Feeds:
Posts
Comments

Archive for July 28th, 2008

This morning’s web 2.0, vc, internety buzz is all about this new search engine, Cuil. Even CNN devoted the top space to it.

Here’s some more fan fare on TechCrunch.

Cuil’s front end is built with CherryPy.

This search engine is not good.

Touting the size of your index is a bit like pitching the world on GHz for chips or telling people what jet engine you use for the FedEx planes.  No one cares.  Does the product work?

Can I find what I’m looking for or be surprised by what I didn’t know I was looking for?

Cuil doesn’t work.  Usually it doesn’t bother me when a new internet product doesn’t work.  It’s the web.  Fly be free.  So what’s my beef here? This engine is getting major buzz and its a complete let down.  I want to believe.

Search Logic:

Try typing basic queries like restaurants in Chicago, then use their sidebar to drill down.  Ask.com works better than this.

Try your name.  If you have even a slightly non-obvious name you’ll get 0 results.

Try SOCRATES.  For goodness sake, SOCRATES.  0 results.

Layout:

Everyone tries alternative layouts for search.  I’ve seen and built so many myself.  Why? Perhaps if it looks different it will perform differently? Perhaps a different look will make it seem better? newer? improved?  Generally it boils down to “Google has been this way for 10 years it’s time for a change.”

I posit that most new search engine builders don’t actually look at the user behavior though.  If they did, they would never try these new layouts.  There’s simply no demand for them.  Users are confused by them. And the information these new engines show case doesn’t require a new layout.

Images, in non image based searches are highly distracting.  Most of the thumbnails an automated crawler picks up will not index correctly for the search at hand.  So the results relevance, visually, is diminished.  Images attract the eye too much and provide too little directional cues to drive a click (and you need clicks for the user to get the payoff).

The grid layout versus the rank ordered list also provides no cues for the user on which links to explore first.  The grid is easier to visually scan the whole set but again this reduces the click behavior of the user because they explore through scanning versus clicking.

The lack of bolded words in the title and URL is pretty bad too.  They took away yet another cue for the user to click.  Users don’t read on screen.  They rapidly scan, bouncing back and forth between interesting representations on the screen.  Without the bolded words, the eyes have no anchor points within the actual results.

Let me explain that point.  Finding the best results is only a small part of what makes a successful search engine.  Users need to visit the sites a search engine returns.  The only way to validate the relevance of the search results and to get what a user actually wants is to go to the sites.  Users often will find what they want and return to the results more satisfied.  Creating and interface that delays that behavior reduces the result relevance (perception) without having anything to do with the actual algorithms!

The tabbed results are interesting and tabs are a well known visual element now.  However, sometimes the tab contents bounce around and its slightly unnerving.  Worse using the tabs often produces worse results.  This hurts the experience because the user isn’t forced into producing a new search behavior – essentially they never shake off the bad initial query.  Instead the tabs are used in one “continuous” behavior and provide feedback to the initial search behavior which might lead to the user never trying a better query.

Pagination, results numbers and other numeric cues.  These are fairly inconsistent – that is, often the number of results bounce around page to page within the same result set.  Hard to trust the results when you have numbers changing and the layout already keeps you off balance.

Business:

If they did find a cheaper way to scale, that’s interesting.  If they can really handle query massaging better than google, that’s interesting.  This experiment might be worth a lot just for those two facets alone.  The user won’t appreciate those in the least, but acquirers and techs will.

Conclusions:

I think this is getting press partly because the media loves telling an underdog story and we love to take on the incumbent.  That’s great.  No harm in that.

Heck, there’s no harm if people think Cuil is great when it isn’t, except for Cuil and the folks who put in $33mil.  If they want to be a contender and not a “so what” like Clusty, Ask and all the other non-yahoo/non-googles they better keep it real.  If they had product bravery they would kill off access to google within their offices so they were forced to use Cuil for everything.  I’ve worked at a lot of search engines and not a single company was willing to kill off access to google and completely entrust their product to find what they need.

The pitch of privacy as a good reason to use Cuil is folly.  Few users hold back using google because it tracks you. No one stops using Facebook because your data flows freely.  These products deliver value.  If your pitch to users for using the search engine is not 100% because the search kicks ass you’re going to lose.  The behavior will be disconnected from the consequences and when that happens you have no hope of keeping the users searching.

I didn’t even get into the business model, advertising thing and how that is a huge feedback loop for google.  That’s for a later post.

Give me $33 mil and I’ll show you a search engine that has a shot 😉

No, really, if you have $33 mil and you want to try something that has a chance, call.

Read Full Post »