August 26th, 2010 | Published in Google Code
Over the months, we’ve had many requests to explain the way we rank applications in the orkut directory. Developers often wonder why one of their very popular apps doesn’t appear as high up in the directory as they believe it should. Well, it’s not exactly magic but simple math, and we wanted to share with you how our algorithm works out the rankings.
As you’d expect, we rely heavily on stats that tell us not only the number of users who have installed your app but also the number of users who actively use it. The number of installations is further broken down into the number of weekly as well as total installs. We hope you’ll agree that counting the number of users who uninstall your app is also crucial, since that is an indication of which apps didn’t live up to user expectations in some way and could be improved, and we lower the ranking score by a few points to account for the weekly uninstalls.
However, it’s not enough to judge the popularity of an application by the number of its installations alone – how often it actually gets rendered is a definite index of how addictive, useful and well-designed it is, and you can surely expect us to feed those numbers back into the formula, too!
Besides these, we think apps that users find good enough to put up on their home page should be given some weight, thus the number of weekly renders of those apps in profile view figures into our calculations too. We then add one last parameter to this equation: a popularity index that is a function of the weekly renders of each app over the number of it’s total installations.
In short, the formula looks something like this:
Total Score = Base Score + Popularity Score
Base Score = Score (total installs) + Score (weekly installs, adjusted for weekly uninstalls) + Score (weekly renders in canvas and profile views)
Popularity Score = Score(weekly renders / total installs)
We hope this gives you a clue to the “mystery”. We look forward to hearing your comments and feedback on the forum!