7 Points on “The End of the Email Era”

"The End of the Email Era", a Wall Street Journal article by Jessica Vascellaro ignited, somewhat ironically, a flurry of "have you read this?" emails in my inbox. I'm a bit late to the party of dissecting Vascellaro's piece. All of last week, I was cranking on a new version of reMail. Yet I felt I'd write about it, since I feel pretty qualified to comment on email-related topics.

WSJ's 4 Points

In case you haven't read it, here are the points that the WSJ article makes:

1. IM is better than email because it gets you faster responses.


2. Twitter and Facebook updates are better than email because they're informal and fun.


3. All these updates will cause even more overload and filtering needs to improve.


4. Facebook gives you context about people's location, mood, and current activity. You need to coordinate less than if you were using only email.

Gabor's 7 Points

Most people misread the WSJ piece as "email is dying". But email isn't dying, it's being complemented by new modes of communication. And despite Paul Graham's warning about "lists of N things", here's my list of 7 things to contribute to the social network updates vs. email debate.

1. Twitter and Facebook updates are orthogonal to email. Looking through the last 200 tweets on my Twitter feed, I didn't find a single update that I would have sent as an email had Twitter not existed. The use cases are too different. Thus, Twitter is a parallel world to email.
   
   This HuffPo article puts it best:
   

   If you're like us, you still send text messages on the weekends, check voicemail at work, post photos to Facebook, watch viral videos on YouTube, and Tweet your favorite news.
   
   In other words, we haven't "killed off" our previous tools: we're actually adding, not abandoning, platforms. And when we do ditch, it's because of forces more complex than seasonal trends (or the news cycle).

   
   
2. Email is private, Twitter is public. Twitter and Facebook can't replace email because they're public or semi-public communications channels. Direct messages in Twitter and Facebook messages are bad, low-fidelity clones of email functionality. You shouldn't use them.


3. Your work email belongs to your employer. You can't use Facebook for work. The messages and the intellectual property you create while at work belong to your employer. If you leave the company, you shouldn't be able to take them with you.


4. Email is about task management. The reason why your inbox is a source of stress and your Twitter feed is not is because email is a task manager. Twitter and Facebook are entertainment. Your boss wouldn't assign a task to you via a Facebook update. But if your boss sends you an email, you better read it and get that work item done.


5. The unread messages counter. Unlike Twitter, email has an unread message counter. If it didn't have that counter, email would make you far less anxious. But it would lose its work value as a task manager.


6. The future of email is not to become IM. Part of the value of email is that it's asynchronous: While you're getting actual work done, new messages pile up. You don't want to give everyone the chance to interrupt your work flow. You wouldn't get things done. And that's exactly the problem with turning email into IM, whether it's with push notifications or Google Wave: Yes it will get you answers instantly, but it would make everyone less productive.


7. The lack of innovation in email is because the underlying protocols suck. If you have a great idea about how to use or display the data in Twitter, all you need to read is the Twitter API docs. If you have a great idea in email, you need to know MIME (the encoder), SMTP (the message protocol), IMAP or Exchange (the access layer), and your email client (the viewer). The email technology stack is huge, wobbly, and antiquated.
   
   Take IMAP: a hugely inefficient, stateful protocol with an ugly message format. State-of-the art in the late 1990s, yes, but if you were to reinvent it today, you could do a much better job.
   
   We need to make it easier to innovate around the mail client. We could rip out everything (maybe save for SMTP) and build a great new stack that allows fast iteration. Make it easier to move the needle in email, and the needle will move.

What Makes an Email Important?

In my last post, I said it would be good if email clients only notified you of important emails, rather than popping up a toast for each email that arrives. One of the commenters asked me to point to some research about this topic.

What makes an email important? In this Microsoft Research report [1], the authors have conducted surveys of email usage inside Microsoft. One of the questions they asked was "When is an email particularly important?". Here are the responses:


Note that 5 out of the 10 factors are directly related to who sent the email. (This would indicate that filtering or auto-classifying emails by sender could be very effective.)

