Why Ecommerce Research Is So Bad and What Could Soon Fix It

On the website of one of the largest online retailers of consumer electronics, a search for “beer cooler” returns a bounty of over 500 relevant results, but a search for “beer cooler” does not not even provide one. Type “something to chill the beer” and the only result you get is a Lego set.

Do the same exercise on Google or Bing and the experience is completely different. The two most popular motors seem to understand that “cooler” and “colder” are synonymous, and it even performs quite well on the “something to cool” test.

What do search engine giants know that e-commerce sites don’t? The difference is “vector searcha technology rooted in artificial intelligence research that represents information as numbers rather than text.

Once content is converted into search factors (which are essentially strings of numbers), machine learning algorithms can find similar content by comparing distances between vectors to understand how different words relate to each other. They can also analyze surrounding content to understand the context of search queries, so that “bad company songs” return results to 1980s supergroup tunes and not unwanted guest laments. If you want to dive into vector search technology, this post on the Google Cloud blog should satisfy your inner geek.

When research needs a human touch

However, this is not how most e-commerce search engines work today. “Great search is actually a data and machine learning game, but none of the major search technologies available today do that directly,” said Hamish Ogilvy, CEO of Search.io, which makes a search engine for e-retailers based on vector technology. The bottom line is that “search quality is fundamentally determined by the ability of humans to configure and connect to other systems.”

In other words, the search engines on most commercial sites are only as good as the human beings behind them. Giants like Amazon.com have been able to outsource the hacks needed to deliver relevant results to data science teams over a period of years, but most retailers are stuck with the vendor’s default search engine. of services they use.

Most are not well served by this. A recent survey of search performance of the 50 most profitable e-commerce sites in the United States by the Baymard Institute said the state of e-commerce search was “broken”, noting that only 34% of sites could handle queries that use themes, features, or symptoms rather than specific product names. “A whopping 70% of search engines are unable to return relevant results for product type synonyms, forcing users to search using the exact same lingo as the site,” the company claimed.

It costs sellers a lot of money. A recent Google report estimates that e-commerce businesses lose $300 billion a year in the United States alone because visitors can’t find what they’re looking for.

Adjustments and Unintended Consequences

Traditional search relies on matching text strings, Ogilvy explained. Therefore, a search for “crew necks” will not return a result related to T-shirts unless the relationship is defined by hard-coded rules in the index. To handle a laptop search, for example, the engine needs to know that the words “laptop”, “laptop”, “laptop” and “MacBook” are functionally identical. The manual effort of coding these relationships multiplied by thousands of products each of which can be referenced in multiple ways is almost unimaginably complex.

And hand coding creates its own problems as the number of rules accumulates. Ogilvy cites the example of a company that programmed a workaround that reformatted searches for “USB C” to “USB-C,” which was the syntax it used in its catalog. The unexpected result was that when visitors searched for “USB cable”, the hyphen was automatically appended to the text string and the resulting query – “USB cable” – was empty.

“It’s very difficult to write thousands of these things and not cause trouble,” Ogilvy said.

These limitations had prompted most e-commerce site operators to optimize for the largest queries and effectively drop the 70% of queries that make up the “long tail” of rarely used search terms.

The good news is that the situation will improve in the not too distant future. E-commerce search engine makers are “all jostling for the vector,” Ogilvy said. “That’s how research will be done in the future.”

The question is not whether vector search will become widespread, but when. “I expect almost everyone to go in that direction,” he said. The transition will not necessarily be smooth. As website operators replace their heavily patched search utilities, many rules will need to be removed and some modified, as machine learning is not magic and cannot anticipate the nuances of each case. ‘use. However, in the long run, everyone will be better off. I bet a case of cold beer on it.

Then read this:

Copyright © 2022 IDG Communications, Inc.

Comments are closed.