Get writings by email.
Get writings and updates by email.
Lessons from Tracking My Productivity and Sleep
May 11, 2015
You can't improve what you can't measure. When I started graduate school at Princeton, I was obsessed with nailing grad school and getting a faculty offer at a top university. I knew that it's going to take an insane amount of effort and I was ready for it. To become more efficient and productive I wanted to track my daily routine. So I started a simple system.

I started tracking sleep (how much, what times) and productivity (out of the last 30 minutes how many minutes were productive). It was just pen and paper records. Tracking productivity was the hardest thing. Every 30 minutes, like a clock, I would evaluate myself. It was often frustrating when all I did in the last 30 minutes was check email or browse the web.
My weekly score cards (2010–2011)
fter a few months, I manually entered raw data into a CSV file and generated plots. The results were shocking.

Sleep and productivity data for 15 months.
This graph is 15 months of data. My target productive hours per week were 60 hours (this doesn't include email, meetings, overhead etc., but actual hours in which I felt I achieved something). My target sleep per week was 44 hours (6 hours on weekdays, 7 hours on weekends).

Here are some observations:
Real data was way off from what I thought it would be. I was surprised at the average monthly productive hours and average monthly sleep, because they were far away from what I wanted them to be, and quite different from what I thought I was clocking in. Lesson: humans are bad at estimating their own performance. There is no substitute for real data.

Sleeping more than a certain threshold makes me unproductive. On days that I sleep more than my target, I was less productive on average. Just having less number of absolute hours in the 24 hour cycle was not the explanation. I was in front of a computer or at a library for long enough, but my productive hours were taking a bad hit. Lesson: I get demotivated by sleeping more and having a bad start to the day messes everything else.

Work spikes are followed by dead weeks
. Whenever I go through a work spike (pulling nighters, doing whatever it takes to meet some deadline), I end up riding the high of the last victory and basically stop being productive for days, even weeks, at a stretch. I could see the spikes and corresponding drops in the plots. Maybe this was a "recovery period", but productive spikes averaged out. Lesson: it's a marathon, not a sprint.
I also started tracking other things like eating three times a day, drinking enough water, having vitamins at night, quality of sleep etc.

Wakemate device for sleep analytics, 2012
Over time, manual data entry was just too painful, even when I wrote software to help me collect data. I replaced manual data entry with RescueTime and highly recommend them. I bought a WakeMate (YC S09) for tracking my sleep but it didn't go that well, since the company shutdown.

More people have started paying attention to personal analytics (mostly health related). However, I feel that this area is largely under explored. I'd love to see applications agreeing on standard data formats and writing data to personal data stores that the user owns.
Most importantly, we need to figure out automated data collection, since it's the single biggest pain point in Personal Analytics, and I hope we get there soon.
Comments? Tweet them @muneeb

Muneeb Ali
Co-founder Stacks, a Bitcoin layer for smart contracts. CEO Trust Machines, building Bitcoin apps. → Learn more