Can Apple and Google predict your next app download? Thanks to social media and humans urge to share almost everything on Twitter and Facebook, the companies might actually predict your next app download. The prediction may not be a 100% accurate but 45% accuracy doesn’t sound all that bad. A team at MIT’s Media Lab are analyzing social media chatter of a select group to predict their next app download.
The MIT team has given Android running phones to 55 students who are living on campus. These phones are metered, meaning, the researches can gather data of the 55 students right from phone calls and text messages to facebook and twitter updates. This data was analyzed in the following manner :
- Find user’s significant friends
- Find the apps, user’s friends use
- Predict if the user will be interested in his or her friends apps.
Summify for app downloads
Much of it works on the similar lines as Summify, the story recommendation engine. If you sign up for Summify and sign into your Twitter and Facebook accounts, Summify recommends 5 stories for you every day. These stories are the most read stories by your Twitter friends. Now do you want to read the stories your friends are reading? Sometimes yes and sometimes no. What summify propagates is ‘group think’. Instead of finding your own stories, you end up reading what your friends are reading. So you are as good as your friends are. Is that something you really want? Probably not. Is there a better way? Not that I’m aware of.
App downloads is much different from reading interesting stories. The approach of social interactions and app predictions might work just well for app downloads than it did for interesting stories. There is a significant time and energy gains, if your friends have found an app, are using it and liking it. You don’t have to invent the wheel again. News stories have a limited shelf life and apps have a greater shelf life. Recommendations can go a long way in app downloads.
Are the app recommendations effective?
The researchers at MIT are finding it difficult to find out the optimum social network. In their sample size a pool of 821 different apps were used and the recommendations based on this data turned out to be 45% accurate. Complete random guesses with no algorithm turned out to be 10% accurate – much like what Apple and Android show up as recommendations in their app stores.
So yes. Your next app download can be predicted. In fact many app recommendation engines are working on this technology.
Your friend and 20 of your Twitter followers have just download “Angry Birds” and are liking it very much. May be you should try it too.
May be I will download an app based on that recommendation. With the number of available apps across app stores fast reaching a million figure, these recommendation and discovery engines become all the more useful.
