Press "Enter" to skip to content

How to enrich metadata for products in your catalog (Artificial Intelligence in eCommerce Part 1)

The gamut of Artificial intelligence technologies such as image recognition, natural language processing, behavior based learning etc. can tremendously benefit as well as change how traditional merchant operations for an eCommerce websites are performed. There are many areas where AI has the ability to positively impact merchandising. My objective in this series of posts is to cover as many areas as I can however for this post (PART 1) I will largely focus on how AI can benefit merchants in creating and maintaining accurate and fresh product content which typically includes product tags, descriptions, features etc.

eCommerce merchants or anyone who manages a product catalog on an eCommerce website understands how critical it is to maintain and display relevant product listings and high quality shopper friendly product content. Maintaining fresh product listings, high resolution product images, accurate product descriptions, competitive product features etc. across thousands of products is where merchant teams spend most of there time and money.

Traditionally merchants have relied on manual quality checks, semi-automated methods, pre-configured rules, inflexible relevance algorithms etc.  to ensure relevant product listings, accurate product descriptions and comprehensive product attributes.  Manual methods work, however they are expensive and time consuming. In most cases their application is limited to best-sellers and top categories. Fully automated methods may address few challenges but are also limited due to rigidness of rules and their inability to keep pace with fast changing data. Above all these systems do not understand product domains, do not self-learn and mostly rely on inputs from the merchants to make decisions and take actions.

Creating, maintaining and ensuring high quality product content is a complex business. “You don’t know what you don’t know”.  In this case what it means is that proactively figuring out if the product content is accurate and of high quality is a tricky task.  Merchants generally deal with manufacturer’s content and do not act till a question is asked, a shopper makes a negative comment or conversion and engagement rate for a product is abysmally low. These steps are reactive in nature and generally become an operational nightmare for even a medium sized eCommerce website. Traditional tools even  lack the ability to integrate actions with real time insights harnessed from vast amounts of user data, and social product content.

That’s where AI capabilities such as image feature extraction, natural language processing, sentiment analysis come into play and fare rather well in addressing day to day challenges faced by eCommerce merchants.

Imagine a system which understands images (just like a merchant would) and enumerates all visual features for that product OR a system which evaluates thousands of user generated product images and predict pairing options or usage scenarios for a product OR one which can weed out low quality low performing images.


Embedded Image recognition and feature detection in merchandising tools can generate powerful insights and present information faster without the need of manually evaluating the image content. Image recognition models can be trained to automate feature extraction and translate them into product tags and features, ensuring everything of importance in a product image is translated for search and SEO. Its like modeling a merchants visual analysis in a machine model.


The potential of artificial intelligence in eCommerce merchandising operations is unlimited. I have covered but a tiny fraction of the possibilities. I will continue to share more in upcoming parts of this blog series however for now, merchandisers looking for “what’s next” should take the leap in 2017. They should initiate planning and implementing systems with AI capabilities at its core to make there websites outshine those of competitors

Be First to Comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Social media & sharing icons powered by UltimatelySocial

Enjoyed reading? Share it with your friends :)