In December of 2019, I started recording how many pomodoros I did each day. I had no expectation that this would last a long time, but it's now been over a year! I've kept a spreadsheet of my daily pomodoros since then, and here's the graph. I did 2171 pomodoros between 2019-12-01 and 2020-12-29, for an average of 5.5 per day.
This post is a retrospective. As such, there's not much of a narrative, just a lot of little details. Hopefully it will be useful to someone.
In case you're unfamiliar, pomodoros (sometimes abbreviated to pomos or poms) are an extremely simple work technique. You set a timer, usually for 25 minutes. You do focused work during that time. When the timer goes off, you take a break, usually for 5 minutes. You repeat this however much you want.
This almost sounds too simple to take seriously, but LessWrong is no stranger to using literal timers to solve problems. The purpose of doing pomodoros is to increase productivity by controlling your focus. Some people are naturally bad at taking breaks, and some people (like me) are bad at focusing on work. The former can lead to mental fatigue or general attentional inefficiency; the breaks force a person to rest their mind. For the latter, setting a timer makes it easier to "stick with it" for that duration. There are a more specific ideas and recommendations from the official inventor of the technique, but honestly it's so simple that you're better off just experimenting with it for a few days, and picking the parameters that work best for you.
For the purpose of my own data recording, I defined a pomodoro as a 20-50 minute period, where I actually set a timer, and worked on an endorsed productive activity. If I zoned out for too much of the time, I didn't count it.
The graph definitely encodes some broad trends, but there is quite a bit of interpretation to be wary of. There were many times when I did work, but didn't time it. This could be because I forgot, or didn't realize I was going to do a block of work until I was halfway through it, or because I was collaborating with someone else and our dynamic didn't need the structure, or because I was doing some physical work where I didn't really take breaks and having a timer around would be awkward. There was quite a lot of this, and I really wish it was more practical to record all the work I do. At the very least it would make me feel better. There were of course also many times where I worked the whole pom, but did so at a glacial pace, constantly having to force myself to refocus on the task, and getting virtually nothing done for my efforts. Pomodoros are not a measure of output. Lastly, while two poms generally equal an hour of work, the duration do vary.
It's a little weird to have the interval be able to span more than a factor of two. But a pomodoro is intended to delineate a time span of unbroken concentration, and depending on what you're doing (and how your brain works) this can be a pretty variable duration. If I'm having a bad day and I can just barely get myself to sit still, I might go for 20 minutes, and if I'm working on high-overhead focus work like refactoring a codebase, I might go for 50 minutes. In practice though, almost all of the poms I did were 25 minutes. I'm not sure why. It is the standard recommendation, and if you're also timing breaks, then doing 25 on and 5 off leads to lining up with the hours, which is useful. I think I also feel something like, I want to get more points on my chart but I would feel bad about doing the bare minimum. This was enough to keep me sticking to 25, while there was never quite an incentive slope that pushed me toward doing 30 minutes or more.
I don't force myself to take the whole break, or to always set a timer for the breaks. Sometimes I work much past the end of the pom, and for me, that's considered a success. Sometimes I'll stop, stand up, walk around the room, and then come back and start another one. I always try to at least stand up before starting another pom.
Some days I think I have to do this task anyway, I might as well get some points for it and then I end up doing sporadic poms for a while. Other times I would decide to work for several hours, in which case I might make sure to time the breaks as well. And also I often co-work with others, in which case it's actually pretty important to time the breaks, because otherwise I can easily spend half the time socializing.
In the beginning, I used a timer in Chrome by just googling "25 minutes timer", which gives you a timer widget on the results page. Surprisingly, I always found this slightly annoying. I didn't like having another tab to keep track of, and when it went off I didn't like that I had to go searching for it while it kept on beeping at me. If I was doing work in another window, then I couldn't see how much time I had left without tabbing over to the Chrome window, which can do a lot toward breaking my concentration/momentum. Lastly, when you do get to the timer tab, you have to actually click the tiny "stop" button. You can't just hit the spacebar like on a youtube video. I also tried using the timer app on my phone, but that has problems of its own, not least of which is the fact that it causes me to continue to be near my phone and look at it often while working.
These details feel sort of petty when I write them out, but in the domain of habit engineering, you need to squeeze out every effect you can find. If I can just barely get myself to do work, then I need to remove every possible obstacle, even if that obstacle is "I have to click a button".