I have a bunch of other interesting research results to point to when I have a little more time. If you've read anything interesting recently, please point me to it in the comments.

[1] Gina Danielle Venolia, Laura Dabbish, JJ Cadiz, and Anoop Gupta. Supporting email workflow. Technical report, Microsoft Research - Collaboration and Multimedia Group, September 2001.

If you find this interesting, you should also read HappyMail.

A Model of Your Inbox

I've been thinking a lot about how users view and manage their inbox. This post is about the model I came up with. I'm basing this on my own experience and on having talked to dozens of people about their email habits. But since this blog also is read by lots of email enthusiasts, I'd love to hear your feedback: Does this make sense? What am I missing?

Four Quadrants

Every single email message youreceive can be classified into one of the four quadrants below. Important emails are the ones you need to take action on. Urgent emails are time-sensitive. Urgency does not necessarily imply importance: Your coworker's cake will be gone in a few minutes, but it's not necessary to take action on that, especially if you're on a diet.


The key insight here is that the stuff you care about are the emails on the left. These are the emails that make it worth checking your Crackberry every few minutes. They emails that keep you awake at night.

Filtering the Important

How can you filter out what's important? My theory is that for humans, that's actually the easy part: You often easily determine importance by just looking at the sender of the message. Roughly, senders fall into three categories:

1. Crap: these are easy to remove: iTunes Receipts, Amazon notifications, LinkedIn emails, frequent flier statements, and the like. Not important.


2. VIPs: You know who they are! Major customers, your manager, and your girlfriend belong in this category. Important.


3. People you know: These are the non-VIPs that you still want to deal with. I like to think of this category as the intersection of your Facebook and LinkedIn connections. Mostly Important.


4. Everyone else: Recruiters and salespeople, senders you don't recognize. Mostly not important.

Research indicates that people use senders as their main importance indicator [1, 2]. The task of filtering out important email is easy, and you could probably train a classifier to very high degrees of accuracy.

Managing Later vs. Now

I believe that this is the heart of email overload. Remember how I said that the left side is what matters. I like to label the two subcategories as "Later" and "Now" [3].


The workflow should be that you decide whether you want to deal with an email now or later. You respond to the "now" emails and they disappear into the archive. The "later" emails haunt you until you're done with them.

This is the point where today's email clients fail. Users try various mechanisms to manage now/later and to do/done: Keeping emails unread, starring them in Gmail, and filing away stuff that is done into an intricate foldering system. The number of email management strategies that users come up with impressive [4].

But none of this works. Emails that are unread and starred disappear from your main view. Out of view, out of mind. Poof. There is no pressure to act upon them. In contrast, filing away emails that are done requires an amount of discipline that few users have. I think that the inability of managing now/later and to do/done is one of the main reasons for email overload.

Wrapping Up

Let's review. I've outlined three theories here:

1. Your incoming all fall into one of the four quadrants of urgency and importance. What really matters is the important stuff, which breaks into "Later" vs. "Now".


2. It's easy to filter important vs. not important.


3. It's hard to manage Later vs. Now because the current tools are broken.

I'd love to hear your views on my theories. Drop me a line or leave a comment.

References

[1] Gina Danielle Venolia, Laura Dabbish, JJ Cadiz, and Anoop Gupta. Supporting email workflow. Technical report, Microsoft Research - Collaboration and Multimedia Group, September 2001.

[2] Olle Baelter and Candace Sidner. Bifrost inbox organizer: giving users control over the inbox. In NordiCHI '02: Proceedings of the second Nordic Conference on Human-computer interaction, pages 111-118, Aarhus, Denmark, 2002.

[3] The "Getting Things Done" school of thought has a more intricate system than this, but I think that Later / Now is the essence of it.

[4] Steve Whittaker and Candace Sidner. Email overload: exploring personal information management of email. In CHI '96: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pages 276-283, Vancouver, British Columbia, Canada, 1996.

