How and why events drive search ROI

Algolia is the world’s only end-to-end AI search and discovery platform.

That pitch is more than marketing jargon. We built our suite of AI products to power the customer journey from their first visit all the way through to conversion. AI helps us nail the answers to the following four questions which allows us to optimize your customer experience, increase conversion, and improve your ROI:

  1. Why is the user here?
  2. What content is relevant to the user’s query?
  3. In what order should that content be presented to the user?
  4. How can we personalize that content to be more attractive to this particular user?

Each of these questions builds on the previous. Once we understand the user’s intent, we’re able to define search relevance in their context and retrieve the relevant content. Then, we rank and personalize that content for an optimal experience.

How do we go about answering these questions? We tackle these problems with AI.

A rundown of Algolia’s top-of-the-industry AI-powered features

Query Categorization

Displaying relevant results involves understanding the query intent. AI can help us with that! If we define a category hierarchy, such as product categories in an ecommerce store, our Query Categorization AI predicts which category a query is most likely searching in. This allows us to automatically apply filtering and boosting and instantly optimize search results at query time. It also gives us category analytics, giving us the power to identify underperforming queries.

AI Synonyms

Setting up synonyms involves sifting through the queries that return few results and, for each, identifying alternative query terms to search for. Doing this manually is cumbersome and expensive to maintain. But your users give you all the data needed to automate this process: if they search for something and get no results but are determined to find it, the next few searches they make are likely synonyms of their original search query. Our powerful AI does the sifting and analyzing for you, finding synonyms you didn’t even realize needed to be set.


Traditional keyword search has its place, but to get the best results, we complement it with AI-based vector search. NeuralSearch uses large language models (LLM) to understand the semantic meaning of a search and match results based on contextual relevance. It strips out the complexities of human language with a hybrid of keyword and vector search, merges the results and ranks them optimally.

Dynamic Re-Ranking

An optimally ranked results set is one that positions the most relevant results first and the least relevant results last. However, the concept of relevance can change over time. The classic example: users searching for “face mask” before COVID were probably thinking about skincare, but times have changed. You can tackle this problem by manually configuring filters and rules, but it’s impossible to account for all the nuances, many of which aren’t as obvious as in the surgical masks example. So, why not let Dynamic Re-Ranking monitor your users’ behavior and take care of this for you? It saves you the effort and is more effective.


Treating every user the same, conforming to the ideal user profile our marketing team cooked up, isn’t going to get us far. A book store customer searching for “harry” may be after “Harry Potter” novels, or they may be more interested in “Harry Truman” biographies. It’s important to personalize search results based on every user’s preferences and history. This used to be an infeasible task in media such as leaflets, billboards, or printed catalogs. But online, you can serve up a new version of your search results every time, optimized specifically for that particular customer.


Even outside of the context of a search query — for example, when an aimless user browses without clear intent — we can maximize the likelihood that they buy something by analyzing their behavioral cues and how they’re similar to other users who have converted in the past. We can use collaborative filtering and content-based filtering to recommend products, which involves parsing a ton of data and making split-second decisions about new users before the page even loads: a perfect job for AI.

Conversational Commerce

Some people like to shop through conversation, and adding that as an option can increase your user retention dramatically. This is especially true in the context of voice search — asking Alexa about the latest items from your favorite clothing store is nice and all until she starts reading the search results like a screen reader, more concerned with structure and semantics than making the results easy to understand and pick through. Again, that’s where AI comes in. By reinterpreting those results in the context of a conversation, AI can allow your customer to interact with what feels like a personal assistant.

Real stories

Take a look at some of the massive improvements our customers have seen after taking advantage of these AI-powered features:

  • Zenni Optical added onto their typical search Recommend, Dynamic Re-Ranking, Query Suggestions, and an early prototype of Neural Search, and they saw a 44% increase in search traffic, a 34% increase in revenue originating from searches, and a 27% increase in revenue per user session.
  • Lacoste, a global clothing company, implemented Algolia’s AI personalization on their site search and saw +150% revenue originating from searches, +210% search use on desktop and +314% on mobile, +37% conversion rate on desktop and +62% on mobile. On top of that, the personalized content kept their customers’ attention, leading to a 88% drop in bounce rate.
  • Idealist, a job site for non-profits and social programs, found that the AI-powered relevance matching, combined with the decisions they made based on Algolia’s analytics, boosted their conversions by 15%.
  • Teachers Pay Teachers, a platform where teachers can share instructional materials across the world, implemented a host of Algolia’s AI features, making it far easier and quicker for teachers to find what they’re looking for. Now, they’re finding an average of 8 new materials every session.

