get zero. Not because your photos are bad or your price is wrong, but because the system simply never considered your item at all. I call that the zero-view trap myself, and it’s the most underrated cause of a listing going silent.

Vinted searches by meaning, not just words

Until a few years ago, Vinted largely worked with simple keyword matching: type “cable knit sweater” and you’d get items with that phrase literally in the title. Since the switch to a new search engine, Vespa, the system works differently. It uses what’s called vector search, through a so-called two-tower model — the same underlying system that also gives the boost feature its momentum effect.

In practice it comes down to this. The system builds a kind of numeric profile of every buyer based on search behaviour, click behaviour, price preference and brand preference. It builds a comparable profile of every item based on title, photos, category, brand, price and the seller’s reputation. The closer those two profiles sit to each other, the higher the odds your item appears in that specific buyer’s feed. That explains why two buyers typing the same word still see different results, and why an item sometimes gets found without the exact search term appearing in the title.

For you as a seller this means keywords still matter, but not as the only factor. Title, photos, price and even your selling history work together to determine which buyers see your item as relevant.

Explore and exploit: why new items get tested first

Every recommendation system, and Vinted is no exception, works with a balance between two modes. In the exploit phase the system mostly shows what it already knows performs well. In the explore phase it deliberately gives new or unproven items some extra visibility too, purely to gather data on how buyers respond.

That explains the temporary boost a freshly uploaded item gets. The system tests your item in the first hours on a small group of buyers. If they respond well, with views, favourites and messages, the system shifts your item toward the exploit phase and keeps showing it more often. If they don’t, your item falls back into the general pool and now has to compete on price, photos and existing favourites against thousands of other items. That first test phase lasts around 48 hours in practice, after which an average item loses 60 to 70 percent of its visibility if the test results weren’t strong enough — exactly the window that also forms the starting point of the day-by-day path for an item that isn’t selling.

The three-favourite threshold

One concrete signal within that system is the number of favourites an item collects. Once an item reaches three favourites, the system appears to mark it as “popular” and shifts it noticeably toward the exploit phase, with a higher chance of appearing in other people’s feeds. For children’s clothing that threshold is notably lower, where a single favourite already seems to be enough. That’s exactly why the first days after upload matter so much: every early favourite contributes to crossing that threshold, which in turn produces more visibility. A flywheel that reinforces itself, in both directions — and also exactly why what you do with a favourite is so much more than just waiting it out.

The 60 percent price rule

Price doesn’t only play a role in the buyer’s decision, it also seems to play a role in how the algorithm evaluates your item. Items priced around 60 percent of the going market value for comparable pieces appear to get a noticeable extra visibility boost in the feeds of new users. That’s a deliberate reward for what the system recognises as a good deal. Go too far below that, well under 60 percent, and the effect seems to flip: a price that feels suspiciously low tends to trigger distrust rather than enthusiasm, both with buyers and, apparently, in how the system treats the item.

Reputation as a ranking factor, not just a trust signal

Your completion rate and your average review score don’t just play a role in what a buyer sees when they click your profile. They also appear to weigh in on how often the system shows your items in the first place. Sellers with a strong reputation and few cancelled transactions get structurally more visibility on comparable items than a new or inexperienced seller. That makes sense from the platform’s perspective: Vinted benefits from transactions going well, and rewards sellers who deliver that consistently.

What this practically means for your listings

This technical background translates into a handful of concrete habits.

Signal	What the algorithm appears to reward
Title and description	Specific search terms: brand, model, colour, style
Upload timing	Uploading when many buyers are active
Response speed	Replying quickly to messages and offers
Price	Around 60% of the market value of comparable pieces
Seller reputation	High completion rate, few cancellations

Make sure your title and description contain enough specific words, brand, model, colour, style, so the system can link your item to the right search profiles in that first selection stage of 200,000 candidates. Upload when many buyers are active, so the 48-hour test phase generates enough signal to reach the exploit phase. Respond quickly to messages and offers, since a short response time also appears to count as a positive signal. And price your item deliberately, not blindly rounded or high, but in line with what comparable pieces actually sell for, somewhere in that zone of roughly 60 percent of retail or market price that the system seems to reward.

Why it still sometimes feels random

Even with this knowledge, there’s still a layer of unpredictability, and that’s inherent to how these kinds of systems work. The model keeps testing, keeps adapting, and the exact weights it assigns to each signal aren’t public and undoubtedly change with new versions of the system. What can be said with confidence today is based on patterns from my own sales data and what Vinted itself has publicly shared about the underlying technology, not on complete internal documentation. Treat the principles above as strong working hypotheses that keep getting confirmed in practice, not as an exact manual.

Vinted works with a candidate pool of up to 200,000 items per search, followed by a ranking system that picks the 20 to 50 items a buyer actually sees from that pool. Whoever understands that stops hoping for luck and starts deliberately working with a system that, under the hood, just follows consistent logic.

Frequently asked questions

How does the Vinted algorithm work?

For every search, Vinted first pulls a candidate pool of up to 200,000 items, and a second layer, the ranker, picks the 20 to 50 items a buyer actually sees from that pool. That matching happens through vector search based on title, photos, price, category and seller reputation, not just keywords.

What is the zero-view trap on Vinted?

The zero-view trap is when an item never makes that first candidate selection and ends up with zero views, not because the photos or price are bad, but because the system never considered the item at all. A title and description with specific search terms like brand, model and colour reduce that risk.

How many favourites does an item need to rank better on Vinted?

Once an item reaches three favourites, the algorithm appears to flag it as popular and shifts it noticeably toward more visibility. For children’s clothing that threshold seems lower, where a single favourite already seems to be enough.

How long does the test phase last for a new item on Vinted?

Roughly 48 hours. During that period the system tests a new item on a small group of buyers. If they respond well, the item keeps getting shown more often. If they don’t, an average item loses 60 to 70 percent of its visibility.

What is the best price according to the Vinted algorithm?

Items priced around 60 percent of the going market value for comparable pieces appear to get a noticeable extra visibility boost in the feeds of new users. Go far below that 60 percent, and the effect seems to flip toward suspicion instead of enthusiasm.