[teahat]

What this article misses out is that the metrics will be present and used whether they are articulated or not. By making them explicit they can be improved, validated, debated, and removed. Standards are consistent. When they're implicit, you get none of those benefits and instead promote an insider culture of unspoken biases that get no scrutiny at all.

The choice isn't between metrics and human intuition, it's between explicit and implicit metrics.

Of course there are bad metrics; to use an example from the article - measuring the output of analysts finding OBL on a binary metric is simply a bad metric. But if we start from there, we can improve it. How many leads do they have at the current time? What stage is each lead at? What intelligence has it generated to this point?