Why Your Metrics Suck

I 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.

Because they're pseudoscience? No, wait, that's this article.

Metrics 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.

The 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.

Simple 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?