A Year(ish) of Emails

I have a difficult relationship with email. 

On the one hand, it helps me to avoid meetings - a constant and passionate fight. I really, really hate (most) meetings, and trading a few emails is much nicer than shoving ten distracted people into a room for an hour so they can feel like they did something today. Email also leaves a paper-trail, so nobody has to rely on hastily written meeting minutes that Phil will probably forget to send out anyways.

But while a meeting is a semi-frequent major disruption in any given day, email is an nonstop barrage of micro-disruptions. Every few minutes all my devices light up, and it's nearly impossible to know if its urgent or if Phil just wanted everyone to know he left a half-eaten muffin in the kitchen because he's that kind of guy. Email is like your friend's annoying toddler that won't shut up until you pay attention to it, and even then you've got no guarantees. Ignore it for too long, and a grumpy co-worker will inevitably storm in demanding to know why I haven't responded to their question about how to format a status report (damnit, Phil).

Email has me by the balls and it knows it. I'm currently trying to retaliate with a guerrilla campaign to slowly move our company onto the new messaging platform Slack. It's a great product, but I'm still not sure it won't just add to the mess - a second loud toddler that I have to ignore. Time will tell.

Sent Items

In the meantime, I thought I'd at least quantify the problem. The graph below shows volume of Sent messages in 30-minute intervals, for every day of the past 12 months. You can play with some different 3d perpectives by clicking on a different shape to the right. 

It makes for a pretty picture, if not a very efficient one. The mountain ridges to the right of the valley make it pretty clear that my co-workers won't hear from me before 8:30a. There is pretty good activity between 6-9p as I catch up from all of that email ignoring I've been trying to do during the day. There's also a fair amount of activity between 12-3a - hopefully folks silence their notifications when they sleep.

Received Items

Now on to the inbound emails. Since I recently archived my inbox this data only goes back to August or so. I may try to pull a full 5 years worth of messages at some point, but not today.

It looks like my colleagues are a bit less keen on emailing after-hours. Data clumps more tightly around the 9-5 bands. There's a bit of action in the 12-1a time frame. Some small ridges in the early morning are automated notifications letting me know our software is still alive and kicking.

It's certainly a lot of noise to have to drown out during the day, but I'm actually surprised at how low the volume is - less than 100 emails per day. I think when I checked in 2012 I was well over 200.  This could either (a) represent progress in more efficient/effective communication or (b) mean that people just don't like talking to me anymore.  Either way I'm ecstatic.

A Crummy Visual

What about the visualization choice? It makes for some very pretty pictures, certainly. I think they are also pretty intuitive - people are familiar with time-series line graphs and the "z" time dimension is a very iterative extension on that. But as a tool for visual investigation, it's sort of a disaster. Only the most obvious patterns and aberrations pop out, and 60% of the chart is a sea of scrambled lines. 

A simpler summarized graph below provides insights that would be impossible to infer from the previous graphs.

A simpler 2d summary provided much more insight than the chaotic 3d spark lines.

Here I can see that, on average, my email volume drops pretty dramatically at noon.  I can also see that my email rate at 7p is about half what it was earlier in the afternoon. My co-workers, on the other hand, drop down to around 25% their earlier rate. This graph also has the benefit of not bringing older internet browsers to their knees, which I'm not certain can be said about the 3D graphs.

Finally - the trusted scatter plot. Each email gets its own dot (17K in total), and the resulting mosaic paints a really detailed picture of the year. I dulled the "Received" messages to gray to avoid triggering migraines.

A dense scatter plot picks out some detailed patterns across both the Time and Date dimensions.

I see that September and October were full of a lot of late nights. I used to step away from my computer between 12-1p, but stopped in mid-October thanks to some DIY Soylent.  Other patterns like Vacations are easy to pull out, too.

Email took my Saturday Morning

At some point in the future I'll get to the meat of this analysis - the content of the email.  I'd love to test some hypotheses about what the most popular business buzz phrases are (so we can build scale and synergy and disrupt the market), which of my contacts swears the most (it's not Phil), and how many more typos I write after 10p (don't drink and email).  

For now, though, I've let my old nemesis Email consume my entire Saturday morning.  What a jerk.