nicolas leroy

TheFind launches price comparison service

November 19, 2009

Today, TheFind introduces a new price comparison service on top of its product search. Not really innovative as all the “classic” price comparison services (think Shopzilla, Shopping.com, PriceGrabber, Kelkoo…) do it for years? True, but TheFind had some real challenges to solve on introducing such service:

Quick review of this new service called “Best price search” (launched in beta, so likely to evolve)…

Overview

From TechCrunch:

When you do a search for a product on the find, you can now choose to see the “best price” view. For now, the feature is only available for electronics and appliances, which amounts to around 40 million products (or 10 percent of the total index) indexed by TheFind. The search engine will do a lot of the heavy lifting for you. It will automatically try to detect your location and list the best prices from local retailers. The engine will also list the items that have the lowest verified price.

Few comments:

Extended view with

Clustering vs matching

Doing a search on “canon eos 40d digital camera 28-135” and finding similar offers… Most of the offers displayed are relevant for the query, but not all of them sports the new “Find the best price” link, suggesting some improvements are still needed to extend coverage.

Clicking on each ‘”Find the best price” link… Each leads to a different product page, with different offer / store numbers…

Yes, as opposed to other price comparison websites that most of the time rely on matching against a product database to build product pages, TheFind uses clustering technology. The main benefit is that such technology works on hard goods but also on soft goods… I.E: a search on “digital camera bag”. Should be quite computing-intensive on a data set of 400M offers!!!

I have noticed each “product cluster” triggers offers with similar images… I may be completely mistaken, but TheFind may use image similarity along with text analysis to create those clusters… When they tested visual search a few months ago, such image similarity was used to create the same kind of “product cluster”…

Three different clusters for Canon EOS 40D

Product pages

Let’s have a look at the product page itself:

Product page

Final thoughts

At first glance, the introduction of price comparison on TheFind is a follower move; that’s basically what the TechCrunch article published today and all its comments say. But looking more closely, we can feel the approach is quite innovative. Clustering technology, while requiring further tuning, fixes the problem with soft goods. Well thought filters on merchant reliability and indication of reliability for each offer is a way to tackle the issue of crawling rates and comprehensiveness. Really nice.