You know how it is: all metrics can, and will, be gamed.
So using metrics is always a balancing act: how much can you learn from a single metric, how resistant is it to being gamed, and does it have a counterbalance that avoids pushing the organisation too far in one direction?
The trick is to not use a single metric, but a system of metrics. That way, by improving any single one through gaming, you inevitably create a counter-signal in the other metric.
Let’s take a simple example: maybe you want to measure deployment frequency. Obviously, the more often you deploy, the better.
Except now people get sloppy, and push out poorly tested and badly thought out things.
But what if you started measuring another quantity as well: change fail rate.
Now, if you go too far in one direction, the counter-metric will warn you: deploy more quickly than your system can stomach, and change fail rate will spike. Be too cautious and aim for a low change fail rate – your deployment frequency will plummet.
Thus, when considering metrics, see if you can find another metric (or combination of metrics) which counteracts it. Not only will it make your processes more robust, you’ll also learn something.
(Photo by Jean Scheijen from FreeImages)