A Cuil Search Engine is Born

“Cuil” supposedly means “wisdom” in Gaelic. In English, pronounced “cool,” it means business when it comes to Internet search. That’s what its founders, nearly all ex-Google employees, want you to believe. Let’s take a look at their claims and their performance to see whether it’s wise to take them seriously.

Cuil is the brainchild of Tom Costello, Anna Patterson, Russell Power and Louis Monier. Costello is an ex-IBM employee and former member of Stanford University’s Research faculty, and his work at both organizations focused on search. Anna Patterson joined Google in 2004 after it bought Recall, the search engine she designed, which indexed 12 billion pages (a record at that time). At Google she handled the problem of scaling architecture – keeping up with the growth of the web. She left in 2006 to build Cuil.

Power and Monier, like Patterson, boast Google on their resumes. Both dealt with issues of scale as part of their day-to-day work, like Patterson. Monier also did serious work on search at eBay and AltaVista. When founders have this kind of background, even Google has to sit up and take notice. One has to ask: how does Cuil approach search, and how is it different from Google’s approach?

Cuil points to four guiding principles in its philosophy. First, it believes that size matters, which is not surprising when you consider the background of the founders. Second, popularity is useful, but not always important. This is reflected in its claim to index pages based on content and context rather than just the number of links leading to a page. Third, it believes that organization is fundamental, and it uses this principle to classify the information it finds so you can find it better. And fourth, privacy is paramount; Cuil explains that it analyzes the web, not its users.

{mospagebreak title=Debunking and Exploring}

Danny Sullivan does a good job of debunking each of these points in his Cuil review. I’ll give you the quick version. First, size isn’t as important as relevance; you can index billions of pages (Cuil claims 120 billion) and still deliver crappy results. Unfortunately, there is no good metric for relevance, but size doesn’t work as a stand-in.

Second, Sullivan ran some tests that seemed to indicate that Cuil does depend on links to some extent when figuring out its results. He also found that Google does some link analysis, so Cuil and Google might not be as far apart here as Cuil would like you to believe.

Third, while Cuil does use its layout and certain features to improve organization of its results (as I’ll show you in a minute), Google results are also diverse. And fourth, while privacy is good, it doesn’t seem to have won any battles for other search engines – and lack of privacy hasn’t exactly hindered Google’s growth.

All of that said, let’s take a look at what Cuil actually offers. I did an idle search or two on the search engine on Monday night, partly out of curiosity and partly because I knew I’d be writing this review. I read a lot of bad comments from other sites about the results. But the interface itself earned a lot of praise, and the screen shot below shows why.

What you see here is a three-column format, very similar to a magazine-style layout. You can also change it to a two-column layout if you wish, simply by clicking a link in the lower right corner. For each result, you get a much longer blurb than Google delivers with its snippets. You also get an image from the page (at least in theory; more on that in a bit) to help you decide whether to click through to the actual web site.

Do you remember that Cuil said one of its guiding principles is organizing your information? That gray box in the upper right hand corner of the results helps with that. Let me give you a closer look. For reference, I searched for “washington dc tourism” (without quotes).

Those are all logical categories. Now watch what happens when I hover over a category:

I can click on those links in turn, to do another search. But before I even do that search, I can just hover over a link if I want more information, and Cuil gives it to  me. Take a look at what happens when I hover over the Washington Monument (virtually of course, not literally):

{mospagebreak title=Tabs and Suggestions}

These aren’t the only ways that Cuil organizes your information for you. When you first do a search, as you’re typing the words in, Cuil gives you a drop-down list of suggestions. It’s similar in some ways to Yahoo Search Assist, but less sophisticated. Here’s a screen shot.

That box to the left of the top suggestion really should contain an icon in the shape of the site’s logo. Anyway, if I had clicked on that top item in the drop-down list (Washington Mutual) I would have been taken directly to that web site. Clicking on any of the other items in the drop-down list causes Cuil to perform a search using that term.

After you do a search, Cuil will sometimes include tabs in the gray bar above the search results. These give you more options, more ways to organize your search and focus more tightly on what you’re seeking. Here’s a screen shot from a search for apple (which doesn’t even offer the fruit):

Clicking on the “more” tab delivers a drop-down list similar to the one Cuil uses to make suggestions when you start your search, while clicking on the other tabs causes Cuil to perform new searches using that term.

All told, this is a very nice interface, and it’s very smoothly executed. In the time I’ve spent reviewing new search engines, I’ve seen a lot of different interfaces. Some were good, and some were truly awful. This one is both easy on the eyes and seemingly more informative than Google’s standard ten-blue-links-in-a-list. It actually feels more natural than the standard search interface, which is quite an achievement.

{mospagebreak title=Relevance Matters}

Now for the bad news: Cuil’s results are uneven at best, and truly awful at worst. In my initial search on Washington, DC tourism, the web site for the Washington DC embassy of Turkmenistan came up. Now I have nothing against Turkmenistan, but its embassy in DC is probably near the bottom of the list of places I plan to visit on my fall vacation.

Danny Sullivan tested Cuil’s relevancy in some detail and also found it wanting, especially when compared to Google’s results. But that’s not the only problem. Cuil presented other peculiarities that turned up even in the few searches I did.

For example, when I clicked through to “Washingon Monument” as a category suggested by Cuil on one of my searches, I got a page that said “We didn’t find any results for “washington dc tourism Washington Monument.” That’s ridiculous on the face of it – but also, if Cuil didn’t have any results for that, why did it direct me there? Another time I tried clicking through a category, Cuil said it found a few thousand results – but only showed me one, which wasn’t actually relevant.

The issue of relevance haunts Cuil in other ways. I have seen comments around the web that many of the images Cuil uses in results aren’t relevant to those results, and don’t even appear on the page with which Cuil associates them. The most egregious example of this was reported by The Register, when one of their readers, a quantum researcher named Dr. Jonathan Grattage, did an ego search. The results that turned up, though apparently leading to information about the good doctor, included images of American soldiers and other males that (one assumes) looked nothing like him.

What is going on here? Cuil, since it is new, is getting slammed with lots of searches – more than its founders expected this early in the game. But it is returning results quickly. Rafe Needleman, in reviewing the issue in his blog, explained what was going on. The way Cuil is set up, a traffic spike won’t slow it down, but it will make its results less accurate. “This is because Cuil isn’t set up as a massively parallel search network the way, say, Google is…Each of Cuil’s search appliances is specialized to a particular subcategory of results. There are machines that understand and index sports; other are experts on medicine…As these search machines get overloaded…they drop offline for some queries, and the machines left online return less-than-relevant results that then appear at the top of users’ pages.”

This probably isn’t going to win Cuil many friends or fans. Some observers said the search engine launched too soon. That’s entirely possible, though it had to launch sometime, and obviously the founders believed it was ready. Remember, though, that it is literally a brand-new search engine. How accurate was Google when it first came out, when compared to today? Cuil needs a little time to get its feet under it. Whether it will get that time in this highly competitive field is another question. I wish them luck. 

[gp-comments width="770" linklove="off" ]