Quantified Self EU 2015 Conference: one year of time tracking

Two weeks ago we went to our first QS conference, the Quantified Self Europe 2015 in Amsterdam. It was a great experience, and an incredible opportunity to meet with interesting people and get the conversation going. I might have finally found a way to resurrect my Zeo sleep tracker! We were also very happy to see how many people got on board with Smarter Time and our goal to track all of our time automatically.

Anis and I led a group talk on time tracking and productivity, which he will comment on in a later article, and I briefly spoke about my time tracking experiments with Smarter Time.

I have gathered more than one year of time data since I started building Smarter Time. That’s 7349 timeslots in 440 days. It took me about one minute per day to input new places and activities and correct the app’s automatic guesses.

The accuracy test

Before analyzing the data, my first step was to check how accurate it is. Smarter Time has a very innovative way of guessing activities, which comes with a small margin of error. To assess it, I measured all my computer activities with Rescue Time for the month of August and compared them with what Smarter Time tells me:

My main activities have very similar durations on both apps. They are slightly higher on Smarter Time because it only measures the main activity for a given period of time, and not the short breaks, such as grabbing a bite, or quickly doing something else.
The one worrisome difference is programming. A lot of my programming time is spent offline, but I still think Smarter Time overestimated it. There is probably an “office hours” effect: unspecified work is classified under the most frequent work activity. I may also have been a bit sloppy in my classification (blame August).
At any rate, we are working on better ways to measure infrequent activities, which we will implement soon.

What the data tells me

Here are the in-app stats of my time use:

Unsurprisingly, I mostly sleep and work. Not enough sport, too much professional stuff. Only 2% of my time was unclassified.
The thing is, data aggregated over long periods of time is not useful so much for the total numbers, as for the evolution and patterns it reveals. One of my main goals with Smarter Time was improving my work life. Let’s see how I fared in that respect.

Working hours evolution

This is my monthly work time, averaged by day. Holidays are represented by the drowsy guy.

Do you see how wavy the curve is? Even if you ignore holidays, heavy months are followed by lighter ones. That goes to confirm the theory that our work time is in limited supply, and longer hours have to be compensated for afterwards.
The pattern is less obvious in weekly data, as it suffers from bigger variation.

Here are my main work activities:

News browsing is one of my main tasks as a CEO. I need to keep myself up to date with everything about Quantified Self, Productivity, Privacy, Mobile, Marketing, Startups… But I realized it was taking up too much of my time. That prompted me to attempt cutting it back to more reasonable proportions — which I have mostly succeeded at.
Programming is clearly my “filler” activity and varies depending on my total work time and other more urgent work duties.

Weekly patterns

I used to spend most of my time working from home. Here is a typical “remote work” week from the app calendar view:

You will notice how I used to wake up late and work long sessions, and how there is no clear distinction between the weeks and the week-ends.

Then in September we moved to a startup incubator and started working with two interns. My rhythm totally changed:

It is nice to be around my colleagues, but the trade-off is that I am usually too tired to work in the evenings.

Here is the breakdown by category for those two weeks:

I work a bit less, but since I have to be more efficient my productivity feels roughly similar.
However my sleeping hours have clearly suffered from the lack of morning sleep, and I feel really flat in the evenings. My entertainment time has increased, but mostly because I cannot work in the evenings as I used to. I also enjoy my off-time less now than before, when it was a choice I made freely.

Leisure and entertainment

Looking at my trends, I was able to match specific games and books with my activity times:

My first reaction was “wait, are video game releases the main factor influencing my work time?” But when I compared them with my monthly data, I realized they are simply what happened to pique my interest during the “down” months, when I was tired and had to recover from more intensive work periods.

What happens next

My goal is for Smarter Time to be an efficient and accurate tool. I want it to record our personal chronology in such a way that it allows us to get relevant insights about our time use and helps us manage it better.
The current version already provides interesting data but we will keep refining it and adapting it to new challenges.

The next big step for us will be to match time data with other, more qualitative factors. You will be able to grade how productive you were at work, how restful was your night’s sleep, or how satisfied you were with your day.

As for myself, I will continue studying my time use and improving it through better data gathering and analysis. I will keep sharing my findings with you on this blog, so stay tuned for more chapters of my statistical life!

Written by Emmanuel Pont