The First Self-Serve Personalization Engine: Amplitude Recommend

Bilal Mahmood

Head of Product

people reacted
6-minute Read,

Posted on April 25, 2021

With this new product, marketing and product teams can customize the digital experience for each customer—all within minutes.

Amplitude Recommend: The First Self-serve Personalization Engine

Today we’re launching Amplitude Recommend, a new personalization product that gives all digital companies the power to produce personalized experiences at scale; accelerating your roadmap by years, increasing conversion lift—often by double digits—and reducing the cost by millions.

Personalization in the style of Netflix or Amazon is the dream of every company—optimizing the digital experience for each user so that it feels like it’s built just for them. But most companies face a significant barrier to entry when attempting to automate these 1:1 experiences. It requires sophisticated identity resolution to reach the right user, machine learning to find the right message, and real-time delivery to identify the right time. Personalization at scale requires product, marketing, and engineering to collaborate.

This can take years of investment and millions in development cost. Digital companies in turn face the tradeoff of locking up resources for months at a time or losing out on millions in revenue.

With Amplitude Recommend, that tradeoff is no more.

A New Solution For a New Age

Recommend is a new product in the Amplitude Digital Optimization System. Powered by first-party data collected via the Amplitude Behavioral Graph, you can now automate an end-to-end personalization workflow in minutes. Recommend alleviates the technical burden of personalization so that digital experience owners—marketers, product managers and growth teams—can manage it directly.

How Amplitude Recommend Works: Cohorts, Predictions, Recommendations and Computations

This self-serve platform guides you through each step of personalization—from mapping the right user to the right message at the right time, via a collection of three new sets of functionality:

  • Segmentation: Enabling marketers to find the best customers to target via Cohorts and Computations
  • Recommendation: Enabling product teams to automate the next best message or content for those customers to receive via Predictions and Recommendations
  • Delivery: Enabling anyone to export their segments and recommendations into any digital channel via APIs and Syncs

Self-serve Segmentation to Find the Right Customer

Amplitude Recommend Computations

The first step of any personalization initiative is identifying the right customer to target. Amplitude Recommend provides two sets of functionality to build lists of users to synchronize into your downstream digital channels: Cohorts and Computations.

Cohorts are the core form of segmentation. With Amplitude Recommend, you can create clusters of users grouped together by any events they have performed in the past, like adding to the cart over the last 24 hours, or starting a subscription. This is all done through a self-serve interface, with no SQL or code required. And because Recommend is part of the Amplitude Digital Optimization System, any cohorts created in Amplitude Analytics are immediately available in Recommend—and vice versa.

Computations are an advanced form of segmentation. With Amplitude Recommend, you can convert behaviors into aggregate user properties over time, for more sophisticated filtering. For example, in a matter of seconds and without any data engineering, you can count the number of times users perform an add-to-cart event in the last 24 hours, or aggregate the average order value over the last 30 days.

Combined, both Cohorts and Computations allow you to identify behavioral segments to trigger engagement-based marketing conditions.

Automated Recommendations to Find the Right Message

Amplitude Recommend Recommendations

For many, a “personalized” experience has meant inserting a user’s name in the home screen or swapping out a picture based on their location. While that type of creative, demographic personalization is encouraged, the impact is limited. Products that fully adapt to each individual user based on the past and predicted future behaviors is what realizing the full potential of personalization looks like; a dynamic product experience that is now possible with the Recommendations feature in Amplitude Recommend.

Powered by Amplitude’s AutoML system, the new Recommendations capabilities enable you to create personalized experiences just like Netflix or Amazon. Once the right user has been identified, Amplitude Recommend determines the right assortment of content, products, and messages that are going to increase the likelihood to convert.

 

From a self-serve user interface, select any event property—an SKU, item name, product category, etc.—and an outcome you want to optimize (purchase, subscription, etc.). Amplitude Recommend will automatically rank all potential values for the property by their likelihood to increase the outcome, personalized to the individual user.

