For the past few days I've been using Snap, the new search engine unveiled at the Web 2.0 conference.

At first glance Snap struck me as cluttered and busy, very different from other SE's.
One of the first complaints I saw was that Snap disables the scroll wheel on the mouse to scroll through the results, but this is missing the point of the new approach in Snap - the old behaviour of scrolling through pages of results or even below the fold of one page is what is being avoided here, even the familiar link to the next page of results you see at the end is almost unnoticeable.

Snap allows you to filter your results after you've searched - say you do a search for cruises, then want to drill down to something more specific like Mediterranean cruises, you type Mediterranean into a second text field that reorders the results as you type, this is very cool and doesn't even require another page to be loaded.

In addition to typing in refinements, the results can be sorted by columns with different functions, clicking these columns reorder the results in descending order, clicking again sorts by ascending order.
The default columns:
Popularity - how many users clicked that link directly after doing the exact search you just performed
Satisfaction - number of pages on the site visited by users directly after doing the exact search you just performed
Web Popularity - the amount of Snap network visitors to that page
Web Satisfaction - the amount of pages viewed on that specific site by members of the Snap network

The Snap network are users who have searched on Snap, combined with anonymous browsing history data Snap acquired from ISP's (more about this here).
I haven't found much use for these options myself yet, the second query field being the most powerful option I've tried, but I will give them a further test as I use Snap.

Searching for shopping queries like mp3 players will give you even more new columns with which to sort the results, based on price, memory, PC interfaces, manufacturer and other info directly related to the query - this is an excellent feature, it's the first instance I've seen of a search site actually rearranging its layout to present the best info possible.

Certain queries also get special treatment - for instance a search for yahoo will add an extra section between the query box and the results that contains refinements within Yahoo (mail, login etc), the latest news and company information.
The criteria for receiving this special treatment is hard to decipher: it works for broad generic searches like games or news, and some major companies get it like Yahoo & Google, but strangely, not Microsoft.

Another aspect of Snap getting a lot of attention is their unusual approach of opening up all their statistical info (searches performed, incoming links, advertisers and revenues earned) and displaying it front and center on the home page.
This is almost the mirror opposite of Google, who are famously secretive about the usage data on their engine.
Apart fom just being an opposite to Google and being a useful tool for advertisers though, I can't really see the reason or use of it to a user except as a novelty.

I did notice a few flaws in my initial test of Snap

• the small index - the refinement tools would work excellently for ferreting out pages within one website, but the providers (Gigablast & Looksmart) don't have near the same coverage of the Web that the large engines have; the coverage is far from complete and this is a major drawback on the refining aspect.
• Loading results pages hangs Firefox for me momentarily (like Gmail) while it loads tons of JavaScript into the browser. This makes the load time significantly longer than Yahoo or Google, though I can see that the reorder-as-you-type feature requires a lot of code to be loaded into the browser.
• Refining your search down works well for broader searches, but it comes apart on more specific queries. eg a search for President Bush, if you try refining down to one of his famous quotes "axis of evil" there are no results, even though the exact page to match that search is in the Snap index.
  This is likely because of a limit on how much code can be loaded to the browser after each query, and shows the drawback of relying on client side to provide the refinements.

These are my impressions after just a few days of using Snap, and I probably haven't seen everything it has to offer yet (it's that feature-heavy), but apart from the flaws above (which I'm sure will be looked at) I am very impressed with it so far.

For more info:
· Official Snap weblog
· Why we built Snap - By founder Bill Gross
· New Snap site thinks outside the search box
· Snap Gets Into Search Game
· Snap, But No Crackle Or Pop
· Snap: The Revenge of Bill Gross
· Bill helped Google with paid placement, can he now snap its search?
· Snap.com: "A different kind of fish."
· Jeremy Zawodny liveblogging from the Web 2.0 conference