Recommended content on Vinted

We use your listing history and personalization settings. As you type, we show keyword suggestions ranked by relevance. We don’t use (nor try to predict) any personal data that might be considered sensitive, such as race or religion.

Our recommender systems use the following parameters to select and rank listings displayed to members (and to visitors, in some cases):

• Time since item upload: this parameter helps to evaluate the chances of an item being sold
• Number of interactions an item has already collected (including impressions, favorites, clicks): this parameter indicates the previous level of interest in the item and how likely it is to be sold. This parameter can also apply to visitors who are not logged-in
• Item and buyer location country: this parameter ensures that items can actually be purchased and are not subject to shipping and other logistical restrictions. The seller’s country is also a parameter that can be used to make recommendations to potential buyers
• Item lister and buyer preferred language: this parameter is used to show items to members in their preferred language, which can help to improve the relevance of recommendations
• Item price: this parameter aims to recommend better deals to members, so the item price is used as a parameter to determine its relevance
• Item ontology: this parameter includes the category, brand, size, color, listing photos, and other attributes that help define an item's characteristics. This parameter also helps to ensure that items shown are relevant to potential buyers’ interests
• Item condition: this parameter is used to show items in the appropriate condition, based on buyers' preferences
• Listing frequency: this parameter is used when sellers upload a number of items within a specific timeframe, showing those items to buyers in a proportionate way
• Seller type: If the seller is a regular user or a Pro Seller (in markets where Vinted Pro is available)
• Expressed filters and search query: this parameter is applicable only for the search query and is used to show only the items matching the expressed intent
• Item descriptions: this parameter is used to evaluate the relevance of an item to the search query
• Illegal content: certain measures we take to limit and stop illegal content may impact the visibility of members' legitimate listings
• Image resemblance: this parameter is used when the buyer searches by image, showing items that look similar to the uploaded image 
• User history: this parameter helps new members improve their selling experience by promoting the visibility of their listings. These listings are recommended to other members who have seen or purchased similar items in the past
• User’s following list: this parameter helps sort items based on the members that a given user follows

We also promote some segments of content, such as new listings and listings which are part of paid promotion services — like item Bumps. Members can purchase these promotions to make their listings more visible, which gives them more chances to sell.

The most important parameters are:

1. the filters that members apply, such as search filters or customized sizes, and
2. the filters that our system applies, such as filters of items available in a member’s country.

Promoted content (such as new listings, or listings as part of our paid promotion services) is also an important factor. These items get increased visibility through rules that limit how often they appear to avoid overwhelming members.

The next key factors are item attractiveness criteria (such as recency, the number of interactions, price, and condition) which help us evaluate sales likelihood and how attractive an item is to potential buyers.

Personalization based on historical interactions also matters. Though past activity may not always reflect current interests or needs, it still improves recommendations by showing relevant items.

We believe these are the most important parameters as they help provide the best search experience for our members. With these parameters, we ensure that the listings presented higher in ranking match search expectations. However, these priorities can vary depending on the specific use-case and members’ behaviors.

Our recommender systems are used in several parts of Vinted. We’re continuously improving them and their usage, so we can show more relevant items in all parts of the platform.

Here are the main areas where we use recommender systems:

Newsfeed

We use two types of algorithms in the homepage newsfeed: recommenders and re-rankers. The recommenders select the listings that are recommended to you, and the re-rankers rank the selected listings by estimated relevance to a member.

Search results

When searching and browsing the catalog, the results can be filtered based on criteria such as category, brand, price range, and more. Once filtered, our system ranks the listings based on relevance to the query and to you. The ranking algorithm takes into account several factors, such as item attractiveness, query relevance, and your historical interactions.