When Google introduced its Priority Inbox last year, it was with the promise that the system could help us manage our important email more efficiently. Now Google has published a paper on how the Priority Inbox learns what’s important, and on how much time that may end up saving users.
The research comes from Douglas Aberdeen, Ondrej Pacovsky and Andrew Slater from Google’s Swiss office, and the data comes from Googlers’ usage of Priority Inbox – so take the results for what you will. But according to the findings, Priority Inbox users spent 6% less time reading mail overall, and 13% less time reading unimportant mail.
The paper addresses some of the challenges that come with designing the Priority Inbox’s ability to learn what’s important. The system looks at several elements, including the social features – the relationship between email sender and recipient and content features – what the email contains and how the recipient has responded to similar content in the past. The system also looks at the speed with which a user responds to an email in helping it just what’s important.
Google has developed models to help predict people’s responses, both at a global and individual level. The paper notes that “scaling learning to millions of users is as dif?cult as tuning the algorithms for a single user. To store and serve models, and to collect mail training data, we make extensive use of bigtable, which combines features of a distributed ?le system with a database.”
Spending 6% less time reading email sounds like a great result. Do you use Gmail’s Priority Inbox? If so, do you think it’s offering you a substantial time savings?