Upcoming Talks: Geneva, Sydney

I've been invited to give two talks while I'm traveling around the world:

• University of Geneva, Switzerland
  Tuesday, September 16th 2008, 5 pm (Unitec Seminars)
  Auditoire du Rez, Bat A, Battelle, 7 rte de Drize, 1227 Carouge (map)


• Macquarie University / CSIRO, Sydney, Australia
  Wednesday, October 15th 2008, 11 am (HAIL Seminars)
  Building E6B, Eastern Road, Macquarie University (map)

In both places, I'll be talking about the Future of Email and how we can improve email clients far beyond where they are today. I'll also chat about my experience with startups in Silicon Valley and what lessons can be learned from here. The talk in Sydney will focus on the former, the one in Geneva will focus on the latter.

I'd love to see you there! I hope both talks will lead to inspiring discussions.

Thanks to Andrew Lampert and Matthias Kuhn for organizing these talks.

How Researchers are Reinventing the Mail Client

For the last 10 years, the three-pane has been the standard view of looking at email. A pane for folders, a pane for folder contents, and one showing the selected email. Even though mail clients are highly configurable, this has been the standard view of many users. It isn't likely to change soon: The beta of Microsoft Outlook 2007 – pictured below – sticks with conventions.


Email today has many annoyances. Even though we now seem to have a grip on the spam problem, many users are suffering from email overload: There are just too many emails flooding the inbox. Many are drowning in heaps of emails that aren't even important – it's just a colleague at work Cc-ing everyone evenly remotely connected to his project.

There are plenty of ideas on how to improve the current state of mail clients, and I'll present some of them here. None of this is my work: I'll give references to publications of others. There are literally hundreds of papers on this subject, so I've chosen to present my selection of personal favorites.

Here are the three ideas I'll present:

• task-driven email organization
• smart organization structures
• cool new features

I'll present one example from each category.

Task-Driven E-Mail Organization

People's lives today are organized in their mail client. It's not just communication that takes place here: Meetings are organized, lists of todos and deadlines are exchanged, documents are sent around.

In effect, what you're keeping track of in your email client are tasks. Most emails you get are part of some project, belong to an event you're attending or organizing, or are part of a greater plan, e.g. keeping in touch with a girl.

That's the idea behind TaskMaster [2], a tool developed at PARC in 2003. All your emails, drafts, attachments, and bookmarks are mapped to "thrasks". Emails in the same thread are grouped automatically, but the user still has to assign other mails, links, and deadlines manually.

Thrasks can have associated actions, such as "call this person", and "review this". You can also add deadlines to each task: they are shown as green and red bars as they approach. Documents can be previewed right inside TaskMaster's UI, as seen with the Word document on the bottom.



I think the great advantage of this approach is that items that belong together are displayed together. Instead of using email folders to hold related messages, the central element is the task, with all the associated deadlines, todo items, and documents.

Here's a quip from the paper's usability interviews:

"It's just nice to be able to have the control over mixing [...] related things together, even though they might not be [...] the identical kind of thing."

What if we went a step further and looked at workflow patterns? For example, at a company where you interview candidates in a formal hiring process, you get automatically generated messages reminding you of the interview, requesting feedback after the interview, and a notification of the final decision. In the future, we might be able to automatically identify the structure of such processes [4] and classify email into these activities [5] – both of which goes beyond Taskmaster's model, which requires some manual effort.

Creating Smart Organization Structures

Almost everyone I know keeps incoming email entirely in the Inbox. Newly arriving messages join the 500 messages already marked as unread and are displayed at the top of the pile. Is there a better way to organize this view? Can we sensibly restructure incoming mail?

Bifrost [6], a plug-in originally conceived at Lotus Research, that takes this approach. The idea here is that the people are the main indicators of whether an email is important. After installing Bifrost, you're asked to sort your contacts into five groups: Your own email addresses, "VIP Platinum" (extremely important people, e.g. your manager), "VIP Gold" (important people: friends and family), as well as small and large distribution mailing lists.

Bifrost then reorganizes your inbox and displays your email in a number of predefined categories:

• Timely: Emails that contain today's or tomorrow's date in the subject line. They'll likely be important today, but not next week.
• VIP Platinum: emails from your manager.
• VIP Gold: emails from friends and family.
• Personal: replies to emails you've sent out, emails sent directly and only to you, and any unclassified emails you receive.
• Small distribution: Intended for group messages.
• Large distribution: Large-distribution mailing lists.

