nicolas leroy

TheFind’s new merchant feed, and what it could mean for search

March 22, 2010

Even if I really appreciate TheFind for the innovations it brings to the CSE world, there is one area where I think its product currently lags: search relevancy. Indeed, with an index of 400M merchant offers (as claimed in press releases), noise in search result pages becomes an issue; and TheFind currently doesn’t have that many filtering options to refine queries and let users remove this noise; the strategy for users is therefore to change their search term.

I.e: a search result page for “canon eos 50 that brings 11 accessories for 1 camera, in the 12 first results:

TheFind search result page

Last week, TheFind has announced an enhanced feed specification for merchants, which could give hints of how TheFind wants to improve the findability on its site:

TheFind, which hosts 17 million unique monthly shoppers and sends quality traffic to 150,000 stores, today introduced an enhanced feed spec and a new merchant center that allows online and local stores to submit their product feeds directly to TheFind, for free. […]

Merchants can use popular feed formats such as GoogleBase, or they can use TheFind’s new enhanced feed spec, which includes more detailed attributes relevant to a growing number of online shoppers.[…]

“We wanted to make it as fast and easy as possible for merchants to get their feeds directly into TheFind, which is exactly why we are perfectly happy to receive the GoogleBase feed that many merchants are already using,” said Jordan Keffer, director of merchant programs at TheFind.  “We also know – based on our own data – that existing feed specs don’t cover many of the product attributes consumers search for, so we’ve taken this opportunity to also define a new feed spec that better matches how shoppers actually search for products and stores.”

At this point, this is pure guesswork based on the feed specification:

So far, that is pretty common for CSEs to ask those kind of fields to merchants. I’m however intrigued by the following fields that have the potential to greatly improve the search for similar or related products (whatever similar or related could mean for the end users):

Style_ID – The id number for a particular style of a product. The style should be labeled distinctly in the Style_Label field. Styles may have the same SKU but different prices, UPCs, available sizes, etc.

Style_Name – The style name given to a variation of a product. Styles often share the same SKU number but may have different colors, materials, or price. (e.g. a particular shoe’s style may be “Black snake” or “Silver soft kid”)

Compatible_With – A list of SKUs that are compatible with the product (e.g. ink toner with a list of printers, or belts with a pair of shoes)

Similar_To A list of SKUs that may be similar to the product.

Color – A comma-separated list of colors a product comes in. For example, “red,green,teal”.

Along with experiments on visual search, the search features that could be built from such fields could greatly improve the navigation on the long tail of products, and the discoverability of “side” products… Can’t wait seeing what TheFind could do!

I suppose we will need a few months before seeing any front-end improvements linked to this new feed format, even if 1000 merchants have started sending feeds in the 48 hours following the announcement according to TheFind.