Brand-level Ranking in Ecommerce: New Model from Alibaba Research

Image for post
Image for post

This article is part of the Academic Alibaba series and is taken from the paper entitled “A Brand-level Ranking System with the Customized Attention-GRU Model” by Yu Zhu, Junxiong Zhu, Jie Hou, Yongliang Li, Beidou Wang, Ziyu Guan, and Deng Cai, accepted by IJCAI 2018. The full paper can be read here.

On e-commerce websites like Alibaba’s Taobao, brand is playing an increasingly important role in influencing users’ click/purchase decisions, partly because users are now attaching more importance to the quality of products and brand is an indicator of quality. However, existing ranking systems are not specifically designed to satisfy this kind of demand. Some design tricks can alleviate this problem, but they either provide less-than-satisfactory results or create additional interaction cost, such as filtering search results with check boxes.

Now, the Alibaba tech team, in collaboration with researchers from Zhejiang University and Northwest University of China, has designed an alternative approach in the form of a brand-level ranking system. This system accurately deduces which brands the user prefers and gives them prominence in search results, without the burden of excessive manual interactions.

The Importance of Brand

There are numerous reasons why brand is having a growing impact on customers’ click/purchase decisions. On one hand, users may prefer items with high quality, and brand is a good indicator of product quality. On the other hand, users may prefer certain brands because of the brand image built from marketing and campaigns. For example, basketball fans may prefer clothes under NBA-star-endorsed brands.

As in the examples shown below, when using a traditional ranking system, most retrieval/recommender systems aggregate the items of different brands together. This does not satisfy users with strong brand preferences, since they have to waste time browsing through large numbers of products from other brands. Users can choose to only display Levi’s brand products by selecting the Levi’s checkbox. However, only a few brands can be displayed here and there is no personalization. Moreover, users must check multiple boxes to browse several brands, increasing the interaction cost and creating negative user experience.

The research team developed its new brand-level ranking system by conducting feature engineering specifically tailored for the personalized brand ranking problem, and then ranking the brands with an adapted Attention-GRU model. With this system, items of the same brand are grouped together and brand groups are ranked based on users’ brand preferences. In the example below, jeans are first grouped by brands, e.g. Levi’s and Wrangler. The system will then monitor the user’s activity and change the brand rankings when it learns that the user prefers one brand over another.

Personalized Brand Ranking

At the core of the brand-level ranking system is a personalized approach to brand ranking. Different user actions (e.g., clicks and purchases) reflect different levels of user preferences towards items and their brands. For example, making a purchase normally indicates a higher level of interest by the user than simply clicking on the item. The key challenge was how to exploit these characteristics of user action sequences for brand ranking.

The team’s Attention-GRU-based system formulates this as a point-wise ranking problem. Specifically, a classifier is trained based on the various information on the e-commerce website. Then given a user, his or her probability of preferring each brand is predicted by the classifier, and brands are ranked based on their probabilities. Items within the same brand group can be ranked by traditional ranking methods.

The team conducted extensive offline and online testing of this system on a large-scale e-commerce platform. The test results and user feedback both indicate the higher effectiveness of their adapted model and the brand-level ranking system in comparison to traditional approaches.

The full paper can be read here.

Alibaba Tech

First hand and in-depth information about Alibaba’s latest technology → Facebook: “Alibaba Tech”. Twitter: “AlibabaTech”.

Written by

First-hand & in-depth information about Alibaba's tech innovation in Artificial Intelligence, Big Data & Computer Engineering. Follow us on Facebook!

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store