Proxy Metrics

This is part 3 of our Melzaminder Series. Previously we introduced the Mindful Munching with Metrics program followed by To Weigh Or Not To Weigh, in which we argue in favor of poor maligned scale weight as a useful getting-in-shape metric. Here, we generalize the idea of metrics that aren’t quite the true thing you care about and yet still make for good Beeminder graphs.

Pop quiz! What is a metric?

(a) It’s short for the metric system.
(b) A fancy way to say “a lot”.
(c) A set plus a distance function satisfying the triangle inequality.
(d) Something you can measure numerically and make a graph of.

Got your answers locked in? If you said (a) then no, that “metric” is an adjective; ours is a noun. If you said (b) then no, you’re thinking of “a metric f***-ton”. If you said (c) then you probably thought you were winning all the nerd points but, alas, that’s a metric space. Nice try.

So, yes, the correct answer is (d). We’d be kind of surprised if anyone reading this blog didn’t know that, since we’ve been talking about metrics for a decade and a half. (See, for example, a classic one from over ten years ago on how to pick a good metric.) But if you’ve been reading along in confusion this whole time, now you know.

Alright, and a proxy metric? We talked about a classic one last time: scale weight is a proxy for fitness (or, less indirectly, is a proxy for bodyfat).

A lot of our own beeminding is proxy beeminding. Like how our User-Visible Improvements goal is a proxy for inexorable forward progress and never giving up. When what you really care about isn’t straightforward to measure and track, you can often find another metric that is and beemind that instead. Like Danny O’Brien’s browser hack. He wrote some code to automatically beemind how many websites he’d visited in a day as he found it correlates well with his distraction. More websites = more distraction.

Here’s another example: I track my vegetarian days, and in the past have tracked sugar-free days. What I really want is to reduce my overall meat and sugar intake. But Do Less goals can be fussy, with the Pessimistic Presumptive Reports and such (plus they’re a premium feature). Also keeping track of number of servings of sugar or meat becomes too easy to get vague about. If I don’t remember to update the moment I eat a cookie, it’s very easy to forget later in the day if I ate 1 cookie this afternoon, or was it 10? But sugar-free or meat-free days are easy and binary. Either I ate of the forbidden food or I didn’t.

In theory I could eat twice as much sugar on sugar-allowed days and defeat the whole point. But that would be gross, so in practice this goal succeeds in reducing my sugar consumption just fine.

(UPDATE: Since I wrote that, I switched to a Do Less goal for sugar, set quite generously, since I love sugar, and just have my “vegetarian days” goal.)

Alright, let’s engage listicle mode and enumerate examples, starting with those above, that we’ve collected with the help of the Discord and daily beemail subscribers:

1. Scale weight as a proxy for bodyfat or fitness.
2. Sugar-free or meat-free days as a proxy for eating less sugar or meat.
3. Automatic tally via browser extension of number of websites visited as a proxy for distractedness.
4. Timestamp of first Withings weigh-in as a proxy for wake-up time.
5. User-Visible Improvements as a proxy for inexorable forward progress.
6. Number of gym visits as a proxy for exercise.
7. Words written about a nebulous goal.
8. Time spent playing piano as a proxy for getting better at playing.
9. Time spent on a project as a proxy for finishing it.
10. Focusmate minutes as a proxy for focused work generally.
11. Number of homemade meals as a proxy for healthy eating.
12. Number of comment actions in Google docs as a proxy for progress on a jointly authored document.
13. Strength training reps as a proxy for strength.
14. Triangular beeminding as an elaborate proxy metric for quantifying the binge-iness of one’s alcohol consumption.
15. Hours in bed as a proxy for getting enough sleep.
16. Handfuls of veggies, or servings of protein, as a proxy for healthy eating.
17. Time spent reading as a proxy for mental health (if you’re Nicky).
18. Micro habits like setting foot in the gym as a proxy for working out (reasoning: lower activation energy).

“All metrics are wrong, some are useful” — Theo Spears

We realize there’s a fuzzy line on what counts as a proxy metric vs just a metric. Some goals are very direct, others are very roundabout, others are in between. Some proxies are just about making a tradeoff in favor of easier data collection. Some goals are inherently qualitative but we find creative ways to operationalize them.

As Beeminder user Anita put it, this is central to Beeminder (and life!). “A lot of what we want is pretty fuzzy — be healthy, be happy, be loved. Choosing what actions to take (what metrics to track) in pursuit of those goals is central to a life well-lived.”

This is also the key to choosing SMART(ER) goals. You have an underlying goal that you distill into something Specific that you can Measure and Attain — also it should be Relevant — in a Time-bound way. (And since that’s merely a proxy for the underlying goal, be sure to Evaluate and Reevaluate it.)