Search is increasingly broken, right? That is certainly my experience and that of everyone I speak to, although I haven’t actually seen any evidence. As the web gets larger and people get more sophisticated at SEO (i.e. gaming search engines) then it is getting harder and harder to find what you want.
There are two ways of attacking this problem:
- do better with the query information you’ve got
- get more information into the query
The big success stories so far have all been in the first category, with Google being the latest and most successful example, and there is more to come here, for example by mining clickstream data or taking advantage of user generated metadata from sites like del.icio.us.
Recently however the attention has been turning to the second category. Vertical search (about which I have blogged before) is a great example of this – because a site is focused on say real estate or travel the query is implicitly enhanced to say “and ignore all of the web outside this vertical”. Further, the query usually asks for information which is only relevant for that domain – e.g. postcode of the area you want to look for a property or the dates on which you want to fly.
That is great for vertical niches (some of which are huge) but for general search query disambiguation is much harder.
Google has put a fairly vanilla approach to this problem live at Simply Google. There they offer you 37 different search boxes to choose from – one for each of their 37 vertical search applications.
A more creative solution is to play with the search interface itself. This runs counter to what seems to be the received wisdom at the moment that consumers won’t be bothered to learn anything new or different from typing their average of 2.1 words into Google, but my gut says that this is the way to go.
In fact my first ever VC investment into Wordmap back at the beginning of 2000 was an attempt to do this, we were ahead of our time there and the company morphed into a taxonomy management solution, but the idea remains a good one. (As an aside, being too early is a nightmare for VCs and entrepreneurs alike. Market timing is everything, but getting it right requires a balance of intuition, analysis and good old fashioned luck.)
Quintura (backed by Mangrove) is approaching this problem by adding a tag-cloud style visual interface which automatically adds to the query every time the user clicks on an item in the tag cloud. I like the way the tag cloud updates itself to help you get more precise with your search and I really like the way the results on the right hand side of the page update in real time.
Hopefully these screen shots will give you a sense of how it feels, but I would encourage you to have a go. I have split the screen shot into two so it isn’t too wide for my blog software, so picture the two shots side by side, with the first on the left (and sorry for those of you who saw it earlier on when it was too wide for my central column).
According to ReadWriteWeb 99.99% of web searches are on Google, Yahoo!, MSN and Ask.com – which tells me that this market is a tough one for start-ups – but the prize is huge – that much is clear for all to see.
As ever, I would be keen to hear your views.
Finally, a bit of fun I came across during the research for this post. If you want a bit of a lauch go and see Ms Dewey and search on “visual search engine”.