This article helps you:
Determine if recommendations are a good fit for your organization
Understand the use-cases best suited to recommendations
Understand the data requirements for recommendations
Once you’ve identified a predictive goal for your users, the next step is making the recommendations that are most likely to drive users to reach it. Amplitude’s AutoML determines which items are most likely to maximize each user’s predictive goal, and then places those items in front of the user.
Amplitude Audiences's machine learning algorithm clusters your selected users into groups of similar users. This similarity is determined by shared user properties and behaviors taken in the past. Next, it analyzes historical data to see which items are most likely to increase each cluster’s propensity to convert. Finally, it assigns a ranked list of items to each user, based on their respective assigned cluster.
The algorithm re-trains every hour, so it’s always incorporating new information on properties and behaviors into its results.
Recommendations are available for standard event properties only. They are not available for merged, derived, or transformed properties.
Amplitude Audiences is optimized for user-based personalization, not account-based. As such, its recommendations will be most useful for companies that need to showcase an array of items—products, articles, shows—in some kind of product carousel, product list, or cart flow. In particular, ecommerce and marketplace companies, as well as B2C and subscription software companies, are the best fit for Amplitude Audiences.
Enterprise B2B companies, on the other hand, are unlikely to benefit from using recommendations.
It’s important to keep in mind that Amplitude Audiences is not an analytics feature; instead, it’s a personalization feature that helps you improve in-product / digital experiences to maximize lift. Its recommendations are optimized for user-based digital commerce use cases, and are most effective for three types of personalization: Assortment, next-best action, and cross sell.
Support for other use cases—like in-session recommendations and new item recommendations—is currently in development.
Recommendations are only available to Amplitude Audiences customers.
There are three data components to configuring a recommendation: the outcome event, the exposure event, and the event property. The data behind these components must be instrumented] in your taxonomy for recommendations to work:
It’s recommended that you work closely with your Amplitude CSM to ensure these conditions are met.
To learn more, read on to find out how to build a recommendation and how to use recommendations in your personalization campaigns.
Thanks for your feedback!
May 7th, 2024
Need help? Contact Support
Visit Amplitude.com
Have a look at the Amplitude Blog
Learn more at Amplitude Academy
© 2024 Amplitude, Inc. All rights reserved. Amplitude is a registered trademark of Amplitude, Inc.