« Monad Team Blog! | Main | Monad and WMI »

August 27, 2005

The Relevance of Relevance

A lot of talk in the search engine world these days is about relevance. Meaning how relevant the results are to the query you entered.

Now relevance is nice. But is relevance what drives people to switch search engines (like, say, from Google to MSN Search)?

I don't think too many people remember what the search engine world was like when Google arrived. I recall preferring Google to other search engines (AltaVista was the previous favorite) because of three things:

  1. Google showed an excerpt from the page that had the section of text right around the search terms; other sites showed no excerpts, or showed the beginning of the page.
  2. Google had a cache of pages which you could directly view, instead of being hidden away for use only by the search engine itself.
  3. Google did a great job of finding official sites. Meaning if you wanted the site for the Seattle Aquarium and typed in "Seattle Aquarium" as your search text, Google had a much better chance of returning the official site as the first item.

The third item is not trivial, and it showcases Google's technique of treating inbounds links as "votes". Many people may write about an Aquarium in Seattle, but the most linked-to site will almost certainly be the official site.

But I listed those three features above in what I feel is order of importance. Yes it was nice to be able to find official sites, but in many cases you were searching for information, and in that case the inbound link counting was not guaranteed to give you the best result. This is where the first two features came in. What they did was allow the user to have a holistically better experience with Google. It wasn't so much that Google returned the best results at #1, it was that the user's overall experience with Google was that they found the page they wanted faster. This was because the way Google excerpted text made the results page easier for a human to scan, and the cache made it easier to check out the linked-to page itself (because it would tend to load faster out of Google's cache than from the real site). So the user came away from Google happier, but it was because Google's had two features that leveraged the user's brain to provide the "last mile" of searching.

Note that neither of those two features had anything to do with the actual relevance of the searches; Google just had to get the right site on the first couple of pages of results and the user would do the rest.

Of course Google has improved tremendously in the interim, and has gone far beyond merely counting inbound links (if its algorithm even ever was that simple). Google now does a much better job with searches for information. Meanwhile every other search engine has adopted Google's excerpting and caching features, and they can all find the Seattle Aquarium if asked. Everyone is plugging away at relevance. But my point is that in order to actually pass Google--to draw users away from Google-- a search engine is going to have to go beyond relevance. It's going to have to something fundamentally different in how it presents the results that allows the user to have a faster, more successful search experience than they do with Google.

I don't have any silver bullet ideas on how to solve this. Actually I did have a notion that it would help if the search engines returned a thumbnail image of each site along with the results. What's that you say...it would take up too much bandwidth? Bah, spare me your resource limitery. I want a new pony NOW!!

Posted by AdamBa at August 27, 2005 11:01 PM

Trackback Pings

TrackBack URL for this entry:


The most linked-to site will almost certainly be the official site? In theory yes, in practice, no, bloggers and blog rolls, PageRank abuse, paid search results, SEO gamers and spammers, general link rot and such, sure gum this up. And what would be the official site for say 'Seattle furniture store'? Proper names and nouns, that usually correspond to domain names, are easy grabs, and seattleaquarium.org? That's not saying anything. The cache and the abstract was a hit tho, but forget not the spell check. But moving people to MSN? I think that's a lost cause already, regardless of revelancy or extra features or anything. ;) But Google is not really a search engine company, nor is MSN, they are advertising companies.

Posted by: Christopher Coulter at August 28, 2005 12:45 AM

IceRocket.com has some thumbnail images for web sites on their web search. They're not great but they are something. Right now it looks like the main page from large domains are there for everyone from that domain. It's a start though.

Posted by: Alfred Thompson at August 28, 2005 06:24 AM

If you're a Firefox user (of course you are), there are a couple of extensions to give you thumbnail previews in search: Google Preview (https://addons.mozilla.org/extensions/moreinfo.php?id=189&vid=1077) does exactly what you've asked for, and Better Search (http://bettersearch.g-blog.net/), which does it for more (it didn't work for me with MSN, but that was last year). Looking at these kinds of items would be a pretty easy way to get ideas for improving search, but I suspect the NIH (or as you mentioned before, ISTY) problem will probably mean these extensions are never heard of at MSN.

Another interesting idea is the clustering that many search engines are trying now (I remember seeing Vivissimo first), where domains are inferred from results and the user can select one as a filter. That would have to be done easier, though, maybe like a Google Suggest style interface, where the domains are shown as you type and can be chosen before you submit.

Posted by: Don Draper at August 28, 2005 12:01 PM

Christopher, I'm not saying the site would necessarily be seattleaquarium.org (although as it happens it is), just that the link counting works well in this case. If there was actually a store called "Seattle Furniture Store" I would suspect it would be the first hit on that query...if there is no such store than you are talking about a more general informational query where the search engine has to be more clever.

Yes Google has to deal with link gaming...but that all happened BECAUSE of their ranking algorithm. Until Google started caring about inbound links nobody bothered with that stuff because it had no effect. Now that they are out there, they have to keep ahead of all that stuff.

Google's spell check is also cool although I think it came a bit later.

- adam

Posted by: Adam Barr at August 28, 2005 04:12 PM

I remeber search not working too well until Google came out. It was a whole new web searching experience.
Also, what makes search engines effective is how often sites are indexed. You only posted this a couple of days ago and already if you search for "Bah, spare me your resource limitery. I want a new pony NOW", in Google, it finds it. Are so many people reading your blog that its indexed daily?

Posted by: SuperBK at August 29, 2005 11:10 AM