[conover]

Output metrics are things you can't control. For revenue, you can't force people to by your products. Inputs metrics are the things you can control. You can (to a degree) control price and selection. And you've established via analysis that those inputs metrics are associated with the output metrics.

A key part I think people miss is to make _sure_ your input metrics and output metrics are correlated. You would intuitively think that, say, lowering outage minutes would be correlated with revenue. And as a result, you should invest engineering and other resources in reducing outage minutes. But maybe experimentally or via some other type of analysis it isn't. Maybe users will tolerate some amount of outage minutes without materially impacting revenue. You can then utilize that knowledge to prioritize. Rather than investing in resiliency to improve outage minutes, you invest in something else because that will have a bigger impact on revenue.