There are a lot of things to think about when choosing metrics, and most groups tend to get paralyzed debating measures and ultimately measure nothing.
You Improve What You Measure
So the risk of not measuring anything also means you risk not improving anything.
The above saying is also there to caution against measuring the wrong thing since it’ll improve that wrong thing really well.
A classic example is velocity. If you make measuring velocity a big deal, it’ll magically increase! That has, sadly, nothing to do with what you likely wanted, which is getting things done sooner.
List Your Major Decisions or Interventions
A way to get started is to list major course-altering decisions or interventions. Answering the question, “What would I need to see to change course or intervene,” helps fuel some ideas.
Then you can figure out precisely the few things you need to model to see that moment that would require the intervention or large course change.
Some examples that I tend to harp on a lot are around operational thresholds like API request/response times, concurrent loads, test execution times, etc.
By setting a threshold, I know we have to change course because it’ll soon become a disaster if we don’t. I’ve chosen my intervention point.
What Good Looks Like
Another way to pick a metric is to ask, “How do I know we’re winning?”
You can probably come up with a few ideas there that will turn into quick-to-model measures. With a few measures here, you can ask of just about everything if it will move that needle or not and when we expect to see the needle move one way or another.
If you’ve ever worried about scope creep or that the project might miss the mark entirely, measures tied to this question can help you correct it quickly.
The idea of this article isn’t to solve the whole problem with measures but rather give a step you can take. There is always more. But if getting a measure is a constant struggle, try to come up with just one measure using these questions. Get the habit in place, then grow from there.