Metrics and Targets

A target is not a metrics

In the world of data, businesses are constantly worried about meeting targets. This is not new, and not a uniquely corporate phenomenon. The pirates that worked off the coast of Somalia gave bonuses, including for being the first on the ship, for example. In a corporate job, a target might be finding new targets, rather than armed maritime assault, but the idea is that we want to incentivize higher performance.

What is a metric?

A metric is a way of measuring something. In a hospital, blood pressure and pulse rate are useful metrics. Businesses look at measures such as employee and customer churn, units produced, and the like. The idea is, like a sudden drop in blood pressure, a sudden dropoff in quantity ordered might indicate that the firm is having trouble of some sort. As such, monitoring metrics is a key task of data analysts.

What is the difference?

In short, the diffrence is Goodhart’s Law. My personal favorite example, introduced by Tim Harford’s book “Messy”, is how the APGAR measure went from a metric for newborn babies to the target that drove (at least in part) the rise of C-sections in the United States. Most importantly, this never accuses doctors of actively pursuing the goal. However, simply moving a procedure towards a default selection, rather than an opt-out, can shape outcomes (just as any behavioral economist about nudges).

This has nothing to do with Amazon, in particular. However, in hearing multiple times about monitoring metrics, I wanted to take a minute and say out loud that a target is not a metric. Now if you’ll excuse me, I have to contemplate factor analysis.