So after struggling with the google timer for a while, I went on Amazon and bought this physical timer. It completely solved all those aversions for me. It takes a flick to set it, I can easily see how much time is left with the blink of an eye, and when it goes off it beeps for three seconds and then stops. This is the perfect length to make sure I hear it, but not long enough to be annoying. It has a quiet/loud switch, so if I'm listening to music with headphones I can switch it to loud and still hear it. It's also perfect to use when co-working, because then everyone stays synced and there is mutual knowledge that the official work time is over, without me having to personally interrupt them.
And lastly, I just like objects. I think the timer looks aesthetically pleasing, and I like the feeling of having "equipment". I don't entirely understand this, but buying the timer got me to do more pomodoros, and that's what matters.
I generally try to continuously improve and iterate on my systems for getting things done. In the case of pomodoros, there were three other parts of my system that played a role in causing me to continue doing poms.
I've known about the pomodoro technique for more years than I can remember. But I didn't feel anything especially positive about them. The idea sounds... kind of lame, to be honest. You just, time yourself doing stuff? So the idea itself wasn't enough for me to start doing them regularly.
I'm part of an accountability group where, every two weeks, we each send in our goals for that cycle to an email thread. At the end of the two weeks, we report on how we did, and send in our next goals. It's pretty simple, but it's surprisingly effective. In addition to setting object-level goals, I've also used this mechanism to try a lot of different structural goals, like "Do a mid-day check-in where you write down how you're doing at 3PM each day" or "Rate how well you adhere to your schedule" or "Co-work four times this week". And one day, seemingly at random, I decided to try the goal of "Write down how many pomodoros you do every day". Notably, the goal was not "Do four pomodoros every day" or even "Do pomodoros every day"; just write down how many you do, even if it's zero. I then proceeded to do a bunch of pomodoros, and consequently a bunch of productive work. I ended up liking it enough to keep the same goal for the next cycle, and then the next cycle, and then by that point it was just a thing that I did, and there was no point in taking up my goal list space with it. So without the accountability group, I probably never would have even tried setting that goal, and or followed through with recording how many I did, or had anyone to report it to to motivate me.
A while after, I added a reminder app to my system. Getting prompted to do tasks is a part of GTD, and it was clearly a missing part of my personal system. I spent a while trying out lots of different apps for it. I added a daily morning reminder that just says, "Record how many pomodoros you do today". This still goes off every morning and is regularly the reason why I do my first pomodoro.
The last part is co-working. Co-working itself was somewhat of a productivity breakthrough for me, although I still haven't figured out how to get myself to do it very frequently. I virtually always use pomodoros while co-working, so that I and the other person can alternate getting work done with check-ins and socializing. The person is usually over for a few hours, so that results in a good chunk of poms for the day.
The graph has fairly clear periods of more or less pomodoros. (If someone knows any time-series stats I could run on it, let me know.) As you can see on the graph, I wasn't doing many pomodoros during the first few weeks. I knew I liked it but needed another structural change to bump up the rate. I had figured out that co-working was a great way for me to be more productive. I didn't have a full-time job, and I really wanted to find a way to establish regular, habitual co-working. So I made the following post on Facebook.
...And that worked, actually. I legitimately did not think it would. But I soon found four(ish) people who were willing to meet with me once a week at a regular time and co-work using pomodoros for several hours. It was great! I got more socializing and more work done. The rise on the graph in January and February comes from this.
And then the pandemic arrived, and I had to stop co-working with people. Virtual co-working with them didn't catch on. The dip in mid-March is from stressful COVID negotiations, and then moving into a different unit in the house.
...And then immediately after that, I got way more productive?? Or rather, I should be more careful with that wording; I started recording way more pomodoros. Apparently, like many others, I found the isolation and lack of options useful for sticking to a routine. This didn't last the whole pandemic, but it did give me my longest streak (96 days).
