From intent to content - Site Search and Product Discovery

E-commerce sites are getting intuitive by the day. There's talk of 1:1 personalization, forays made into natural language processing, BI and AI thrown in for good measure and all of that makes for a "smarter" shopping experience. Or is it...Read on.

Sometime in the 1920's, Lucene came up with the idea of document search by means of term frequency. Later the terms were indexed based on their rarity and thus was born the Lucene algorithm. Today, most site search solutions have a come a long way from that, but not enough to actually have a perceptible change for the shopper. Most of the search algorithms actually struggle with long-tailed queries and sputter at the thought of machine learning. What's good on day 1 is expected to be good on day 27. Meanwhile, the catalog changes, seasonal demand changes, shopper profile changes - inherently selling and buying both change.

So what's the answer, what is the Holy Grail of product discovery on e-commerce websites. Some may say the answer lies in the question. However prophetic may that sound, there is an element of truth in there somewhere.

Product discovery solutions have to A. cover the whole array of search query types. So if it's a relational search, descriptive search, non-product search or even a feature search - it has to be made sense of and relevant results delivered accordingly, B. learn on the go as the query terms change - I'm talking aliases, stemming etc. and C. provide enough feedback for the merchandiser to step in as and when needed. A great merchandising tool works not only to up-sell/cross-sell products but also to fine tune what a seller would like to periodically showcase.

Yes these are the basics. But it's a start. More on e-commerce product discovery in the days to come.