How can Algolia’s AIs help me increase my ROI?

The answer is simple: implement event tracking.

Events are small pieces of data sent to Algolia’s servers every time your users do something significant, like click on a search result or go through with a transaction. We use these signals to train the above described end-to-end AI search and discovery products and features. You just need to send us event data and our AIs will start learning from your end-users’ behavior as they interact with your website or app.

You’ve seen how these AI features can dramatically increase your CTR (click-through rate), CVR (conversion rate), and ROI (return on investment). But let’s back up — to train those AIs in the first place, you need to be sending events to Algolia for some time. Think of events as the fuel for the machine learning models to perform their magic.

How do I send events to Algolia?

We’re doing our best to make this process straightforward. Here are the steps:

1. Understand which events to track

When it comes to AI, the accuracy boils down to the right input data. It’s about quality, not just quantity — too much can be just as bad as too little. So, ask yourself this: In my business model, what user actions do I care about? Boil them down to their simplest components and define why they matter to you. Here’s a more detailed guide, but in short, your list might look something like this:

  • I care about which search results the users click on. Why? The site can learn what content actually seems appealing to the users based on their queries, which shows how well our search engine is matching users with revenue-earning suggestions (and how to improve it further).
  • I care about users showing approval for products (i.e., through clicking a like button or adding to a wishlist). Why? This demonstrates what products users want both after a search query as well as independent of search queries.
  • I care about conversions (i.e., users buying a product, listening to a song or watching a movie). Why? Because they signal our user journey worked, and they indicate what we did that sealed the deal for the user so we can repeat it or improve on it.

Think about the journey your users will take on your site; it might even help to create a diagram. Here are a few examples so you can visualize it:

Example events in an ecommerce search-to-purchase funnel


Example events in a media browse-to-consume funnel

As you go about this process, the metrics you care about will stand out. We’ve highlighted the three most common categories of events, but your particular use case might merit tracking other ones. That’s the beauty behind Algolia’s design philosophy: it’s all prepackaged and straightforward to set up, but it’s still powerfully flexible. It adapts to you and not the other way around.

2. Use one of our many event ingestion methods

Once you’ve got that plan worked out, it’s just a matter of copying or writing a couple of code snippets into your site. Don’t worry, we’ve made this painless. Or you can send event streams or batches from your backend straight to our Insights API.

But if coding isn’t your expertise, we get it. Sometimes, it’s much simpler to use an external tool that already did the coding for you.

For example:

  • If you’re using an ecommerce platform, check out our ecommerce integrations and how they can help you capture event data. Read more here →
  • If you’re already tracking events through a tag manager, an analytics platform, or a customer data tool, perhaps one of our connectors is more up your alley. Read more here →

We’re always working to improve these ingestion methods and and build new ones — keep your eyes open for some new releases soon! If you’re not the techy type, we’ve still got your back.

3. Validate your events setup and switch on AI-powered features

Once you are sending events for long enough that the AIs can spot reliable patterns in the data, you’ll be able to switch them on in your Algolia Dashboard.

Of course, different AI features require slightly different input data. For example, Personalization works best with stable or permanent user tokens. Go to the Events Hub to check on potential issues with your events setup and advice on how to improve what you’re sending.

And if I want a test drive?

That’s a fair ask. Any big purchase or commitment merits careful research and unbiased discovery first. That’s why we’re proud to offer a free exploratory service to those ecommerce brands who are looking to supercharge their sites with Algolia’s top-of-the-line AI search platform. It’s easier than you’d expect: just click the chat button at the bottom right, get in touch with a team member, and ask for “Algolia Value Engineering”. We’ll arrange for an expert to assess your current situation (using your current product catalog and data collected by a small code snippet you’ll paste on your site) and build a custom demo for you using your own product and event data. We’ll compile all the data we collect into a detailed report highlighting any current pain points and describing exactly what you can do to increase conversions.

Ready to get started? We’re just a click away! Get ahold of us with the chat button at the bottom right.

About the authorsJaden Baptista

Jaden Baptista

Technical Writer
Ben Franz

Ben Franz

Sales Engineering and Product Leader at Algolia

Recommended Articles

Powered by Algolia AI Recommendations

How Algolia uses AI to deliver smarter search

How Algolia uses AI to deliver smarter search

Julien Lemoine

Julien Lemoine

Co-founder & former CTO at Algolia
The (almost) ultimate guide to site search

The (almost) ultimate guide to site search

Ivana Ivanovic

Ivana Ivanovic

Senior Content Strategist
4 questions to ask for relevant search results

4 questions to ask for relevant search results

Jaden Baptista

Jaden Baptista

Technical Writer