Because of the uncertainty generated by the pandemic, I decided that I really should start doing software work again. So a lot of those pomodoros in the midyear high period were me filling out applications or doing coding problems to practice for interviews. Using the accountability group, I set the goal of doing a coding problem every day. I ended up joining a freelancing network and taking gigs. For this work, I didn't submit hours; I just got paid to do 20 hours per week, and did weekly check-ins with the client. For some reason, this led to me mostly not doing that software work under pomodoros. I would just get up and think, eh, I guess I should start working and then do some work. I would stop when I couldn't convince myself to keep working on it. This isn't how I want to feel about work, but it got the job done. So I transitioned to doing arguably more productive work, but recording far less poms. I'm currently in a period where I do pomodoros regularly but not every day, and I don't have a particular goal with them.
It was also interesting to see the histogram of pomodoros per day. The most obvious pattern is that 0 is the most common number of poms. Even though I did poms on the majority of days (83%) those counts are spread out among all the non-zero numbers. The second interesting thing to notice is that the histogram is virtually flat between 1 and 7! That means that, on a given day, if I have already done a single pom, I'm just as likely to stop there as I am to do one more, two more, etc. up to six more, and in fact slightly more likely to do more! This is a strong argument in favor of adding something to my system that induces me to do that first pomodoro. After seven, the graph trails off in a statistically unsurprising manner.
One major thing that surprised me was how few poms I was doing while still being as productive as usual. In theory, a "full" 8-hour day of work could be broken up into 16 pomodoros. You can see on the histogram that I rarely did this many; over the entire year, there were only 11 days where I did 16 or more poms. A day where you wake up and do nothing but work could be 32 pomodoros; there were definitely days where I subjectively did nothing but endorsed poms all day, yet the most I ever did was 19. I have previously tried a week of working literally all day, and while that felt like obviously more work than 19 pom days, I'm very surprised that it was 68% more work.
Now, the whole reason I'm doing this is because I'm not nearly as productive as I think I need to be, so it's not hugely surprising that I wouldn't be clocking 8 hours of focused work on an average day. But the difference is big. And more importantly, after now being quite calibrated, I can tell you that after doing 8 pomodoros (~4 hours), I got a perfectly normal amount of work done in comparison to the days where I worked a full time job. After realizing this, I actually felt quite angry about the way society structures work. My feeling was something like, are you telling me that for the last several years, I could have been getting paid the same amount of money to do the same work, but in 4 hours per day instead of 9?? (I am not the first person to notice this.) In any case, the finding is interesting, and I would love to compare notes with others who have done poms over a long time period.
Pomodoros work very well for me, and they are now part of my system indefinitely.
Setting the goal of writing down how many pomodoros I did every day caused me to do more pomodoros. Doing more pomodoros caused me to do more work.
The gamification of having a pretty graph to watch and wanting to "make number go up" was a noticeably motivating factor.
Pomodoros on their own are not enough; I need to leverage many other pieces of my productivity system to make them work for me.
Weirdly, despite doing thousands of pomodoros, it's still not a habit. What I mean by that is, when I start my work for the day, I don't just start a pomodoro without thinking about it. I often need the push notification reminder to remember to start using it. And since that particular stimulus is now quite stale, it's often not enough for me to start one either. This needs further debugging.
In a bizarre coincidence, I just noticed that seven years minus two days ago, another person named Alex submitted a post with the exact same title.
(I noticed because the slug my post generated had a -1 at the end.)
We have had all the conversations worth having, and now we are resigned to have them all again. :D
Some ideas are just naturally high on rederivability.
I was working through Alex Vermeer‘s ”8760 Hours” pamphlet today which linked to that post and actually read it today. I was very confused when I saw it show up on the frontpage again today (or so I thought).
By another bizarre coincidence, I planned to start doing and recording pomodoros in 2020.
(For proof here's the blog post from yesterday where I mention it though it's mostly about other things. Also I realize now I published this post after yours but I just checked lesswrong.)
If someone knows any time-series stats I could run on it, let me know.
There's a few different things you could be interested in here for 'time series segmentation', which slightly shift what sort of method you want to approach.
Often, people use ARMA models for time series because they can easily capture lots of different local factors, and HMMs for structural breaks (when you have as many as you do). I'm not aware of standardized methods that are good for this problem because often there's lots of tweaks inherent to your distribution; take a look at all the detail in this accepted answer, for example. But you also might be able to stick your data into seglearn and get something cool out of it.
Here is an analysis of one person's 10 years of using Pomodoros, including the data: http://shape-of-code.coding-guidelines.com/2019/12/15/the-renzo-pomodoro-dataset/