5 ms·
Why Your Metrics Suck
- jsankey 17y agoMetrics suck because they are indirect measurements, so by optimising the metric you may not be optimising what you intend. There's no reason why a more complex metric will necessarily be more directly aligned with a goal. This thinking suggests that by adding more complexity your metrics become more aligned, but it seems these patches could also take you further and further from anything real.
- metafunctor 17y agoBy a "complex" metric, I didn't mean that the metric is necessarily complex to compute or represent. Rather, the complexity of the metric is a measure of the breadth and complexity of the activities which have an effect on the result. Take operating profit, for example. It's fairly straightforward to compute (not as simple in practice as one might imagine, though), and it's just one number. Pretty much anything going on in the company has an effect on the bottom line, in one way or another. It's the result of complex activity. Obviously, this was a poor choice of words. Any suggestions for a better word?
- jsankey 17y agoI see your point: your complexity relates to the system being measured rather than the measurement. But I'm afraid I don't have a single word that could make it clearer.
- mildweed 17y agoBecause they're pseudoscience? No, wait, that's this article.
- metafunctor 17y agoLOL :) I wasn't trying to make it science-y, although now that I look at it it sure comes off like pseudoscience. The graphs just struck me as a good way to explain what I had in mind...
- rwolf 17y agoMultiplying some random functions together caused me to twitch, but I like the "metric of the month". Having got to the end, I think the hand-wavium was an okay way to relate your idea.
- jsm386 17y agoThe argument is an interesting one--the more complex the metric, the harder it is to act on it--but the author provides no reason for this other than his graphs, which are representations of...nothing? Simply drawing a line to represent his thesis doesn't mean his thesis is based on data; it just means he knows how to graphically represent his thesis.
- acgourley 17y agoI don't agree - usually the simple metrics are not actionable because they are in the vein of "uniques per month" which is not something you can direct act on. Specific metrics (breakdown down uniques by cohort, behavior, etc) are actionable, but end up being complex to describe and collect.
- foulmouthboy 17y agoSimple metrics or complex metrics won't get you anywhere unless you have the proper context around them. Why should I care about this metric? How did this metric get calculated? What am I supposed to do now?