Beeminder Forum

The "100%" dilemma

Newbie time-tagging question here

If you want to answer the question “where do my days go” using TagTime (or any tagging system), it seems you probably want some mutually-exclusive collection of categories that add to 100% of the time you tracked.

On the other hand, it’s nice to know that when I was doing nothing I planned to do (for which I have a tag) I was typically doing that by browsing hacker news, or whatever (as it happens I don’t have a tag for that, but probably I should). If you do that, and you ask “how many hours did I spend on tag X” for all X, of course they don’t add to 100% of the time you tracked.

How do people deal with that dilemma?

1 Like

By the way: one way that occurs to me is to strictly divide up one’s tags into two sets: a set of adds-to-100 categories (of which you must pick exactly one each time you see the TT popup), and a set of throw-them-in-as-you-will tags (of which you can add as many or as few as you like).

Since 1. I’m as lazy as the next person, 2. I find that dissatisfying, I’m hoping there’s some better, nerdier answer that hasn’t yet occurred to me?

I think that there are really two separate things that are being conflated when they are both just called “tags”. The way I envision it, there is “what am I doing”, and “what describes the current state”. I’d like to call only the second “tags”, and give the first another name (like “category”).

The difference between the two is something like noun vs adjective, or verb vs adverb.

For instance, a ping right now would have the category “writing”, and tags including “listening-to-music” and “drinking-coffee”. Writing (this reply) is the main thing I’m doing, but there are additional things that describe my current state.

This more or less is the same as what you were saying about dividing tags into two disjoint sets. I too find it a bit dissatisfying, and I think what I’d want is if TagTime supported this distinction natively. The reason I find it dissatisfying, I think, is because it’s forcing two separate things into what TagTime considers one concept (tags). If TagTime’s prompted you not for just generic “tags”, but for a (single, mandatory) category and optionally also descriptive tags, I think it would be a much better conceptual fit.

(Actually, what I’d really like is hierarchical categories. So for instance “writing>beeminder-forum”, “writing>blog”, etc. Within “writing” all the sub-categories would add up to 100%, then at the top level “writing” and all the rest would themselves add up to 100%, etc. This too can be simulated, I guess, under the current setup with tags, it’s just a bit clunky to do.)

That said, there is a different possible approach: trying to zoom in as much as possible on the very split second when the ping occurred. If in that second my finger was hitting the keyboard, then the tag is “writing”, if I was sipping coffee then the tag is “drinking-coffee”, etc. I like that less, because I’m no less in the process of drinking coffee when I’m holding the cup and about to take a sip than when I’m actually sipping, but that is a possible approach, and done properly it can probably eliminate the concept of ambient tags as distinct from categories of activity.


You could just tag everything: writing, drinking coffee, listening to music.

I do that but it’s not about physically sipping or whatever, it’s what has the focus of my attention at that moment in time.

1 Like

Right, but that feels ontologically inelegant. If you do it that way (and I have, at points in the past) you then look back at the data collected and you have no clear view on really on what was going on, what you were spending time on. This is, I think, really the same thing as the point @halfplane was making above.

I don’t understand - why does that not give you a clear view on what was going on? It tells you exactly how much time you were spending writing, drinking coffee, listening to music, doing any two, or doing all three.

1 Like

For example, I might ask myself: “What do I spend the most time on?” If the answer TagTime tells me is “listening to music”, then that’s silly, because “listening to music” is not something I (mostly) spend time on, it’s in the background when doing other things.

So I’d have to skip over that, and go a few more down the list, to see where I’m really spending my time.

Of course, if I’m at a concert, then I actually am spending my time specifically do listen to music. That’s a fundamentally different thing!

The distinction I’m making (and that I think @halfplane is making as well) is that the amount of time spent doing things (in the sense of “you’ve got this many hours in your day, that’s your budget, divide it how you like between your tasks”) is expected to add up to 100%. You don’t “spend time” (in this sense) listening to music (except in a concert or the like.)

This view (on how the time I actually spent lines up with how I was trying to budget it) is something that I and @halfplane would really like to have, but is somewhat awkward to do with TagTime currently.

1 Like

But it’s not silly. It is something you’re spending time on, even if it’s not your main focus. And it allows you to look at your life and see if that’s something you really want - like, do you want to spend all that time with music on when you’re not giving it your full attention? Maybe you look at that and decide to make a change in your life like “I’m not going to have background music on. I’m only going to listen to music when my full attention is on it.” Maybe you decide background music is harmful and you’re glad to know how much time it’s in the background.

(If you don’t think background music is harmful, consider things like having TV or ads on in the background, smoking or being around secondhand smoke, drinking unhealthy beverages, etc. It might be good to know how much time you’re in different states or conditions.)

But if you don’t consider yourself to be listening to music if it’s on in the background, if you don’t think that’s relevant, then why would you even include that tag? It’d be like including a “breathing” tag for everything.

Except that you can do multiple things at once. Drive and listen to an audio book. Eat a meal and spend time with a friend. People do things like that all the time, so why would it add up to 100%?

I see a couple of options:

  • don’t bother tagging the current state
  • tag the current state with a tag like “background-music,” “state-music,” or “category-music.”

Then you have a bunch of “background” tags (what you call “categories”) that can easily be distinguished by their names.

This doesn’t address the issue of numbers not summing to 100% if you’re doing multiple (non-background) activities at once, though.

