You don't always need to measure a metric to improve it
July 13, 2022You don't need a dashboard until the difference you're trying to measure is too small to be casually obvious.
I had a nice discussion with a friend and former colleague about yesterday’s email about DORA metrics.
She asked about my experience using DORA metrics to improve team performance, and I related a story, about the company where the two of us had previously worked together.
At this company, when I joined, the development team typically released new code into production every three to four weeks.
By the time I left, they were able to release within an hour on demand.
I didn’t have any dashboards to tell me this. I just knew from casual observation.
Often this is all you need, especially at first.
The moral of the story?
Don’t bother wasting time building dashboards for DORA metrics (or any other), until you know you actually need them. And you usually don’t need them until the differences you’re trying to measure are so small that they are no longer casually obvious.
Slow stable releases or fast unstable releases?
The real choice: Big pile of poo, or small, managable piles of poo?
Audience question: What will be delivered when?
How do you create a plan for a team as to what will be delivered when?
Why would you ever need to deploy new features multiple times per day?
That's not why we do CD. When using CD, large features are typically behind a feature flag.