Product Recommendations

Feed-Based Product Recommendations

Common Input Signals for Feed-Based Product Recommendations

Temu’s feed-based product recommendations are those surfaced on the platform without your search queries, such as the ones you see immediately when you enter the platform’s homepage.

These recommendations use various input signals to predict how likely you are to click on or purchase the products, and rank them accordingly. While the relative importance of the input signals is subject to change, some signals are generally considered more important because they have a higher likelihood of providing more responsive matches with your interests and are thus used more broadly across different feed-based product recommendation scenarios. The signals about your interactions with Temu (such as your browsing history, searching history, and past purchases) are generally weighted more heavily than other signals.

These common input signals include:

Your interactions with Temu
Your browsing history
Your searching history
Products you have clicked and related time spent
Products you have saved to your wish list
Products you have added to shopping cart
Products you have purchased
Your return and refund history

Item attributes
Product intrinsic characteristics (e.g., size, material, and specifications)
Product features (e.g., product category, title, price, pictures, and affiliated store)
Product popularity and statistical metrics (e.g., sales volume; average product ratings; how often a product has been shown, clicked, added to shopping cart, added to wish list, purchased, or returned in aggregate terms; aggregated click-through rates calculated as the ratio between how often a product is clicked and how often it is shown; and aggregated conversion rates calculated as the ratio between how often a product is purchased and how often it is clicked)
Whether and how much traders have paid to promote the products

Contextual information
Your region, language, and currency settings
Information about the device you are using when browsing Temu
Timing of your interactions with Temu

Feed-Based Product Recommendation Scenarios

Temu offers feed-based product recommendations across various sections of its interfaces. These scenarios differ primarily in the product pool from which the recommender systems retrieve products for recommendations. All these recommendation scenarios rely on the same recommendation logic underlying feed-based product recommendations and consider the same set of common input signals listed above for retrieval, scoring, and ranking. Across these scenarios, signals about your interactions with Temu (such as your browsing history, searching history, and past purchases) are considered more important than other signals.

The “Best-Selling Items” entry point on Temu’s interfaces features the top-selling items in their respective product categories within your current region setting. These products are ranked entirely based on their gross merchandise value over the last 30, 14, or 7 days (depending on the filter you select), respectively. Temu does not deploy scoring or ranking recommendation models or utilise input signals about your interactions on Temu to provide these recommendations.

General Feed-Based Recommendations

The general feed-based product recommendation is the most common feed-based recommendation scenario that surfaces in different sections across Temu’s interfaces, such as the “Explore Your Interests” section on the homepage and product recommendations on the order details page, the order list page (displaying your previous orders), the product listing page, your shopping cart page, the “You” section in the app, in-app review pages, and in-app messages. In these sections, Temu retrieves the full product catalogue on offer in the region from which you are viewing Temu to recommend products we believe you are more likely to click on or purchase.

Certain other sections, including product recommendations on store and brand pages, use the same recommendation logic. The only difference is that for these sections, Temu retrieves products for recommendation from a filtered, context-based product pool rather than the full product catalogue available in your region.

For the general feed-based product recommendations, we use the common set of input signals listed above to surface the relevant product recommendations.

Feed-Based Product Recommendations within Sales Campaigns and Promotional Activity Pages

You may see Temu sales campaigns and promotional activities, such as seasonal sales (e.g., summer sales and winter sales), anniversary sales, coupons, and credits, through various banners across Temu interfaces (such as your homepage), marketing messages, pop-up windows, or your shopping cart page.

Product recommendations within these sales campaigns and promotional activity pages feature only products participating in these campaigns and activities.

We use the common set of input signals for feed-based product recommendations to surface the relevant product recommendations.

“5-Star Rated”

The “5-Star Rated” entry point on Temu’s interface features products with customer reviews averaging five stars. The common input signals we listed above for feed-based product recommendations are used to surface products we believe you are more likely to click on or purchase.

"Local Warehouse"

The “Local Warehouse” entry point on Temu’s interface (available in certain EU countries) features products sold by traders who can ship directly from European warehouses, potentially reaching you faster. Temu’s recommender systems utilise the common input signals for feed-based recommendations to surface these products.

"Top Picks"

The “Top Picks” section on Temu’s interfaces features a selection of products whose traders offer price benefits. The common input signals we listed above for feed-based product recommendations are used to surface products we believe you are more likely to click on or purchase.

Search-Based Product Recommendations

Temu’s search-based product recommendations produce recommendation results in response to your search queries.

In addition to using the same set of common input signals as in feed-based product recommendations, Temu’s recommender systems also consider the relevance of the intrinsic product characteristics and product features to the textual or image query you input to predict how likely you may click or purchase the products, and rank the products accordingly. For example, your search for “size 9 red shoes” will lead Temu’s recommender systems to recommend listings whose titles, descriptions, and sizes match your query.

The relevance of intrinsic product characteristics and product features to your search query, and the input signals related to your interactions, are generally weighted more than other signals, leading to the prioritisation of products that match your query standards and interests.

Advertised Product Recommendations

You will see products labelled with “Ad” from time to time, among both feed-based and search-based product recommendations. The “Ad” label means the products are advertised by their traders at an advertising fee.

How the “Ad”-labelled products are ranked among other advertised and unadvertised products takes into account the same set of common input signals as in feed-based product recommendations discussed above, including whether and how much traders have paid to promote the products.

Whether and how much traders have paid to promote the products, along with input signals related to your interactions, are generally weighted more than other signals in determining the ranking of “Ad”-labelled products.

Fine-Tuning Product Recommendations

Diversifying your experience: We aim to provide a more diverse Temu experience for you. If we find Temu's recommender systems displaying too many products with similar attributes, we will space out those products so you can see a wider variety of options.

Filtering & Sorting controls: To help you narrow down and browse products more efficiently under specific product categories or through your search results, Temu provides user-controlled filtering and sorting options.

Filtering options displayed at the top of each list allow you to narrow results by:

Product features (e.g., category, colour, size, material, brand, seasonality, specifications)
Price range
Review rating range
Shipping location

Sorting options allow you to reorder the recommended products by:

Relevance (default): Products most relevant to your search will be shown first
Top Sales: The best-selling products will be shown first
Most Recent: Products most recently added on Temu will be shown first
Price low to high: Products will be displayed from the most affordable one to the most expensive one
Price high to low: Products will be displayed from the most expensive one to the most affordable one