How Esty Uses Personalization to Keep Commerce Human

With well over 80 million users, marketplace platform Etsy relies on a sophisticated blend of artificial intelligence (AI), machine learning (ML) and advanced analytics to connect buyers with relevant sellers.

One of the prominent examples of this is their personalization engine, an ML algorithm that recommends products to customers based on their past search history, browsing behavior and various contextual clues (i.e. time of year, location, etc.). As explained in a very detailed post on the company blog:

“When a user logs into the marketplace and searches for items, they signal their preferences by interacting with listings that pique their interest. In personalization, our algorithms train on these signals and learn to predict, per user, the most relevant listings. The resulting personalized model lets individuals’ taste shine through in their search results without compromising performance. Personalization enhances the underlying search model by customizing the ordering of relevant items according to user preference. Using a combination of historical and contextual features, the search ranking model learns to recognize which items have a greater alignment with an individual’s taste.”

As for next steps, according to a recent study released by Etsy earlier this year, they’re currently developing a new production ranking system - the 𝑂nline 𝑃ersonalized 𝐴ttribute-based 𝑅e-ranker (OPAR), “a light-weight, within-session personalization approach using multi-arm bandits (MAB).”

Using Data to Help Customers “Think Outside the Bun”

In early 2020, Tex-mex food franchise Taco Bell announced its partnership with AI technology provider Certona to create personalized ordering experiences. 

Embedded into the Taco Bell ordering app, this new tool uses ML and AI technology to analyze the users individual preferences, past dining history, location, weather and restaurant specific menus and pricing. Based on these insights, the application will show users the most relevant menu items, promotions and content. 

Balancing Personalization with Privacy at Apple

Apple uses Federated Learning (FL) - a distributed ML approach that enables training on a large corpus of decentralised device data - to enhance the user experience. For example, Apple uses FL to improve predictive keyboard and error correction capabilities. It also used to recommend news articles and further personalize Siri.

One of the things that makes Apple’s personalization approach unique is that that it processes data on the device itself instead of servers to ensure user privacy - a key issue for Apple these days. 

According to “How Apple Tuned Up Federated Learning For Its iPhones,” Apple’s news personalization engine works by: 

• Derive ground truth and on-device evaluation for news personalization, from user interactions with news content.
• Store information on-device for articles such as: tap and read is a positive label and tap and unread is a negative label.
• The system attributes a range of values for each parameter.
• During tuning task execution, the plug-in runs a randomized grid search and randomly generates configurations.
• These configurations are applied to the personalization algorithm which predicts the likelihood of a user reading an article.
• The predictions are then compared with the ground truth labels to calculate a prediction loss for each randomly generated configuration.

Become a Member of the AI, Data & Analytics Network TODAY!