Separate Volume, Speed, Quality, Cost, and Outcome

Five types, and the classification exists for one reason: teams report five volume metrics and call it a scorecard.

Volume. How much. The easiest to measure and the least informative on its own. Speed. How long. Usually the second thing anybody adds. Quality. How good, which almost always requires defining good and is therefore the one most often skipped. Cost. What it took, frequently absent entirely, which is how a process that works and is uneconomic survives for years. Outcome. Whether the business result happened. The hardest, and the only one that is not a proxy.

Not just activity

There is a line worth quoting almost directly from a real specification of what to measure: completion rate, latency, freshness, exception rate, override rate, and a business outcome appropriate to the workflow, not just model or automation activity.

That final clause is the argument of this whole guide compressed into eleven words. Activity metrics are easy, plentiful, and they measure the thing you control, which is exactly why they crowd out the one that matters.

The effort ceiling

An analysis of conversion against early effort shows the expected thing in two segments: more touches, higher conversion. It also shows the two segments plateauing at very different levels, and names the cause as intent rather than activity. In one segment, effort keeps paying. In the other, it stops paying well below.

A metric that measures the input you control will always suggest doing more of it. The useful analysis is where doing more stops helping, and you can only see that by measuring the outcome alongside the effort.