728_header.jpg (23748 bytes)
 EB Blog 
 AB Blog 
EcommerceBytes-NewsFlash, Number 1963 - January 26, 2009 - ISSN 1539-5065    1 of 3

TIMEblaster's Steve Taylor Talks about eBay Search Technology

By Ina Steiner
January 26, 2009

Email This Story to a Friend

Steve Taylor, founder and Chief Technology Officer of timeBLASTER, has unique insight into eBay's search technology. In 2001, Steve developed a tool to help buyers search eBay more efficiently and was an early participant in eBay's Developer Program.

Prior to developing timeBLASTER, Steve was co-founder and Vice President of The Saddlebrook Corporation where he developed back-office and front-office applications for the banking industry. Steve also co-founded and was CTO of SandPoint Company where he designed and managed the implementation of a Lotus Notes-based application that searched for and delivered text information to the desktop. Introduced in 1990, SandPoint's Hoover product, an intelligent software agent, enabled knowledge workers in Global 1000 companies to query publicly available text information from multiple sources such as Dow Jones, Reuters, and Knight-Ridder's Dialog Information Services.

We sat down with Steve - a third-party developer, search expert, and an avid collector who actively searches eBay - to learn his thoughts on eBay's search technology as part of our ongoing retrospective series.

AuctionBytes: Give us some background on the tool you developed for searching eBay auctions, and why you created it?

Steve Taylor: The timeBLASTER software was originally developed for my personal use. I am a serious collector of US stamps and I found myself spending far too much time every week running searches on eBay and looking at the results...upwards of 15 hours a week! The problems were:

1) Too much pointing and clicking and waiting for pages to download.

2) There was no way on eBay to obtain searches results including only items not previously seen so a lot of time was spent clicking through to items previously viewed and rejected.

3) Images are critical to stamp collectors for judging quality and color. Initially the only way to see images on eBay was to click through to each item page. Later, eBay provided thumbnails as a part of search results...but the images were too small on the screen.

4) The number of searches one could save on eBay was initially quite limited and I had almost 100 searches! As a software engineer, I could solve this quite easily by building an html page containing all of the search URLs for my searches...but it was painful to continually update it as searches were refined and modified.

AuctionBytes: At what point did you decide to commercialize the tool you had originally created for your own use?

Steve Taylor: Almost immediately after the prototype was working satisfactorily. In addition to the problems noted above, eBay's syntax for constructing complex searches was poorly documented and almost impossible for the average PC user to understand. The prototype made it much easier to specify complex searches.

AuctionBytes: When did you join the eBay Developer's Program, and how has it worked out for you?

Steve Taylor: timeBLASTER was one of the first software firms to join the eBay Developer's Program when it was first introduced (and was expensive). We dropped it after eBay was unable/unwilling to investigate why identical searches executed through the API and the web interface yielded different lists of items and after exploring technology licensing and/or joint marketing of timeBLASTER with eBay.

AuctionBytes: Did you ever talk with eBay about licensing your technology?

Steve Taylor: Yes, we had extensive discussions with eBay regarding licensing of the technology. However, eBay was unwilling to offer Developer Program transaction costs which would make an API based timeBLASTER economical for the end user. In addition, eBay insisted on a that a totally different technology (ActiveX) be used instead of the Java based technology used for timeBLASTER. We would have had to completely re-write the software.

AuctionBytes: Tell us about how eBay search has changed over time.

Steve Taylor: The changes eBay has instituted over time are a mixed bag. Some things such as automatic stemming (searching for singular and plural forms of words when the singular form is used in a search), refinements in the category structure, and item specifics are improvements. Other changes, such as "keyword expansion" because it is poorly implemented, are negatives.

AuctionBytes: What's your opinion of eBay's new search, called "Best Match"?

Steve Taylor: eBay's new search and "Best Match" are two different things. The new search involves a few changes to the search engine but is mostly new layouts for the search results pages. "Best Match" is simply an additional way to order the search results and seems pretty useless as implemented. The new layouts are terrible...they involve larger data transfers between eBay and the user's PC and are result in fewer items per "screenfull" meaning more clicking/scrolling.