In minutes, Amplitude Recommend will generate a ranked list of what each user is likely to prefer, up to 100 at a time, in the order that will maximize likelihood to convert. These can be ranked item SKUs by likelihood to be added to cart for assortment, or ranked product categories by likelihood for cross-sell. Instead of three to five different experiences, the system produces millions of permutations of potential experiences, personalized to each individual user. All of this is completed within minutes, without any need to request help from a data science team.

The consumer arm of a top-15 U.S. bank uses Recommendations to improve financial literacy with its customers. Now when users enter the mobile app, they are shown content that matches their bank relationship and their behavior. The article they see is the result of an individualized suggestion powered by Amplitude Recommend. Since rolling out recommendations, the bank has seen a 15% increase in engagement.

Real-time APIs to Find the Right Time

Amplitude Recommend Sync Recommendations

The last step of any personalization workflow is delivery. You’ve identified the right message to target, but need to deliver those messages to the user.

Amplitude Recommend now offers a real-time API and Syncs to connect your cohorts, computed properties, and recommendations into any digital channel.

With Syncs, you can synchronize data objects to any ad, email, or experimentation platform. Sync cohorts to Facebook or Marketo, so when user behaviors change it automatically synchronizes campaigns in the respective ad and customer engagement destination. For example, you can sync computed properties to engagement platforms, like Braze, so when a user’s average order value changes, it automatically adjusts the attribute in the respective email campaign. This all happens with no custom data engineering pipeline—it all works from a click of a button.

With the Profile API, customers can now query a REST API endpoint for any user, and return user data from Amplitude. As a user visits your site or app, simply query the Profile API by users’ unique ID, and return their list of properties, cohorts, and recommendations. Embed that response directly back into your product, and adjust the product experience based on their recommended properties.

Jumbo Interactive, the digital platform behind Oz Lotteries, uses the Profile API via Braze to deliver Amazon-style recommendations in emails and push notifications. After customers make a purchase, they receive a follow-up communication suggesting another game they may be interested in based on their purchase history and behavior patterns. Recommend delivers up to four suggestions for email and a single suggestion for push. The result has been a massive lift that has improved checkout conversion originating from these messages by over 158%.

Built-in Measurement

Amplitude Recommend Automated Lift Analysis

With Amplitude, measuring the impact of personalized experiences you create is easy. Every cohort you build in Amplitude Recommend is available for analysis in Amplitude Analytics. That means understanding the impact of your campaign is as simple as building a chart—no need to recreate the audience or duplicate any work.

For Recommendations, we’ve even taken measurement capabilities a step further, with Automated Lift Analysis. In each recommendation you create, you select the percentage of users that will be part of the control: anywhere from 0-100%. Then, when you query the Profile API for a recommendation, Amplitude assigns users the control or delivers a recommendation. At the same time, we automatically log an impression event to track the uplift that Recommendations have on your existing experience. All you have to do is click on the Performance tab in the recommendation to see it.

With one unified system for analytics and personalization, all built on a unified dataset, it’s easy to close the loop on experiences you create.

Personalization, in Minutes

The power of Amplitude Recommend is that it enables a complete feedback loop of the Digital Optimization System.

From the click of a button, you can create Cohorts and Recommendations in Amplitude Recommend, synchronize them to all of your ads, email, and in-app experiences, and monitor the performance of those campaigns back in Amplitude Analytics.

The automation of this process bridges product and marketing teams to reduce time to launch. The integrated machine learning reduces the cost of engineering and enhances the accuracy of the recommendations. And complementary analytics enables you to reliably measure the impact of your personalized experiments.

See a live demo to learn what Amplitude Recommend can do for your business.

Bilal Mahmood

Bilal Mahmood is the Head of Product at Amplitude, leading development on personalization and machine learning enabled initiatives. He was previously cofounder of ClearBrain (acquired by Amplitude), and studied biology and economics at Stanford and the University of Cambridge.

More from Bilal

Inside Amplitude