Increase conversions by personalizing your User Experience with Recombee AI-powered Algorithms

On this page

In order to provide personalized recommendations, the recommender engine will require data. Integration with Segment provides clean data and organizes the customer events in a meaningful way. 

There are two main types of data needed for AI personalization:

Interactions

User behavioral data such as content views, products purchased, content ratings, etc. You can send these events from Segment using the Recombee Destination.

Catalog of items

Information about the items (content or products) that you want to recommend, such as titles, descriptions, images, categories/genres, or prices.

Once you have this data flowing into the recommender system, the last step is to show the recommendations to your users. To connect Recombee personalization through Segment, follow these 5 steps: 

Step 1: Create your Recombee account

Go to recombee.com and create an instant account. There is a 30-day free trial that will enable you to test our solution. Once you’ve created your account, go to the Settings page to obtain the Database ID and the Private Token (API Key.)

Step 2: Connect your Segment Source to Recombee Destination

Login to the Segment app and refer to the Destinations section. Here click on “Add Destination” and search for “Recombee.” At Recombee’s catalog page, click on “Configure Recombee.” Lastly, pick a Source with behavioral data that you want to be sent to Recombee (views, purchases, etc.)

Now you need to connect this Destination to your Recombee Database by pasting the Database ID and API Key from the previous step to the required fields of the Destination.

If you are using Page events or semantic Ecommerce / Video Track Events, simply enable the Destination and your data will start flowing into Recombee; then you can check that the data is flowing in the Recombee Admin UI.

If you are using your custom Track events, you can set up a mapping that will specify what type of Recombee interaction should be sent for each of your custom event types:

This way, you can send any events to Recombee without implementing changes on your side.

Step 3: Upload catalog data to Recombee

There are two ways to get the catalog data about your items (titles, descriptions, categories, etc.) into Recombee: Using a catalog feed or using API & SDK in a programming language of your choice.

If you go by the first option, navigate to the Catalog Feed section of the Recombee Admin UI, add a new feed of the correct type (e.g. Google Merchant, RSS, CSV, or XML feeds are supported), and set the URL of your feed.

Recombee will periodically fetch the feed, keeping the catalog of items up to date.

Step 4: Configure the recommendations

Now you have all the necessary data in Recombee and you can focus on offering the best content or products to your users.

Every point across the customer journey where you show the recommendations is called a “Scenario” in Recombee.

Go to the Scenario’s section of the Recombee Admin UI and create a new Scenario.

A Scenario is of a particular type, e.g.:

• Recommend Items To User: Recommends the best content/products for a particular user, taking into account the user’s interactions.

• Recommend Items To Item: Recommends related items to one particular item.

• Search Items: Performs a full-text personalized search. The results are based on a provided search query and also on the user’s interactions.

After picking the type of your choice, you can further customize your Scenario by setting Logic (desired behavior of the recommendation model), Filters, and Boosters.

You can use the Preview feature to check the recommendation settings.

Step 5: Show the recommendations to your users

When you are satisfied with the produced recommendations, it is time to show them to your users. You can request the recommendations using the Recombee API & SDKs, or create an HTML Widget in a visual editor of the Recombee Admin UI.

For the latter option, navigate to the Integration tab of the Scenario screen and click on the “Create Widget” button. This will take you to the visual editor where you can edit how the box with recommendations will look.

Once you’re happy with the design, click on the Embed Code button and choose the Segment tab. With this, the ID of the user who is getting the recommendations will be taken from the Segment analytics.js integration, ensuring that the IDs of users are in the widget and coming into the Recombee Destination match.

Finally, copy the embed code into the HTML code of your website where you want to show the recommendations.

Congratulations! You have just added the first personalized recommendations to your website.

There are many places on your site or app that can be personalized. You can, for example, add a Just for you box to your homepage, add the Read Next / Watch Next section to the end of an article or video, or even personalize the offers of an emailing campaign.

Wrapping up

The Recombee Segment Destination is usable across various verticals, such as e-commerce, media, news, and other domains that have a large offering of products or content on their site or app. With just a few clicks, you can improve the user experience across multiple touchpoints, improve conversions, increase click-through rates, or achieve other KPIs.

Here’s what we’ve covered in this recipe:

• Created an account in Recombee

• Connected a Source to the Recombee Destination to get interactions (behavioral events) into Recombee

• Set up a catalog feed in the Recombee UI to provide information about the items that will be recommended

• Configured the recommender to give the correct kind of recommendations for the particular use-case

• Showed the recommendations to the users at our website using a copy & paste embed code

If you have any more questions on how to use this recipe or gain the most of your Segment Recombee experience, do not hesitate to reach out to us at business@recombee.com. Our team of specialists will be happy to explain the specific business use cases or help with the integration and answer other technical queries.