Search Engines: Google Not Cuil

Now that my life is so prearranged

I know that it’s time for a cool change

The Little River Band lyrics date me, I’m afraid. In these lazy days of summer, FM radio channels memories. But those lyrics seem appropriate considering the buzz recently about the announcement of Cuil, a new search engine positioned as an alternative to Google. In a time when we use a brand name (Google) to fully replace its function (search), any competitor will have an uphill fight. How many brands of facial tissue (i.e., Kleenex) can you name with confidence?

From Google’s Loins

The fact that Cuil was founded by ex-Google employees has put it under more media scrutiny than other similar offerings in the past. Cuil founder Anna Patterson worked at Google to refine the index and ranking capabilities of the search engine. Joining Patterson is her husband, Tom Costello, a former researcher at Stanford, and Louis Monier, former CTO of the pre-Google search engine Alta Vista. The company, based in Menlo Park, has about 30 employees.

Cuil has a lot going for it: big VC backing, big ex-Google brains, and a big index. It has a newspaper style UI layout and relational grouping. That is, it clusters results based on cross-associations. That all sounds very promising, if not necessarily unique among Google alternatives. Other search engines like Clusty and Tafiti feel very similar to Cuil, yet they haven’t grabbed the attention of the majority and certainly pose no threat to Google.

slammed in the blogosphere and the tech press, operational site issues, weak search results and strange dynamic associations between search results and images. On the day of the launch at about 3 p.m. E.T., Cuil’s servers went down hard: probably not such a great thing considering the number of reporters, analysts and other influential people taking it for a test drive. It’s tough to test a search engine when the error message reads, “Due to excessive load, our servers didn’t return results.” I can just imagine those 30 employees in Menlo Park running around putting out fires. But hey, they’re not a bank. Cut them a break on day one.

Search and Ye Might Find…

Once the Cuil site was up I tried some searches, with mixed results. When I did this search on Cuil: “+burton group” “+chris howard”, nothing appeared (!!!). After picking my ego up off the floor, I removed the plus signs and tried again. This was more like it, but there was one notable issue: the articles returned were spot-on, covering my recent writing and podcasts, but none of the pictures was of me:

This is Chris Howard, the football player, but I’m quite sure he didn’t record a virtualization podcast.

This guy has cool long hair like some analysts, but he is not me. He might just be able to convince you about the value of BPM infrastructure, though.

I don’t remember seeing either of these folks at Storage Expo, but neither of them is me.

This is a book on Costa Rica, written by Chris Howard (not me). Unless I’m mistaken, the book doesn’t contain any Costa Rican activity diagrams modeled in BPMN.

After having a good chuckle, I dug in to see why these pictures appeared. They were not images embedded elsewhere in the target result, as sometimes happens with Google’s image search. Strangely, searching on just my name returned the same odd set of images. Apparently, the search engine guessed that these pictures might be relevant to my search, turning the result set into a bizarre collage. The real issue with this image disconnect is that it is immediately distracting. The success of the text search may be overlooked as a result.

With a different search, this time for jazz pianist Keith Jarrett, the results were more promising. Almost all the images were correct, if not obviously connected with the target abstract. Because the result set returned was sufficiently large, a “category explorer” appeared that allowed me to browse related subjects. Additionally, tabs appeared across the top that grouped results into likely clusters, allowing me to narrow my search based on useful defaults.

The Value of Cuil

Early difficulties, bugs, and operational issues aside, Cuil has significant value even as a conversation piece. So what if Cuil isn’t perfect? Maybe its (unintended) role is to dislodge our search complacency. Even just at the UI, why are we OK with a somewhat random linear list of results (where companies work hard to manipulate their position)? A well-annotated result set has more value, especially if the first results are the most likely to be relevant. Relevance would be increased based on other people’s searches and click paths. Furthermore, relevance would be more personal if the engine knew my history and used it as a filter. Cuil “reads” the page to understand context and infer relevance. Google uses the number of incoming links to a page to determine its popularity. Some combination of all these approaches will yield better results targeted at a specific user.

Sometimes I want search to replicate the old experience of climbing through dusty library stacks. The most interesting things I learn are not a result of what I set out to discover. Ambling through the stacks at McGill University led me to information I never would have found if I only had performed a surgical title search. Those finds were findable because the resources were grouped in proximity to my initial target. The clustering and category suggestions within Cuil and other marginal search engines bring me closer to this experience.

Even if Cuil does not succeed at displacing a significant percentage of Google’s mindshare, Cuil should be a reminder to Google to keep innovating. It seems that the engineers at Cuil (who came from Google) saw a hole in Google’s offering and took advantage of it. These “masters of the algorithm” lead the charge to better search results. Those engineers back at Google are paying attention and are likely looking at algorithms beyond the Cuil innovation. Like many systems that are reaching peak performance, search can only be made so much better. Google needs to address user experience needs that supersede the core search function.

For now, the majority of users will continue to “google” by default, unaware that their experience might be improved if they chose an alternate search provider. The bigger concern is that our blind adherence to the status quo keeps us from breaking out of the Google paradigm into something (apologies for this next word…) richer.

By the way, this picture really is me:

Chris Howard is vice president and director of the Executive Advisory Program at the Burton Group. Howard is a former university professor with more than 16 years of IT consulting experience.