Monday, February 18, 2008

Google Already Does What Wikia's Trying - Social Search

It wasn't all that long back when everyone was disappointed with Wikia's launch of alpha search. With Wikia Search everyone was expecting it be a Google killer from day one and the disappointment was made very public by a lot of tech bloggers. At that time, I thought that it was a little harsh of everyone to be criticizing something so "alpha". But I guess the Web is now v2.0, its growing rather rapidly and people can't wait for social evolutions any more.

Give Social Search Some Time

Wikia Search is about social searching where people search for a specific keyword, follow links, vote for good search results and the contribute a small relevant mini-article. Darwin realized it long back that evolution is slow and made the world know about it. Social evolutions in behaviour patterns are even slower... Then why in the world would anyone think social search will start working from day one?? The success of social networking in the last 2-3 years may be the primary reason where people have started to expect faster growth from social phenomena on the web. But the reason for success was that people did not change their behaviour. They just got a completely new experience since social networking didn't exist earlier.

Search on the other hand already exists and people are behaviorally already accustomed to using algorithmic search. Better web users can already think of good keywords and get their results google'd (since that's a verb these days). So really, Wikia or for that matter social search will take its own time to grow. And if I were to predict, after it matures it'll be better than algorithmic search.

Google understands Social Search

I performed 3 tests over a period of a 30 days and have come to the conclusion that Google performs Social Search, but seems to be a little confused on how much priority it needs to give to social valuations.

Test 1 (Google Reader & Feedburner): I created 6 blogs with the same content and kept posting and pinging the Google BlogSearch equally. Each blog was updated daily at the same time from the same location. But I subscribed different users to the feeds differently using feedburner and google reader. Initially without any feed subscribers, all the blogs would randomly appear above/below each other for a week. After a week, with different number of subscribers, I found that Google would display that blog first which has more number of feed subscribers. Thus Google very well knows which blogs have more subscribers and displays those in the result higher than the others.

Test 2 (Search Result Clicking): This was a weird one to detect. I found that sites which are clicked more in the results got the better ranking in the future. But this was not a general higher ranking in the search results. I observed that it was only limited to that IP subnet where the clicks were made. It meant that Google was able to understand which sites the user visits and hence brings it above others in the search results.

Test 3 (Orkut & GTalk Contacts): Google's social networking site Orkut was my next target. Considering orkut scraps and communities weren't part of Google's index, I didnt think they would have any effect, but I was surprised!! I couldn't spam a lot of friends with the blog links, but I did some in communities which had foolish admin's accepting me as a member. I spammed about 200 communities with the blog links and found a change in the search results. These weren't as straightforward and sometimes the higher ranking would be lost and would come back again at times. But there were more times when pages with orkut links ranked higher.

When I did similar spams with Gtalk contact though none of sites changed. They were still random in the results.

Random Observations:

  • Blogger/Blogspot sites ranked higher: Initially I didn't realize this, but later found out about 2 weeks into the testing that Google was ranking blogger/blogspot sites higher in the results. It was surprising at first, but then I found out that Blogspot's subdomain had a higher ranking. Other blogs that are subdomained to good domains also had a similar effect. The conclusion I reached is that Google seems to be checking the domain name first and then its content. Even if the subdomain name was more apt to the search keyword, Google cared more about the domain name.
  • Knol could be the missing link: I have not tested Knol yet, but it could be the primary decision maker for the Google's engine on social search. Knol articles and the links it mentions could be more valuable because Google has stated that Knol articles are expected to come from experts on any given subject.
  • Gmail could also be used: I tried sending lots of emails (16000 in all) with blog links. None of it had an effect :( I think Google could use it very effectively to improve search results. But I guess there could be some privacy concerns with that happening, although Google already has the rights to index our emails in the EULA for Gmail.

The tests are completely my own observations and are limited in resources (6 blogs) and time (30 days). But they give us a slight indication that Google may be social searching already. How much they rely on it is completely unknown and will be unknown forever... Google already knows so much about society on the Internet that they are probably the only company (with enough resources) that can accelerate social search and implement it successfully today!!


Akshah said...

nice article dude.. good experiment by u..

Anonymous said...

Good test and must have been quite some hardwork. I think Google is very clever to understand that social search is the future. But it keeps changing its search algorithm so often, that there is no sure method to know whats good at the search results!

Fake Steve Ballmer said...

Well done with the test mate... But we are gonna get yahoo and do better :)) hahah!!