Share on facebookShare on linkedinShare on twitterShare by email

From the outset, Algolia has enabled developers to build unique, differentiated digital commerce experiences using its managed search and navigation API. As a consequence, Algolia has been a significant part of our retail customers’ $25B+ yearly revenue.

Now, with the introduction of Algolia Recommend, Algolia further enables developers to unleash the component of the experience that drives the remaining part of the product discovery experience: product recommendations. 

Just like for search and navigation, we believe developers in charge of product recommendations shouldn’t spend their time troubleshooting an infrastructure, maintaining and scaling a low level codebase, or dealing with various vendors with heterogeneous developer experiences, but rather, they should leverage the building blocks that helps them release their creativity, generate unique experiences, and accelerate their businesses’ growth.


In a world where more businesses are going online coupled with the fact that consumers have greater choice and the cost and ease of switching brands is low, it is essential to gain and maintain a competitive advantage by building differentiated experiences and constantly innovating. API building blocks are critical to build this differentiation – due to their significant flexibility, ease of implementation, and how they allow developers and product managers to iterate on the fly.

With Algolia Recommend, Algolia adds a new and powerful device in a developer’s toolkit – so they have keys to augment 100% of retailers’ GMV. (Gross Merchandise Value).

Let’s dive into what makes Algolia Recommend unique…

Machine Learning recommendation models an API call away

Algolia Recommend generates recommendations based on:

  • Your product catalog indexed into Algolia
  • Your users’ interactions on your digital properties

It then relies on machine learning models to power relevant recommendations, starting with the following ones:

Frequently Bought Together

The Frequently Bought Together model recommends items that are often bought together. For a given item, it recommends a list of items based on the conversion events your users perform on your platform.

This model enables you to create crossselling opportunities by showing your shoppers products that complement their current selection.

Related Products

The Related Products model recommends items that are related to each other. For a given item, it returns a list of items based on the clicks and conversion events your users perform on your platform.

This model maximizes conversions and catalog exposure.

The Frequently Bought Together model uses conversion events as we need to detect how complementary products are during a purchase. The Related Products model relies on clicks and conversion because relatedness can be conveyed through purchases as well as browsing: searching, then exploring options.

For both models, the data corresponding to the past 30 days is collected. The result is a matrix, in which columns are userTokens and rows are objectIDs. Each cell represents the number of interactions (click and/or conversions) between a userToken and an objectID. Then, we apply a collaborative filtering algorithm: for each item, we find other items that share similar buying patterns across customers. Items would be considered similar if the same set of users interacted with them. Items would be considered bought together if the same set of users bought them.

The limitation of collaborative filtering, however, is the cold start problem. For items that were recently added to the catalog, there is a lower chance to generate significant amounts of traffic, so they will be less likely to appear in recommendations and their recommendations might not be as good. 

We are working on removing this limitation by integrating some level of item similarity based on their contents (for example, we know that items belonging to the same product category are already somewhat related). In the meantime, we’ve introduced a Fallback strategy that directly leverages product attributes to solve for the cold start problem and lack of events.

Fallback strategy

It is likely that some items in a catalog will not have experienced enough events to allow the Algolia Recommend models to generate relevant recommendations – an example of this is newly listed items. Algolia considered this and added a ‘fallback strategy’ to ensure relevant recommendations are still surfaced using  fallbackParameters

 

More to come!

Algolia will be adding a slew of new recommendation models. Later in the year, Algolia will add ‘Personalized Recommendations’ that will surface additional relevant products based on a shopper’s preferences.

Algolia will also make it easier to generate content-based recommendations using capabilities from  the Algolia Search API, for example Trending Products, New Arrivals, Best Sellers, etc, by packaging them into Algolia Recommend.

Implement in a breeze, while respecting the look and feel of your website or app

An advantage of our API first approach is that you can display the recommendations generated by Algolia Recomend anywhere you need them using 6 lines of codes.

Following the same standards as Algolia Search, Algolia provides front end widgets to ease the display and formatting of recommendations on your front-end, starting with Vanilla JS and React.

“We’ve been impressed by the ease and speed of implementation of Algolia Recommend, which enabled us to go fully in production on hicart.com in only 4 days.”
– Raul Larion, Head of Tech at HiCart

Where the magic happens: total flexibility

Things start getting exciting when you start displaying recommendations across your digital touchpoints: Algolia’s API-first approach combined with front end widgets enables developers to  completely control and fine tune the recommendations returned based on the business’ needs, using Algolia’s facetFilters attribute.

Store-wide filters

Algolia Recommend allows you to apply filters on all the recommendations displayed across your entire store. For instance,  most of the time it makes sense to exclude out of stock products from the recommendations:

But you can go way further than that! Let’s explore some more sophisticated use cases:

Contextualized recommendations

With this filter logic, you can decide to filter recommendations based on the category the shoppers are currently exploring, or even specific attributes of the products they’re currently viewing or have in their basket. Here are some examples:

Show only items of the same color than the displayed item:

Show only items of the same category than the displayed item:

Show only items on which you make more margins than the displayed item:

Localised recommendations

One Algolia Recommend user is an off and online retailer based in California. Their network of brick and mortar stores is a real strength for them, so they leverage it as much as possible in their online experience. When you browse their website, you can pick your preferred location. Once you do that, the recommendations will be personalized to showcase only the products available in your preferred store.

Visibility

Algolia Recommend comes with a simulator so you can assess the quality of the recommendations generated by the models before implementing them in production. 

The advantages of a unified platform for your entire experience

A key advantage of Algolia Recommend is that it relies on the same fundations than Algolia Search, and has been developed with the same standards, bringing your developers and your business various advantages:

Do not duplicate implementation efforts

Leveraging Algolia Search + Recommend to power Search, Navigation and Recommendations, you only have to create one indexing integration for all your product catalog, and send your user events to one API.

Cohesive experience for your users

As the index Algolia Search and Algolia Recommend rely on is the same, your users will see the same product information, with the same freshness, across your entire online store. 

One developer experience

Algolia Recommend is built with the same developing standards and developer experience excellence in mind as Algolia Search, meaning developers will benefit from the same implementation principles, and the same quality of developer tools and documentation to build your entire Search, Navigation and Recommendation experience.

Scalable and reliable

Algolia Recommend is built by the same team and on the same foundations as Algolia Search, powering 100B+ requests per month for more than 10,000 customers. So you can focus on creating the experience and don’t have to worry your recommendation engine will break during Black Friday.

Index once, deliver everywhere. Get started today

Combined with Algolia Search, Algolia Recommend allows you to index your catalog in one place, and deliver it everywhere you need it. Our customers are already seeing great results leveraging it on their digital stores:

“Orange Romania has been using Algolia Recommend technology to retain and convert shoppers landing on out-of-stock products. By recommending them relevant products, we unlocked 8% additional revenue on our online store.”

– Florin Spataru, Digital Marketing Manager at Orange România

“We are already seeing a significant increase in product page views per session since implementing it, and are foreseeing a great impact on our revenue.“

– Raul Larion, Head of Tech at HiCart

 

If you are ready to test Algolia Recommend, you can signup today to Algolia.com or log into your account and benefit from 10,000 free Recommend API Requests per month. You will find everything you need to implement it in our documentation.

If you want to learn more about Algolia Recommend, do not hesitate to contact your Customer Success Manager our contact our team by going on algolia.com/contactus.

Share on facebookShare on linkedinShare on twitterShare by email
About the author

Loading amazing content…

Subscribe to the blog updates
Thank you for subscribing, stay tuned!