Take back your Inbox!

Disclaimer :- The opinions expressed below are purely mine and a couple of friends. None of us are connected to xobni, or anyone working there.

This is the punchline for a stealth mode startup called xobni (inbox spelled backwards) currently based in San Francisco, California. What do they do? They do email analytics. That is, trying to analyse emails with help of stastics and machine learning. The founders are Adam Smith and Matt Brezina from MIT. Backed by khosla ventures and YCombinator.

That’s the only info i had about them from their website. I just wondered, what enhancements do i need in my email? My email flux daily is around 20-25 on an average. I recieve all kinds of mail from newsletters, forwards, friends, family, office and tonnes of spam. When i open the inbox in morning, i see almost 15 new mails daily. As i go through them, i am in a hurry to finish the reading. Thus, i do not reply to any of the mails, even if it is important (Reading takes a priority over replying). After i am finished reading mails, i jump on to my daily dose of RSS feeds (can’t resiste those either.). In this process, i many times do not reply to mails that i should (blame it on my laziness), and then they are lost.

I started thinking what could they be doing? Later i was invited for dinner at Pratyush’s place, where me, RG, Pratyush and Mitra discussed it further. Following is the summary of that.

Some of the desirable improvements in email per se are.

1. Ordering of emails :- I need the ordering of new mails in such a way that mails that i am most likely to reply, appear on the top. Of all the unread mails, i need an option (a tab would be great), that shows me the mails that i should reply to.

2. Autoclassification of mails :- This feature can be like smart filters. Mitra works with me. An email from him asking for a review on th spec of new upcoming product should be classified different than an email that is an invitation for a party in the evening. It should be able to classify the newsletters from general mails.

3. Modes :- There should be a concept of modes. During daytime, office mode is on (which displays the official mails with priority), and in the evening, i switch it to personal mode(which should display the mails from the family and friends as priority).

4. Information Box :- Feeds, IM’s, Social networking should be integrated with the email. It is annoying opening google reader, then gmail, then yahoo messenger, then facebook for all these purposes.

5. Optimizing delivery time :- Every person attends to her email at different times. Pratyush’d replies generally come around 11PM. RG’s replies come around 6AM. Mitra’s replies come around 11 AM. Based on this info, the mail delivery to these people should be scheduled appropriately.

6. Automated followup notifications :- Many times i seek some information from Mitra. He being a busy man, sometimes forgets to reply(same problem that i mentioned in point 1). I should be served with timely notifications on such mails, in i anticipate a reply, and reply does not come within the average period that the other fellow takes to respond.

Many other things can be built around email. It is the still the primary, trusted and valuable means of information delivery. All our registration confirmations are done through email. All invitees that we send are sent through email. All social networking sites use email as propogation medium. One thing that comes to me is that information in To: CC: fields is very important and valuable. In a way, it is worlds oldest social network. We can infer who knows whom from the mails within these fields, subject line and real message.

Availablity of API’s would be a great help. If those are available, i can write a widget that sits on my desktop and starts blinking whenever i recieve a new email from my fiance , or i can write a widget that shows me the email on my desktop on top of the topmost window (IM kind of experiece). I’m sure that other develpers in the world can use these API’s effectively and in better fashion than my silly ideas.

Aah! this takes me few years down the line back to grad school. The machine learning course instructor asked us to modify mozilla thunderbird. The challange was to tweak the ordering of emails. The ordering should be such that the messages that we are most likely to reply, should appear at the top.

Text mining has given us some very usefull tools like support vector machines, which i feel can be utilized to solve these problems. Out of other classification strtategies, things like Conditional Random Fields, Bayesian tree networks can give better solutions.

Please suggest any new features that you need in your email. Lets build a comprehensive list.