Resources Administration & Configuration


Show what’s relevant to each user profile. Personalization learns the affinity of each user and boosts results that match the user profile.

Why Personalization?

Algolia Personalization works by learning about your customers’ preferences and then weighting products that fit those preferences more heavily when they search or browse. It does this by looking at the attributes of products the customer clicks on or adds to their cart and building an anonymous profile. For example if a customer buys purple leggings for yoga, if they search for a t-shirt Algolia will promote purple t-shirt suitable for yoga.

For this to work well your product data needs to have attributes that might match well with the customer preferences, like color, movie genre, author etc. Personalization works best on large product ranges that are in a particular niche like ladies shoes, books, video games and the like.

With diverse product ranges preferences can be harder to define; a customer’s purchase of soft drinks doesn’t tell you much about what kind of sofa they will choose. It also helps if your customers are regular visitors so the profile can be built up.

Collecting user data for Personalization

  • Plan, which events to track.
  • Send click and conversion events with user tokens to the Algolia Insights API. The user token allows Algolia to connect pseudonymous user profiles with actions performed with Algolia, for example, search queries.

NOTE: Sending Events to Algolia requires development work, check with your development team to ensure these are set up in place. 

Configuring your personalization strategy

  1. Go to the Personalization section in the Algolia dashboard.
  2. Select the Strategy tab.
    Personalization main
  3. Give a score to each event and facet to weigh their importance for the personalization strategy. Events and facets with higher weights have a larger impact on Personalization. When weighing events there is no hard and fast rules, but try to think about how indicative of each action might be of customer's preference. For example, a purchase will be a better signal than a click and adding to a wishlist might be between the two.
  4. Set the impact of Personalization on the search results between 0 and 100. The higher the score, the more a search result, which matches the personalization profile, is boosted.
  5. You can test your Personalization strategy with the Personalization Simulator. Select a user token, an index, and enter a search query to see, which search results rank higher or lower due to Personalization.
  6. Select the User tab to view each user profile and behavior to get more insights.

NOTE: Before you enable Personalistion across the board we highly recommend A/B testing it.