There are no gold medals in DevOps
“Oh, so this dashboard shows the lead time for changes from the first deployment to the development environment and production, not the time since the first git commit and production?”
I can see this DevOps manager frown with discontent at the realization that they were not getting a dashboard that exactly reflected the DORA metrics.
“That’s going to be a problem…”
I fight the urge to sigh. This kind of target fixation is, unfortunately, all too common. Teams see the latest DORA report with eye catching graphs and easily consumed categorizations like Elite, High-Performing, Medium, and Low-Performing and believe that getting changes from commit to production in under a hour is the purpose of a DevOps team.
This misses the point of targets. All targets are wrong, but some are useful.
The purpose of targets, like those measured by DORA and DevEx, are not to literally deploy multiple times per day or get changes to production within an hour, but to remove the hiding places where bad practice and inefficiencies take root.
领英推荐
If you think about it, do your customers actually sit around waiting a few hours for the next production deployment to unblock their progress or unlock a new workflow? Short of fixing a critical issue, the answer is no.
The point of a metric like mean lead time for changes is not because your customers demand a steady drip feed of new features and improvements throughout the day.
The point of these metrics are to highlight the structural changes that must take place in your teams and processes to improve their performance. Being able to get a commit into production within an hour means tightening or removing pull requests, replacing change advisory boards with reliable tests, automating the entire CI/CD pipeline, and implementing feature toggles for fast rollbacks. Of course, no one wants to have to measure all of these incremental changes, but years of research have shown a metric like lead time for changes meaningfully reflect their combined improvements.
It doesn’t even matter what value you measure from a metric like lead time for changes, because the actual value is not important. DevOps is not like running a race where a fraction of a second difference can lead to a gold medal. Every improvement that DORA metrics seeks to embed into your DevOps teams can be achieved by knowing only if the metrics are moving in the right direction, without ever knowing their absolute value.
There are no gold medals in DevOps, so fixating on the target misses the point. The ability to measure even a rough approximation of a good quality target is more than enough to shine a spotlight on the hiding places lurking in your processes and to then begin implementing meaningful changes.
Now I just need to work out how to say that in 15 words or less.