1 Like

I think you already understand this, but since you ask the question: First, of course nobody truly records “everything they’re doing” taken to the extreme: for example, we can’t record every detail of our thoughts (and if we tried going very far in that direction we’d get noisy results anyway). So we’re inevitably making a choice in tagging. The most important constraint that I think I should impose on my choice is that it should support my goal of figuring out how I could change how I spent my time to better accomplish my goals. That’s the reason I’m tracking time in the first place. The most obvious way to do that is to end up with a set of tags whose pings add to 100%, so that I can say “if I did less X, I could do more Y”.

I’m not saying there’s no other way to approach this than mine – in fact, that’s exactly what I’m fishing for. I haven’t seen any yet that I really understand, though. In particular, I don’t understand why zzq’s “zoom in” prescription is an alterative. That’s what I do already? But they still won’t add to 100: e.g. I have a workcycles tag and a my-secret-project tag. If I count pings for workcycles and project tags, they’ll add to more than 100%.

1 Like

@dreev, are you saying you basically use only one tag per ping (except for afk etc.) and therefore they do add to 100%? Do you tag in a fairly coarse way – i.e. few tags (say 10 that cover 85% of pings)? I’d be interested to hear what kinds of categories you use – e.g., making this up: UVI, support, top-secret-project, daydreaming, eating? Or are they significantly more detailed than that? I see from the article they at least include sleep, kids, and hacker news :slight_smile:

Oh, no, it just means I have perhaps somewhat less overlap in pings. I still have various state pings and ping hierarchies. I’m very interested in answers to this question! Fun fact: I originally called TagTime “TimePie” because I wanted exactly what you’re describing: a pie chart of where my time was going so I could decide how to reallocate it.

In fact, even funner fact: I saved the chat transcript from when TimePie/TagTime was born in July 2007:

D: Do you track where all your money goes? I keep experimenting with elaborate incentive schemes and such for time management but I think foundational to it would be a way to know where your time is currently going. So here’s my idea: a stochastic self pinger. At entirely random times the program would ask you “what are you doing RIGHT NOW?” And you would answer with a little tag. Then over time it could establish reasonably tight confidence intervals on the fractions of your time you spend at various activities. WRK, FUN, EAT, WEB, etc. It could, like, pop up a pie chart characterizing where your life is going and you could click on the pie slice for RIGHT NOW. There’s probably a big difference between the actual pie and your ideal pie so it would motivate you to move in that direction, because the Self Pinger could ping at any moment.

C: So it works on the “if you write down everything you spend you’ll be smarter about it” school of thought? [or the principle that if you chart your weight over time, you’ll be able to change it]

D: Exactly. I’m thinking an exponential distribution where there’s some fixed probability each moment that it will decide to ping. It has the property that any time from now to forever is possible for the next ping and you can adjust the expected duration between pings by changing that “ping me now” probability.

C: That’s an interesting idea for behaviour modification; relying on our shame, to keep us out of denial.

D: Yeah, it would so work on me. Ooh, especially if you made your pie public.

C: That would be quite a motivator!

D: Ooh, or you could do this with a group of friends; it could ping you all in sync to keep things fair and it would be like a contest. You’d set your ideal pie and then try to make your actual pie come out as close as possible to it.

D: K, hold on. … What’s a better name for this than selfpinger?

C: Hmmm. It should involve a pun.

D: Timepie?

C: I like it.

D: K, stand by again. … [creates repository]

C: And so timepie is born.


No, I don’t understand this. I mean I get the idea that IF you did only one thing at a time, you could look at the percentages and see what you wanted to adjust, moving time from one discrete category to another.

But in fact, that’s not the case. People don’t do only one thing at a time. So how could the categories possibly add to 100%?

I think the answer is that you’ve just got to hand pick a set of mutually exclusive and exhaustive tags and make sure that, whatever other tags you use, each ping gets tagged with exactly one tag from that set. Then you can make that time pie.

An example of a set of mutually exclusive and exhaustive tags might be something like {productive-at-work, distracted-at-work, sleep, nonwork-alone, nonwork-social}.

1 Like

Well, maybe all I can say is that I’m guessing that for my purposes “I do one thing at a time” is close enough to the truth to be a good-enough model to help me change my behaviour (and that more complex models may actually do less well for that purpose).

I think I’ll likely just need to be disciplined enough in my tagging to do as zzq suggests and partition my tags into the adds-to-100 subset, and the “meta” ones like “afk” and “workcycles” (I can think of others like that I might want like “le” / “he” for low energy / high energy, for example, so I can get some information about what I’m realistically capable of doing instead of my normal un-directed default behaviours).

Random tag chat: right now I have a lot of pings tagged with just “n”, which is my tag for “nothing I planned”. I think I need to split that up at least into at least 1. aimlessly browsing the web (soo much time) and 2. other, otherwise I won’t know what specifically I should try to do less of…

1 Like

Right. I think I’ll certainly aim for more detail than that (I already have about 25, most of which are flat categories, and I’m quite sure it’ll grow). Probably I’ll end up with a bit of hierarchy too, maybe not explicit in the tags though. I’m sure it’ll take me a while to figure it out, which is fine.

1 Like

Yeah, just prefix the “meta” tags with “meta-” so you can clearly sort them out. I like the idea of tagging things like your mood or energy level.