The only time I used Etsy, I was looking for something pretty specific: a red scarf. I just wanted a snazzy, simple option for autumn days. I figured that if my purchase could support a real craftsperson, even better.
Etsy’s search engine had other ideas. The results were surprisingly jumbled: women’s headscarves, men’s scarfs that were not actually red, and a variety of items that weren’t even scarf related. Eventually I gave up and ordered from a different site.
If I were to try the same search again today, however, I might not have to leave empty handed. Etsy just launched a brand new search feature, that’s much more user friendly—with results categorized by hand.
The New Etsy
“Red scarf” isn’t highly specific, but it’s actually a lot more particular than most Etsy searches. About a third of the site’s search strings are extremely broad, like just “scarf” or “jewelry.” Etsy realized that if they could tease out more user intention, they’d be able to return better results and less potential customers would bounce like I did.
One simple way to do that is to improve the categories that items are filed under—and make them more prominent. Now, a search for “jewelry” will return regular results, but with specific categories highlighted, like “necklaces,” “brooches” and “body jewelry.” These categories are presented above the organic results, in a photo-rich grid layout that makes them temptingly clickable. If you have something specific in mind, this will pull you down the right rabbit hole.
The Human Touch
What’s surprising about Etsy’s new category system is that it was categorized by live human beings. Etsy hired a library scientist to spearhead the categorization, and asked sellers to help.
To say this is unusual in 2015 is an understatement. At the turn of the millennium, it was a point of pride that Yahoo search results were categorized by real humans—this set the site apart as “better” from competitors. Since then, however, the combination of big data and better algorithms (together known as “machine learning”) have proven they can outperform human beings. No major search engine uses human categorization.
But Etsy didn’t have the luxury of machine learning. The company says they actually tried it first and got bad results, because they were short on one of the key ingredients: data. Etsy’s existing product listings weren’t labeled in a consistent enough way for a machine to learn from them and scale up the process. They had to start from scratch, with old fashioned human ingenuity.
But it sounds like Etsy’s algorithms are learning now. After their librarian established the categories, they asked their sellers to pitch in by categorizing existing listings. Sellers have a natural incentive to do this: better search discoverability means more sales. As the algorithm learns from their hand categorization, it will eventually be able to categorize other listings on its own, and the whole site will run better.
It’s fascinating watching a single e-commerce site go through a microcosm of 15 years of internet history, almost overnight. The new site definitely works better (on both desktop and mobile). I recommend checking it out. And please, let me know if you find that red scarf.