A paradigm for developer productivity (2022)

Software productivity is all but impossible to measure and insanely detrimental. We tried lines of code, story points, key process indicators, and others, but all they do is pervert the process into an exercise in cooperative stupidity. Without metrics, ratings are popularity contests. With metrics, ratings are popularity contests skewed by metric gaming. You cannot get rid of popularity contests. You can only ensure that you punish even more the people whose drives to work on important things is more powerful than their drives for success.

I think measuring developer productivity is fundamentally nebulous because every task is unique and it's hard or impossible to estimate how long something should take unless you have a room full of people do exactly the same task (and even then, everyone in the room's product will be different both quantitatively and qualitatively).

It's like asking "how long does it take to write a novel?"

Measuring developer productivity is trivial because everyone already knows who the productive engineers are. They're the ones who can fix critical incidents quickly, can implement any feature, can fix any bug, and everyone goes to them with questions. The problem is that nobody wants to act on this information because it implies radical changes that no large organization is willing to make.

Metrics for software development are pretty hard unless you are just grinding out the same code again and again. From my experience most productivity can be gained by all stakeholders staying actively involved and making quick decisions. I guess that's what Agile was supposed to be before it got perverted into micro-management hell.

It is imposible to measure developer productivity. Software development is more like research and it's output (software) IS the increase of productivity - someone else's productivity usualy.

The majority of buisnesses do not understand the economics of software so I am not surprised we live this productivity metric chaos

Looks like people once again are trying to make very complicated measurements that micromanage workers instead of using classical approaches like "The iron triangle of project management" - good, fast and cheap.

Why is there no tool that shows a triangle with features, bugs and support quality. Because it would obviously show that there are no new features, terrible support and tons of bugs. Everyone is doing everything except the job that needs to be done. They do it to meet the requirements of newly created 20 indicators. And who is to blame for this ? Obviously not management, they just sit and watch indicators waiting to send those who are underperforming on the bar charts to the HR department for a little talk.

Splendid