Monthly Archives: May 2015

Techstars: Sleep analysis

So I’ve published a book book[1]. It’s really Jon Bradford’s. After all, it was him who suggested that I blog every day. Then Techstars asked him if they could use excerpts from my blog in their [blog series][techstars blog] about what it’s like to go through the programme[1].

Jon’s response was:

[…] Actually we would love to turn this in a full ebook […] and then it would make sense to publish extracts.

Dom – What were you thinking?

Now, to be fair, I am in the process of writing a proper book that I’d love to get properly published, but I read Jon’s email as:

Dom, turn your blog posts into an ebook.

So I did. It’s free and, when I last looked, listed on the best sellers page for the “startups” section. People have also elected to pay for it. By my reckoning this makes me a best selling professional author. I’m pretty sure Chris would say it makes me a douche.

Ultimately it’s a tidied up, corrected and properly formatted rehash of the blog posts I did while I was at Techstars. But it got us thinking. There’s some interesting data buried in it.

Each day has a time it was posted, which I can roughly correlate to my bed time. Most posts were written in bed as the last thing I did. I would literally post, shut the laptop lid, put it on the floor next to the bed, roll over and die. Where I didn’t have proper times (Friday nights and weekends mainly) I could take a good guess at them from my regular routine.

I also knew what time I woke up each morning (basically 6am for most of them). From that we can estimate how much sleep I was getting. And then we can chart it.

What’s interesting is that I wasn’t actually getting by on as little sleep as I thought I was. The actual amount of sleep in a day varies wildly, with some nights seeing me have as little as 3.5 hours, and as much as over 15 hours (although that was in a few chunks). Bung a linear trend line on the data though and that starts at just over 7 hours sleep a night and ticks up towards 8 hours sleep a night as the programme ends. Which sounds like I slept quite well.

A chart showing how much sleep I got on any given night at Techstars

Indicative sleep durations for each day I was at techstars

Slice the data another way, however, and you get to see how much of a sleep deficit I really built up. For the first 2 and a half weeks I built up a deficit of one full nights sleep a week! This then levels of until after Christmas where the deficit start increasing again, peaking at just over three and a half nights sleep lost.

A stacked area chart showing the increasing sleep decifit

The running sleep deficit (assuming I need 8 hours per night)

Its hard to know how my sleep was affected by Christmas. I was at home, so not doing insanely late night, but I had a young daughter who has me up early and a heavily pregnant wife who would get up multiple times in the night. In the end I’ve treated that period as neutral and ignored it in the data.

It’s also hard to know how much of the time marked as sleep I was actually sleeping for. I seemed to sleep remarkably well in the flat in London, but without a sleep tracker I can’t know how quickly I fell asleep and how much of that was good sleep.

Still, as a set of indicative data it provides an interesting conclusion: if you want to have, on average, a fairly good nights sleep, go to Techstars.

In future posts I hope to do things like sentiment analysis on each of the entries in the book and correlate that to the amount of sleep had, and the current deficit I was running. It will be interesting to see fi there’s any correlation there.

1 So technically this is me blogging about Techstars blogging about me blogging about Techstars… It’s turtles all the way down.