EcommerceBytes-Update, Number 321 - October 21, 2012 - ISSN 1528-6703     2 of 6

eBay Cassini: Best Match on Steroids

Email This Story to a Friend

Imagine typing in a search term on eBay and getting back a small subset of results that are targeted just to you. That's something eBay has been testing as part of its new search engine, Cassini, which has yet to be deployed. But will Cassini be able to figure out what you're looking for when it tries to match listings with your search query? And what about sellers whose listings don't appear in personalized search results?

Best Match Search Algorithm
With 350 million listings on eBay, it's easy to get overwhelmed with the volume of search results. That's why eBay launched Best Match and made it the default sort order - to try and rank listings by relevancy.

Shoppers have a choice to sort by other factors (price and time ending) - so here's the challenge: be overwhelmed with a large number search results that may be less relevant, or stick with the default Best Match algorithm but be uncertain whether eBay's ranking was effective - are those listings at the top the most relevant to you? It's fairly impossible to validate if the results are really sorted in the "best" order, as you can do when results are sorted by price.

The challenge of knowing whether Best Match is bringing back all relevant listings first is compounded because eBay uses a number of factors to rank Best Match search results, and furthermore, it doesn't explain exactly which factors it uses and how important each factor is. In addition, the importance of each factor varies by category, and eBay periodically tweaks the algorithm (just as Google does with its search algorithm).

eBay's New Search Engine: Cassini
eBay has delayed the launch of Cassini - it had been scheduled to roll out this year, but at Wednesday's third-quarter earnings conference call, eBay CEO John Donahoe told Wall Street analysts that Cassini would roll out gradually over the coming 12 to 15 months.

"And it's, frankly, it's already running in certain areas and certain parts of the site," Donahoe said. "And we'll just continue to roll it out wherever it tests positive, and we'll aggregate it over the week. The first customers - this week is the first time we're putting any reasonable portion of customer-facing volume against it, and it's doing well. So I put that into the bucket of a series of technology and product continuous improvement that we'll be making over the next year or two years."

From the searcher's point of view, Cassini could be seen as Best Match on steroids. That's because it uses 100s of factors instead of 10s of factors when looking for relevant listings to display in search results. And it goes well beyond looking at structured data such as title, price, item specifics (eg, condition), shipping and return policies, and whether the seller is Top Rated.

Cassini will also use unstructured data - for example, it will mine listing descriptions for product attributes to enrich the catalog.

The challenge: how do you create product attributes out of free form text? And this is not static data eBay can mine once - millions of listings are added to eBay each day.

As difficult a task as that sounds, eBay wants to do many more impossible things before breakfast. It wants to mine the text included in buyer-seller communications, feedback comments - perhaps even written and phone communications sellers have with eBay customer service reps - to "read between the lines" about a seller's performance and trustworthiness.

And eBay will use behavioral data to make assumptions about listings and about buyers - not just whether listings are clicked on and if those clicks result in bids or purchases (which it already uses), but information about what buyers have purchased in the past.

See Monday's EcommerceBytes Newsflash article, eBay Poised to Turn Search on Its Head (Oct. 15, 2012) for more information about Cassini, and here are "Ten Things You Should Know about eBay's Cassini Search" (EcommerceBytes Blog).

Information Blackout
While eBay has briefed Wall Street analysts on Cassini, the topic is verboten when it comes to sellers. Repeated requests for more information about Cassini have gone ignored by eBay's Public Relations team since May, when Johnna Hoff said it was likely that "when Cassini implements any business-impacting updates, that we will give sellers appropriate notice, as it our practice."

News that eBay is testing Cassini in the third week of October is likely to make sellers nervous, as site stability is a key issue during the busiest time of year, and most sites put a moratorium on technology changes during the holiday shopping season.

Compounding matters, sellers are unable to measure any impact of Cassini testing thanks to a glitch. After reports from sellers that their traffic (and in some cases, sales) had plummeted, eBay said its Omniture traffic reports were broken. Despite writing on September 27 that it would continue to update its post with the most up to date information, there's been no update in 3 weeks.

Imagine running a business without key metrics during your busiest season, and not knowing how technology changes will impact your business in the coming year.

Comment on the EcommerceBytes Blog.


About the author:

Ina Steiner is co-founder and Editor of EcommerceBytes and has been reporting on ecommerce since 1999. She's a widely cited authority on marketplace selling and is author of "Turn eBay Data Into Dollars" (McGraw-Hill 2006). Her blog was featured in the book, "Blogging Heroes" (Wiley 2008). Follow her on Twitter at @ecommercebytes and send news tips to ina@ecommercebytes.com.


You may quote up to 50 words of any article on the condition that you attribute the article to EcommerceBytes.com and either link to the original article or to www.EcommerceBytes.com.
All other use is prohibited.