How do I bring out recommendation in B2B site without using rating as a parameter?

Question

I am doing project on Recommender systems for an ecommerce B2B website, I have even gone through recommender systems course from University of Minnesota from Coursera. I could not implement Recommender Systems in traditional way of recommending things as the website I am working on does not have a rating option or click history. I am finding difficulty here to implement recommendations, as the website does not have any required parameters. And I have worked in R language with sample dataset which contained rating as parameter to compute recommendations but I am facing difficulty to recommend things to website without the parameter rating . Kindly give a solution on how to make recommendations for my B2B website.

Answer 1

I'm am always amazed at how slowly the world seems to notice that ratings are a horrible way to make recommendations. You have a much better way with sales events. No major player in recommendations uses ratings for recs. This started with the Netflix prize, where they defined the judging criteria as ratings. The effect has been so long lasting that Netflix decided not to use ratings (in their own recommender) before the event was even finished. Thought Coursera might be more nimble than Academia at large.

In any case most modern recommenders will handle non-rating based recommendations. If you want one that can ingest many different indicators of user preference take a look at the Universal Recommender. It can use purchases, views, category preferences, even user profile information and is implemented end-to-end.