Below is a mock up of what this looks like in practice. (I had to draw this up myself as the screenshots in the paper were too small).


This structure is helpful in identifying important messages and weeding out the less interesting ones. A quote from their user interviews:

"If I am running through an inbox, I might be tempted to read a title and get sucked in because it is interesting. Whereas if it is in a pile of listserv stuff, I just ignore it altogether. That was a nice thing when I was busy, to not get distracted by unimportant mail."

It's interesting to note that except for differentiating small and large distribution messages, this approach can already be replicated in today's email clients. You can simply create search folders or message filtering rules which simulate the Bifrost behavior. However, this would put emails into folders and wouldn't offer the one-page overview that Bifrost has.

Cool New Features

ReMail [7] was a project at IBM Research that ran from 2001 through 2004. It was basically a reimplementation of an email client from the ground up and had several cool new features. I'll describe two of my favorites below.


Thread Arcs visualize relationships between email messages. Instead of wasting lots of space with a tree view that Thunderbird has, it displays the thread structure in a little image. This feature helps you see where you are in a long conversation. For example, in the picture below, emails B, C, D, E, F, and H are all direct replies to A, while email G is a reply to E.



The advantage of thread arcs is that you can see the position of the email you're viewing in the larger conversation, without having to switch to a tree view: your main inbox pane remains sorted by arrival time.

Contact Maps offer a different view of the address book: Senders from which you have received email are grouped by domain. Each person's name is shown with a different background color, depending on the time of the last email exchange. This offers a better view of your contacts than the traditional non-grouped lists where your least important contact looks just like your most important one.



Many of ReMail's other ideas can be found in today's popular clients: Instant messaging is now integrated with Gmail, which also groups emails by thread. The collection mechanism in ReMail is semantically equivalent to Gmail's labels. Outlook integrates emails and calendaring and has list separators ('today', 'yesterday', 'last month'), just like the ReMail prototype.

Conclusions

It seems like the ideal email organization tool would be like your personal, smart secretary: It knows what's important or interesting, and deals with stuff you don't want to be bothered with. That would be perfect.

Today, we seem to be at a point where it seems like we might be able to solve the spam problem. But the problem of figuring out which of the non-spam emails is important, and what it relates to, still exists.

One solution – the one I presented here – is to add nifty features to the mail client. But would all these features really be understood and used? Users today seem to be using a very basic set of mail client functionality. Anything we add should not only solve a painful problem, but also be easy to use. I'm not even sure this applies to the applications I've shown here: You don't know until you've tried.

What do you think? Are these good ideas? Would normal people who are drowning in email use these features? What features can't you live without? Post a comment and let me know.

--

Thanks to Keno Albrecht, Bálint Miklós , Markus Egli, and Fabian Siegel for reviewing drafts of this.

--

References

A more thorough and academic overview of the subject:
[1] Steve Whittaker, Victoria Bellotti, Jacek Gwizdka: Email in personal information management, Communications of the ACM, 2006

Examples of task- and activity-based systems:
[2] Victoria Bellotti, Nicolas Ducheneaut, Mark Howard, Ian Smith: Taking email to task., CHI, 2003
[3] Michael J. Muller, Werner Geyer, Beth Brownholtz, Eric Wilcox, David R. Millen: One-hundred days in an activity-centric collaboration environment based on shared objects, CHI, 2004
[4] Nicholas Kushmerick, Tessa Lau: Automated email activity management: an unsupervised learning approach, IUI 2005
[5] Mark Dredze, Tessa Lau, Nicholas Kushmerick: Automatically classifying emails into activities, IUI 2006

Bifrost:
[6] Olle Bälter, L Sidner: Bifrost inbox organizer: giving users control over the inbox, NordiCHI, 2002

ReMail:
[7] ReMail: Reinventing Email Website, Collaborative User Experience, IBM Research, 2003