AuctionBytes: Do you think relevance search makes sense on eBay? Why or why not?

Steve Taylor: "Best Match" is an attempt to provide relevancy ranking to search results. Relevancy ranking typically is successful when dealing with collections of text where most documents contain a considerable quantity of text, when automated dictionaries are used to provide synonyms to the search words, and when "natural language" queries are processed. The tiny amount of text (typically users are searching only the Item Title) associated with an item make relevancy ranking almost impossible to effectively implement. And searching the item descriptions is quite problematic because sellers frequently add all sorts of words describing other items they are selling which have little to do with the specific item in question. Number of bids would be a pretty good proxy for relevance since the "best" items frequently receive a number of bids...but both sellers and active bidders would be VERY upset by this capability.

AuctionBytes: eBay said its recent acquisition of Positronic (and hiring of its cofounder Christopher Payne) will help it leverage machine learning to provide a more predictive and compelling customer experience. What does that mean to you, and does it make sense to do that on eBay's search engine?

Steve Taylor: I can't find enough information on Positronic's search technology to offer an opinion. In general, sophisticated search technologies (such as natural language search engines) work better when a) each unit of text (article, item description) meet stylist standards and b) each unit of text contains enough words to capture complete descriptions of the information. eBay items meet neither of these criteria and often contain as much, if not more, information about a seller's other offerings than about the specific item being sold.

AuctionBytes: John Donahoe indicated eBay would begin to use closed transaction data in search. What does that mean to you, and does it make sense to do that on eBay's search engine?

Steve Taylor: The use of closed transaction data helps more with the presentation (ranking) of the results of a search than it does with identifying the items to include in the list of results (eg, eBay's "Best Match" initiative). The use of feedback data, dispute data, seller volumes, etc. can place items more likely to satisfy buyers towards the top of the list.

As I previously pointed out, using number of bids to rank search results for auction items would serve as a pretty good proxy for Google's successful relevancy ranking which uses counts of other web pages that refer to a specific web page in order to prioritize search results. eBay has not implemented this capability (which would be fairly easy to implement within their current software structure) because some sellers would object rather strenuously ("Why should my identical item appear at the bottom of the list simply because it hasn't attracted bids when I paid the same listing fee?") and buyers would be less likely to bid in advance on an item knowing that a bid would simply bring on more competition by making the item more prominent. The use of closed transaction data can have a similar though less direct and less obvious an effect.

AuctionBytes: What is eBay doing right with search, and what could eBay do better?

Steve Taylor: eBay's search capabilities are generally decent. The key strengths are a detailed categorization structure, automated stemming, and item specifics where they exist. However, having both the "old search" and the "new search" running in parallel for a year or more is foolish. Users are confused as to what is happening and many do not understand how to force the use of one or the other. eBay's quality control for software changes to either search engine has declined significantly over the past year; both search capabilities have experienced a growing number of problems and errors. eBay's user documentation regarding either search capability is poor, incomplete, and confusing.

AuctionBytes: If you ran a marketplace that offered auction and fixed-price listings, and offered a broad range of categories, from cars to collectibles, commodities and electronics to antiques, how would you design search to help people find things? Is it possible to have one search engine for such a broad marketplace?

Steve Taylor: One search engine CAN perform the necessary functionality. Adequate quality control over the software and functionally stability over time are the keys to success. The vast majority of the user audience are not expert searchers and much of what eBay has done and continues to do represent attempts to serve this audience. However, there are also millions of users who do understand sophisticated searching and eBay has been serving this latter audience poorly. What is needed is an "expert searcher" search form that allows the user to more clearly and easily control whether certain search engine features are used or not used for a given search (such as automatic stemming, keyword expansion, etc).

Comment on the AuctionBytes 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.

Email This Story to a Friend
Email this story to a friend.

1 of 3

Sponsored Ad