A third of all searches on Etsy generate over 10,000 items each, but over 80% of purchases via search come from the first page of results. (Not surprising, since what shopper is going to wade through 10,000 items?!) That's a problem for sellers whose items don't make it to the top of search, and it's also a challenge for Etsy, which uses algorithms to influence which items appear highest in search results.
Etsy CEO Josh Silverman said he's making a change to his site's approach to its search algorithm, calling it a "cushion-to-couch" strategy.
Should Etsy give more weight to lower-priced goods (the "cushion"), since that would boost conversion rates?
Or should it give more weight to higher-priced goods (the "couch"), since that would boost gross sales? In other words, more shoppers buy cushions than couches, but it takes a lot fewer sales of couches to generate the same amount of sales dollars than couch cushions would.
Etsy had been doing the former, as the CEO explained during a conference call with analysts last week, but is in the process of moving to the second approach:
"Our search engine has historically used conversion rate as its objective function. Meaning when you type in a search query, it's trying to show you the top 50 things that it thinks you're most likely to buy. There's an inherent bias to low priced items then because you're much more likely to buy a $20 item than a $200 item.
"So if you were to use GMS (Gross Merchandise Sales) as the objective function instead of conversion rate, you might buy fewer things but of higher value, resulting in higher GMS. We talk about this as the cushion-to-couch strategy inside the company. People right now buy cushions, how do we get them to not only buy cushions, but also buy the couch,... And there's actually some inherent bias built into our search algorithms driving them to lower priced items."
Some readers might ask, why not show cushions when a person searches for cushions, and show couches when a person searches for couches? It may not be quite as simple as that for certain search queries, but on the other hand, marketplaces often get it wrong when they think they know what shoppers want. The law of unintended consequences rules when it comes to algorithms, as sellers on eBay know all too well - just ask them about "Best Match."
But Etsy is going all in - Silverman said its Google cloud migration is enabling Etsy to leverage more sophisticated algorithms thanks to "elastic compute power."
Etsy is also embracing machine learning and personalization with the help of Google technology - "With our cloud capacity kicking in, we are now beginning to enrich recommendations on listing and landing pages by incorporating more information and leveraging more detail about buyers' previous purchases, item views, and items they've added to their carts," Silverman said.
Sellers may be skeptical about any kind of search manipulation. After all, sellers pay the same listing fee for a cushion as a couch, yet the marketplace has not been giving them equal exposure in search results.
But a move to show more higher-priced goods in search could help sellers of quality goods compete with low-priced imports. One question is whether Etsy will swing too far in the other direction.
It's also interesting that Silverman framed the algorithm in terms of Etsy's bottom line (conversion rate vs gross merchandise sales) instead of framing it around shopper satisfaction.