How Wolt’s search and recommender systems work
How Wolt’s search works
Wolt Search is designed to help you quickly and easily find products and services that are relevant to you. Here you can find information about why certain search results are shown to you.
Note that Wolt has two search functions that work differently: the standard search in the Wolt app and on the Wolt.com website, and the partner-specific search within a specific restaurant or business.
Standard search
The standard search is available from the “🔍 Search” button at the bottom of Wolt’s Discovery page or from the search bar on the Wolt.com website. Standard search allows you to search the entire Wolt platform for products, services, and venues.
When you enter a search query, Wolt matches your query with our venues (e.g. restaurants, grocery stores, pharmacies, etc.) and their product assortment (e.g. dishes, groceries, clothing). The search results are grouped in different sections that you can use for narrowing down the results into the specific domain you are interested in. For each query, Wolt adjusts the order of these sections to present you with the most relevant ones for your query first. Within each section, you will find relevant results for your query. Each resulting venue will be presented together with their most relevant products for your query.
When using the default sorting (“Recommended”), the most relevant venues and products are shown first. Wolt uses machine learning to understand which of our venues and products are relevant for your query and for you. The system uses information about our venues (e.g. their ratings), our products (e.g. their names and descriptions) and your personal preferences (your favourite venues) to find the best results for your query. The most important factors in the ranking (in the order of priority) are: delivery and pricing features, query to venue matching, venue quality and popularity, query to product category features, other factors (e.g. favourites). Additionally, venues that are closed or with “pick up only” options are downranked in the results.
If you cannot find what you are looking for, you can also sort the results by other criteria by clicking the “Sort by” button on the filter bar. Here you can also apply other filters to narrow down the results to your current preferences.
Wolt automatically stores the user's search history locally on the user's device to enhance the user's experience by suggesting previous searches. If a user wishes to delete their search history, the user can remove the previous searches manually by clicking on the “clear” option in the Search menu or by signing out of the app.
In-venue search
You can also search the products that you are looking for within a specific partner venue page. In order to do so, simply open one venue and type your query in the search bar at the top of the venue’s page in the Wolt app or at the right hand side of the menu navigation bar in Web.
Wolt uses machine learning to find the most relevant products for your query within the venue. The system uses information about the venue’s products (e.g. their names, descriptions or categories) to understand which ones are relevant for your input query. Roughly, the most important factors for the final rank of the results are (in the order of priority): item to query matching features, item popularity and engagement, item product family relevance, venue-related features. The results are grouped in sections according to the partner’s menu or selection, appearing the most relevant products for your query at the top of the results.
Category pages
In addition to the Wolt search, there are a series of landing pages allowing you to browse by category (e.g., restaurants, groceries, electronics). The category landing pages allow you to find restaurants or relevant shops in your area. These sections can be accessed in the Wolt app or website Wolt.com via the navigation bar. The category pages are by default sorted using personalised ranking and internal business rules as described below. You can also sort & filter the pages using different options, such as distance, delivery price or the partner’s customer rating. Offline or currently unavailable venues are sorted at the end of the page.
Restaurant ranking
When you open the restaurants page, the order in which the partners are shown is determined by an automated ranking system. The ranking uses a combination of contextual information and (where available) your past interactions with Wolt.
Main parameters we use:
Your location
We prioritize partners that can deliver to your current delivery address and that are closer or otherwise more feasible to deliver from.
Date and time
We take into account when you browse (e.g. lunch vs. late evening), including whether a partner is currently open and available for delivery at that time.
Partner attributes
We may consider attributes displayed or available on Wolt, such as eligibility for certain benefits (for example, Wolt+ participation), as well as general popularity on Wolt (partner sales figures on Wolt). We may also use performance-related signals derived from aggregate user behaviour (for example, how often a partner is selected after being shown)
Your past purchases (if any)
If you have previously placed orders, the ranking may give greater visibility to partners you have ordered from before and to partners similar to those you have ordered from, based on partner attributes (such as cuisine category and other listing attributes available on the platform).
Your historical clicks and views (regardless of whether you’ve purchased)
We take into account which partners you have clicked, viewed, or otherwise interacted with in the past. This can increase the visibility of those partners or similar partners.
Whether this is your first order on Wolt
If you have not placed any orders yet, we may adjust ranking to help you discover relevant partners (for example, by relying more heavily on location and time-based availability, rather than purchase history).
