Monday, December 31, 2018

Uncatalogability

At one point, I could make a list of everything I owned. I could even carry it all! That was neat. However, that doesn't work too well when you're older than a college sophomore. Nowadays, I could not make a list of everything I own. Even categorizing it gets into long-tail problems; I can fit a lot of things into "clothes" or "books", but then there's a lot of weird one-off things. The airlock jar topper that I use to make kimchi that lets fermentation gases bubble out but doesn't let stuff in? Where does that go?

I could probably at one point make a list of everything I spent money on, too. Magic cards, a couple of Star Wars figurines, and ... I don't know, that's about it. That ended probably by middle school. Now, even with credit card records and Mint, it's hard to do.

I'm gonna call these domains, where I can't make a list of them or fully categorize them, "uncatalogable." This makes it difficult to reason about these domains. Examples of difficulties:

1. how much do I spend on restaurants? Is it like... 10% of my expenses? How about, which ones are for social things with friends vs when I just don't feel like cooking? Should I try to eat out less? And if I did, would I save $100/year or $5000/year?
2. are there weird things falling through the cracks? Did Amazon fine-print sign me up for some new nonsense that charges me $7.99/month for extra Prime TV channels or something?
3. how much did I pay, the last time I booked a hotel? (I don't remember when or where that would have been.) I have very little idea of how I'd find that.
4. when I move... how long will that take? How many boxes will I need, and for what?

One partial solution comes to mind: search. This has worked for email, which long ago became uncatalogable. But that only solves problem 3 above (and is hard to apply broadly, e.g. to "things I own"). Problem 2 is mostly solvable by being sorta vigilant on Mint or my credit cards - I'm pretty good at that, I think.

But problem 1 is pretty unsolvable. And that (low-key) worries me. I feel like I'm just blundering around in the world, just trying not to own or buy "too many" things. Right now, my current sense of "too many" means I am saving a good chunk of money and I am not deluged with things. So I mean, that's fine, I guess. But it's also just coincidence - if my job paid less, or if we had more room in our house, I'm not sure I would know how to compensate accurately.

2 comments:

Daniel said...

Re: 1: last year, Killian downloaded our card records as CSVs from the bank and manually sorted them. It was a pain, but doable in ~1 day?

Mint doesn't do it with credit cards? That's annoying.

Idea: look through the past month and assume it's a representative sample? Or the past 1-2 weeks?

I'm pretty distressed by 2, myself. These leaks are probably a negligible amount, but it bugs me to imagine little remoras clinging onto my bank account.

Dan said...

I mean... I can get a csv from mint, and indeed I did! It was hard to figure out a couple things though:

1. what all the things were - they're all lines like "CC * SQUARE ** ALS SAN" and then I have to think "huh, did I go to an "Al's sandwich shop" around this time? or maybe I donated $10 to a friend's ALS fundraiser in San Francisco?" I'm sure you know this feeling. Each line is a mystery that can be solved, but given a year's worth of data, it gets kind of unmanageable.

2. what the categorization scheme should be. Mint has a scheme that has about 20 categories by default, and they auto-categorize stuff, so that saves 80% of the work... but then you have to keep realizing when they got it wrong and go back and fix it and then export the CSV again.
Then I end up with "... wait, did I put bike parts in the Sports and Recreation or Public Transportation category?" So like, I should be able to tweak the categories as I go, which you kinda can, but it's also a big pain.

(this is a special case of another interesting problem: categorization tools that learn as they go but let you adapt stuff like this, to make this experience not suck! also good for: labeling training data. I have no idea what this software would look like, but I would like someone to build it.)

I'm a little worried about just doing a month, because there's things that don't really repeat on a monthly schedule (like travel.) Yearly is pretty safe though. A month is probably pretty good to catch the remoras.

The remoras are awful! Isn't it a thing in England that everyone just pays straight with bank accounts, and then you can get an audit of "what are all my remoras" just by asking the bank? If so, that sounds nice.