The research is coming from Douglas Aberdeen, Ondrej Pacovsky, and Andrew Slater in Google’s offices in Switzerland, so it’s a tad self-serving, but the research team seems to have tackled their topic with some serious scientific objectivity.
The problem Google’s Priority Inbox faces is a tricky one, from a philosophical point of view: How do you prioritize the incoming stream of emails to a Gmailer’s account without “explicit user labeling”–where users’ input is required to train the code? There’s an extra spin on the problem caused by the fact that the nature of email is constantly changing (so working out what’s important and what’s not is tricky), there’re terabytes of reference emails to store and trawl through for training Google’s algorithms, and “users disagree on what is important, requiring a high degree of personalization.”
It turns out one of the most important ways Google works out what emails are important to users (versus just interesting) is how quickly the user interacts with it in the moments after its delivery. The other key metric is whether users consistently mark their messages in a particular manner–when Google detects that you’re doing this, it uses the information to adjust how it prioritizes information to you. By applying some bulk decisions to all mail flowing through Gmail, then adding a layer of personalized processing to each user, Google’s worked out how to make prioritizing work.
How well does it work? The research found that compared to Google employees without Priority Inbox, PI users spent 6% less time reading email, and 13% less time reading unimportant stuff in their inboxes. PI users were also “more confident” to bulk archive emails, or delete nonsense.
Considering how many trillions of dollars of business rely on email as a tool, and that the volume of spam email is running at ridiculous levels and growing fast (some put the figure at 97% of all email traffic), any system that means we as users spend less time on dealing with the unimportant trivia in inboxes is a good thing. Plus this research gives us a small, but rare insight into the clever technology that works behind Google’s smooth, featureless user-facing interfaces.
To read more news on this, and similar stuff, keep up with my updates by following me, Kit Eaton, on Twitter.