Infinite Analytics


Wanna Grab a Cold One? Eliminate the Cold Start Problem with IA’s Personalized Recommendations

by Akash Bhatia, February 8, 2017

Few things match the appeal of a chilled ale on a sunny day. Fewer still match the agony of watching a prospective user get cold feet. With Infinite Analytics’ IA Recommend we address the cold start problem through intelligent personalization and matching of users’ tastes with real products!

The cold start problem is the Darth Vader for all the e-Commerce Jedis. It is usually characterized as a problem of cold-start items or of cold start users. The cold start items problem is caused by new items that do not have enough ratings or other user interactions to be properly recommended to users. The cold start user problem occurs when a new user’s tastes are unknown, thus making it difficult to present her with options that match her taste.

Personalized recommendation must overcome this problem; however, the two common recommendation systems — collaborative filtering and content-based recommendations — have limited utility. Content-based systems use metadata of the products being promoted. The question then is which metadata are important? Collaborative filtering, at least in its naïve form, doesn’t care about the metadata, it just uses people’s ratings or behavior to make a recommendation. The problem with collaborative filtering is that you need data on ratings and behavior.

We, at Infinite Analytics, use a hybrid approach. Through a combination of social analytics and NLP we provide retailers with the most advanced analysis of a user — describing them based on various attributes including but not limited to their likes, interests, profession, sentiments in their posts and others. Cutting edge predictive analytics then match what the user is looking to purchase with the most relevant products in the retailer’s product catalog to provide personalized recommendations to the user. By combining data sources and machine learning approaches, we achieve a greater probability of predicting what the user would like to look at and eventually buy, regardless of how new the users or products are. Through IA Recommend, the same search string yields different results depending on the user’s interests.

Such a hyper-personalized experience helps not only in engaging user interest and driving traffic, but in reducing the time spent by users looking for products and hence increasing the propensity to convert cold starts to hot pursuits!

We talk through experience. Selling paintings online is a challenge, especially those of lesser known artists. Through personalized recommendation solutions IA helped increase engagement and conversions for a client. This led to a 33% reduction in the average number of clicks until purchase (120 to 80). Given peoples’ perennial scarcity of time, it was a massive improvement in user experience.

Don’t turn to the Dark Side, get your light saber today! Reach out to us on for a demo.

Leave a comment

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