The parameters above are used together to increase the relevance of the list for you (personalization) and to ensure practicality (e.g. deliverability and availability). Not every factor is used for every user. For example, purchase history is only used once you have placed an order.
Stores (non-restaurant venues) ranking
When you open the Stores page (non-restaurant venues), the order in which stores are shown is determined by an automated ranking system. The ranking uses a combination of (i) contextual and delivery-feasibility signals, (ii) store performance and catalogue information, and (iii) where available, your interactions with Wolt.
Main parameters we use
Your interactions (where available)
We consider your past interactions with stores on Wolt, such as which stores you have clicked or viewed, and how often you have been shown a store (impressions). This can increase the visibility of stores you have interacted with, or similar stores.
Delivery and price-related features
We take into account delivery feasibility and convenience, such as the estimated delivery time, delivery price/fee, and other delivery-specific features that affect whether an order can be fulfilled efficiently.
Store popularity and conversion-related performance
We consider store popularity on Wolt (partner sales figures on Wolt)
We may also use performance-related signals derived from aggregate user behaviour (for example, how often a partner is selected after being shown)
Store retention and reliability
We may consider the share of customers who return to order again from a store (repeat ordering / retention).
We also consider indicators of order fulfilment quality (for example, fulfilment success and operational reliability), which can affect ranking.
Assortment
We may consider the size of the store’s assortment as a signal of how likely the store is to meet typical shopping needs.
Discounts and basket-level effects
Active discounts may influence ranking.
We may also use signals derived from basket behaviour, such as typical order size and differences between customer basket sizes and store basket sizes, as these may correlate with relevance and fulfilment patterns.
Category and country context
Ranking models and the relative importance of signals may differ across product categories and countries to reflect differences in local selection.
The parameters above are used together to increase the relevance of the list for you (personalization) and to ensure practicality (e.g. deliverability and availability). Not every factor is used for every user (for example, interaction history depends on whether you have previously used the service). In general, delivery delivery feasibility acts as primary constraint, while estimated delivery time, performance and interactions signals refine the order of stores shown.
Discovery page
The Discovery page is the entry point for every customer who visits Wolt. This page is designed with several carousels, each containing multiple items. These items can be either venues, dishes, products, or banners, which link to venue(s), menu items, or an editorial article.
The vertical order of the carousels on the Discovery page is determined based on the city popularity of each carousel, measured by customer interactions. This ensures that the most engaging and relevant content is presented to you. Some vertical carousel positions are manually determined.
The venues displayed in these carousels are selected based on their availability to deliver to the customer's current location and internal business rules. Their horizontal order follows the same algorithms described above in the “restaurant ranking” section. Though some carousels follow a different order, eg. “Order Again” is ordered by recency of the purchase. For banners, the horizontal order is defined manually.
If you follow the “see all” link next to a carousel, a list of venues is shown. They are ranked with the algorithms described above depending if they are restaurants or stores.
To maintain a diverse and engaging user experience, we've implemented a system to reduce duplicate content. Venues that appear higher up in a carousel are pushed towards the end of any subsequent carousel further down the page. This approach keeps the Discovery page fresh and interesting for you, while ensuring that all content is presented fairly and transparently.
Recommendations in multi-venue ordering
In certain orders, you may be able to add items from other venues after placing your initial order. Venues recommended to you for such orders are selected based on their location, being in the vicinity of the original restaurant or store your ordered from or the delivery address of your initial order. In addition, venues are limited to those that are readily available to deliver your order while certain venue categories like charities, gift-card venues, etc. and venues delivering with their own delivery personnel are implicitly excluded. The venues are generally ordered by popularity (partner sales figures on the Wolt platform) and how well their offerings complement your original order.
Recommended items for each of the venues are again selected based on the availability to be delivered, selected to best complement your original order where possible, and finally ordered by their popularity.
Paid placement
The Discovery page, the order of your search results in the standard search and product lines and categories section can also be influenced by paid placement. In this case, such results may be ranked higher than they would be based on the parameters listed above. Partners whose position is influenced by paid placement are marked as “Sponsored”.
---
Wolt reserves the right to amend this information as it may change its algorithms, practices and parameters from time to time. Wolt may also experience technical issues or conduct certain trials in a limited manner that have an effect on the visibility, rankings and search results that will not be reflected in these descriptions due to their limited nature.