Over a couple days starting with its launch July 28, Cuil gained unprecedented media attention for a startup, crashed when too many people tried to access its search engine and angered people by not putting the correct pictures with some search results.
Three weeks later it is clear that the company, which aims to one-up established search providers Google, Yahoo and Microsoft by indexing more than 120 billion Web pages and providing meaningful pictures next to search results, has made some considerable improvements with regard to the pictures it places with results.
For example, Trenton Baker, a marketing manager for network storage vendor Aberdeen LLC, told me July 29 that several of his company’s industry-specific search results ended up with misused Web site photos and misleading content results.
At that time, a search for “Storage Servers” returned a news release about a product from Aberdeen rival SMC Networks. Yet the product picture Cuil gives for this result is lifted from an array on Aberdeen’s front page. When you clicked on the picture link, instead of going to www.aberdeeninc.com, you were taken to the SMC press release.
Baker felt this discrepancy dinged his company’s brand. Do a search on Storage Servers today and the picture next to the Aberdeen search result for networked attached storage is Aberdeen’s logo for its rack-mount five-year guarantee.
The result and picture are appropriate, signaling that perhaps Cuil’s context-based algorithm for providing relevant pictures alongside search results has been finer tuned.
“This is not an easy thing to do, to provide a relevant image next to as many result links as possible. Which is why this feature isn’t being offered elsewhere,” Cuil Vice President of Communications Vince Sollitto told me recently. “There are plenty of search engines one can use that do not attempt to provide this feature if one prefers.”
Well, sure, but you don’t want to tempt people to use another search engine. At least, not until you get pretty big, powerful and as close to perfection as, say, Google.
Aberdeen isn’t the only company to notice the difference from July 29 and the present.
Denise Enck, proprietor of www.emptymirrorbooks.com, left this comment Aug. 3 that when searching Cuil for keywords closely related to her Web site, she found that her company’s logo (and other proprietary images) appeared next to descriptions for, and links to, unrelated Web sites as well as those of competitors.
For example, a search on “beat generation books,” her company’s partial logo appeared with a link to the United Kingdom newspaper The Guardian. She e-mailed Cuil July 28 and July 30 until Cuil told her the issue would be resolved promptly. I circled back to Enck yesterday and asked if Cuil had improved. She wrote me:
“Yes, it’s improved. I really raised hell with Cuil, and though they never answered my last e-mails (sent to several individuals within the company), they did clean up their act very shortly after that. They’ve stopped using my logo & proprietary images to promote other people’s Web sites.“
Cuil Search No Google but Getting Better
=Is Cuil Current? }
Despite this improvement, Enck said the search results are still wonky. She said she ran quite a few searches on Cuil since its launch and has seen little improvement in the results for any keywords since then. For instance, a search for one of her other Web sites, Quanta Webdesign, returns only sites that link to Quanta.
“I think Cuil has a long way to go before they can begin to give Google a run for their money. I think it’s stunning that they went public with the site before they worked out these rather serious issues.“
I’m glad to see Cuil is improving, though I still have a quibble with the service. In doing a Cuil search on my colleague Roy Mark, a picture of our former boss at InternetNews.com comes up, which is pretty inaccurate given that Roy has been with eWEEK since last September.
Cuil’s Sollitto tried to clarify for me that, unlike standard image search based just on names to photos, Cuil uses an algorithm that is applied more broadly across the Web based on concepts, content, context and relationships to provide images for non-name searches as well. He added:
“This is why a picture of one of your colleague’s former colleagues may appear next to a result mentioning him; the former colleague is “related” to the search subject. While the relationship may not be current (as you note), it exists on the Web and is a relevant result to the search subject.“
He added that Cuil recognizes that people doing searches solely on names may only prefer images of that person, rather than text results that relate to the subject and image results that relate to the subject, so “we are working to tweak our algorithm to increase its bias in that direction and demote other images for those specific name searches.”
I accept Sollitto’s statement, and I appreciate Cuil’s hard efforts to meet customer concerns. But as a journalist I can’t help but thinking of relevant searches as ones that are the most “current.”
What do you think? Am I wrong to expect this of Cuil?