[disclaimer: I am Evangelist and CEO for Zooomr]
Bill Tancer – Hitwise US: Google Properties – The Extended List It’s interesting to note that according to Hitwise’s breakdown of Google’s traffic yesterday that Google Image Search is second only to Google’s text search representing 9.2% of their traffic. This is almost double many of Google’s other top properties including Gmail, Google News, Google Video, Google Maps, Blogger, etc.
It’s amazing to me that Google is not doing a better job with image search given how important a chunk of business that it represents for them.
Do a Google Image Search for any random word. Here, try monkey (I’ve got Peter Gabriel’s “Shock the Monkey” in my head this morning which is why I picked that term). Ok, so let’s look at Google’s results, there are a couple of ok shots of monkeys there but there are also shots of fish and chips and Star Wars and some random shots of websites, some kids, etc.
Now go over to Flickr and do a search for monkey and rank it by interestingness. Yes there are some erroneous photos there too. There are some photos of people and a horse, but there are less and the actual photos of monkeys are much better.
The reason why this is is because Flickr is using their social network to measure activity on a photo (faves, comments, notes, views, where it’s blogged, who faves it, etc.). They are also trying to patent this which is total BS.
What I don’t get is
1. As image search represents such a large chunk of Google’s business, why aren’t they creating their own social network to rank their image library (even if you only attacked the short head you could dramatically improve overall the relevancy of most of your searches.
2. Why after a year and a half can’t Yahoo! get it together and get Flickr’s interesting photos integrated better into Yahoo! Image Search? It’s amazing that Yahoo! has a legitimate trump card where they could dramatically show their superiority to Google and it largely gets ignored.
3. It wouldn’t take much for Flickr, Yahoo, Google, etc. to hire a small group of college interns and use them as editors to further scrub their image search results. It would be relatively easy to build a ranking score of all images and then manually review the tags for appropriateness, neutralizing tags in the background where they are inconsistent with image search.
Ask is also likely up to something but I’m not sure what it is. They’ve been talking about improving their image search for a while. Take a look at their search for the term monkey and you’ll see that it’s better than Google’s. Not quite as good as Flickr still in my opinion but getting better over the past year.
Given how important image search is and the deep pockets at Google and Yahoo! I’m surprised they are not doing more innovating here. The best it seems Google has been able to come up with is their lame image labeler game (about as fun as a day at the dentist) and Yahoo has the actual fun tool in Flickr but they can’t get it together enough to use it.
Image Search is one of the things that we are working on at Zooomr. Image search can and will get amazingly better over the course of the next year.