Toys ’R’ Us announced earlier this month that they were closing their doors after 70 years in business, ending an era of timely Christmas bargains and birthday toy-store visit traditions. Amidst last minute scrambles for discounts, nostalgic videos of how it all began, and discussions over what’s left to salvage, the story remains the same — online beat retail yet again. We’ve heard the diatribe about innovation disrupting normalcy and changing how things are done in a hundred different contexts before. The industrial revolution made mass manufactured goods accessible and affordable for a huge populace that was waiting to consume. It gave way to a better way of doing things, lead to the technological revolution, to abundance and eventually lead to the luxury of choice. 
A search algorithm should be able to surface a 40-inch television to anybody searching for it, irrespective of whether they look for a 40’ television, a 40-inch television or a 40 in television. Feature extraction allows algorithms to break down search queries to its components and assign weights to each attribute contained in it. This allows the algorithm to figure out which attributes in the query are most important and so must necessarily be present in the search results while which attributes can be placed lower in the list of priorities. A smart site search is capable of understanding the intent behind a what the shopper is looking for when they enter a query and offer relevant suggestions to the shopper. A lot like the satisfying experience of being handed the right product or a relevant variant when you ask for something specific in a store.
The Tides are Turning
Dwindling sales and smaller profit margins at Victoria’s Secret aren’t much of a secret anymore. Sales during the last holiday season fell by 4% in comparison to the same period the previous year. The problem? Women just don’t relate to the brand anymore. The last few decades have seen a seismic shift in priorities and ways of thinking. As a result, the brand’s mantra of putting sexy over practicality is increasingly becoming obsolete. Evolution and change are the only constant that mark every shift in commerce. The ones that survive are those who know how to ride the wave rather than collapse under its weight.Staying Relevant in the Face of Change
What the stories of Toys ‘R’ Us and Victoria’s Secret symbolize, in varying degrees of severity, is how the inability to change as per the shifting priorities of the consumer impacts the biggest of companies and the most proliferous of brands. Change affects them all and evolve they must. In the context of eCommerce, the story of evolution begins with the introduction of basic search. Search, in its most rudimentary form, involved picking up a query and matching it to existing products with the same text description from an online catalog. Thankfully, we’ve come a long way from there. Semantic search, as it exists today, involves understanding user intent when they visit a site. And understanding what people are looking for when they type out a query on a site’s search bar involves a lot more than a basic text match.Understanding the Context of Search
A sophisticated site search algorithm understands shopper queries in the same way a human would. Relevance, in search, is the ability to accurately map queries to products that match it. This remains, at its core, the most fundamental need that all eCommerce businesses strive to solve in different ways. Quite understandably, that makes it their biggest challenge as well. Solving for relevance involves a number of important factors. Here are the top five elements you should incorporate into your site search to fine tune relevance on your site.Specification handling
Understanding the parts of a search query to identify specifications or attributes of the product are an important part of intelligent search. A search algorithm should be able to understand that in a query such as “iPhone X cases”, the shopper is looking for a specific accessory suitable for a particular product subtype. Though it sounds like quite a basic ask, many sites still fail to incorporate specification handling in their site search.Interpreting Dimensions
Different strokes for different folks. No place this phrase is more relevant than in online commerce. Especially for hardware and electronics where different geographies have their preferred scales of measurement, it becomes imperative that sites are equipped to handle dimensions intelligently.
A search algorithm should be able to surface a 40-inch television to anybody searching for it, irrespective of whether they look for a 40’ television, a 40-inch television or a 40 in television. Feature extraction
When you ask for a blue cashmere sweater at a store, it doesn’t take too long for them to understand that you’re talking about a very expensive but extremely gorgeous piece of winter clothing in a particular color. Those are the simple nuances of offline shopping that online misses out on.
Feature extraction allows algorithms to break down search queries to its components and assign weights to each attribute contained in it. This allows the algorithm to figure out which attributes in the query are most important and so must necessarily be present in the search results while which attributes can be placed lower in the list of priorities. Relational queries
As we get deeper into making search algorithms think like their human counterparts, it is essential that they decode the relationship between words in a query rather than treat them as individual planets floating in isolation. This is all the more reason why the old text-match method doesn’t work anymore because thanks to Google, shoppers are looking for almost anything online and they’re typing it like they’re thinking it! Understanding that when a shopper looks for ‘evening dresses for parties’, or a ‘round bedside table lamp’, they are describing features additional to the basic product itself is important in showcasing the most relevant results to the shopper.Phrase identification
The final test of relevance lies in the ability of identifying phrases in a search query, irrespective of length or complexity, and tie them together while presenting results to the shopper. This is especially true for search-intensive shoppers who come to the site looking for something very specific. For example, a shopper looking for a replacement headlight for his car will go straight for the search bar and type out the requirement, rather than go through the navigation to get there.
A smart site search is capable of understanding the intent behind a what the shopper is looking for when they enter a query and offer relevant suggestions to the shopper. A lot like the satisfying experience of being handed the right product or a relevant variant when you ask for something specific in a store. 
