[_rerr]

In my previous company, they has a name to distinguish these two kinds of metrics.

- Inputs are the things employees can act on easily, like the number of LoC written or the number of features developped.

- Outputs are related to business success, for example the number of customers, or the revenue.

The key here is to consider the external business environment as a complex machine, in which you put some effort, measured by inputs, and you observe the outputs as a result (reward). Because you don't know how the environment is going to respond, running the activity consists in adjusting the measures of inputs according to previous results, and trying to maximize those.

This is sane for business because executives can focus on maximizing the input metrics, without caring too much about the outputs, and evaluate separately the assumptions they made about the business (e.g. that x number of features requested by customers developped within 1 month increase the userbase).

It kind of looks like a reinforcement learning process actually.

[humanrebar]

> This is sane for business because executives can focus on maximizing the input metrics, without caring too much about the outputs, and evaluate separately the assumptions they made about the business...

That's basically the problem right there. If you treat the employees as a giant black box, you don't observe when workplace policies (that might not even work!) get in the way of human decency, let alone helping a diverse group of employees each flourish in their own way.

Employers instead sit around wondering why there aren't enough Latino engineers without even considering a remote work policy that would let engineers live in El Paso, Phoenix, Los Angeles, or Miami.