[jackhiggs]

I'm an Engineering Manager for a large company and have been for a number of years. For our large team we will rank the team based on perception of managers. After that we will then manually collate coding stats from all repos we work on. Unfortunately for the consensus view here and in the article, you get a 100% hit rate on who you think poor performers are and the lowest coding contributors.

For top performers it's more nuanced. In general they will be top of the contribution stats but sometimes if they're doing R&D or hard work then the stats are not very meaningful. But that's why we don't rely on them.

So metrics have their place to inform and color existing perception. But they will rarely change perception completely.

[LaGrange]

> Unfortunately for the consensus view here and in the article, you get a 100% hit rate on who you think poor performers are and the lowest coding contributors.

This is circular logic. If you measure that "coding contribution" nonsense, people's "performance" will be perceived based on that, _especially_ by their direct managers.

[AgentOrange1234]

I’ve seen cases where folks completely checked out and were contributing nearly nothing, making no commits, writing no code, and faking it at standups. Simple metrics can help surface cases like this.

I agree that it’s something a manager could over-index on. I’m not sure how to avoid that beyond adopting a mindset of “this is noisy data that sometimes gives you important insights.”

[JohnMakin]

> I’ve seen cases where folks completely checked out and were contributing nearly nothing, making no commits, writing no code, and faking it at standups. Simple metrics can help surface cases like this.

Of course we've all seen varying degrees of this - but these kind of people can only exist because of terrible management. Throwing metrics at the problem just introduces a more insidious version of this individual, one that knows how to game whatever metrics are used (managers especially will do this). I've been on teams where such an individual could thrive for years, even with promotions, and on teams where such an individual would be outed within a week.

[AgentOrange1234]

In one of those cases I was the bad manager. The data was a wakeup call that made me understand the magnitude of the problem.

In another case, I knew the manager well, having been on his team before. He was effective, empathetic, and inspirational. He was also overworked and perhaps a bit naively assumed good intent from everyone. The data let me explain to him that the coworker was not contributing and he had a real problem.

[mrguyorama]

If you need metrics to see an employee isn't doing what you assigned to them, what are you even being paid to do?

[AgentOrange1234]

You’re being paid to make your whole team as effective and capable as possible while satisfying your leadership and stakeholders. And to help your boss do that with his team by developing your peers. And…

Detecting slackers efficiently is simply never going to be a top priority. You have to trust your people. Usually it’s incredibly rewarding.

[LaGrange]

Inflating the engineering department head-count, a very important metric for c-level resumes.

[LaGrange]

> “this is noisy data that sometimes gives you important insights.”

This is called "broken clock shows the right time twice a day."

Also c'mon, fooling commit metrics is silly easy, just _also_ one of the most burn-out and check-out inducing things ever. And at senior level you _HAVE_ to do that these days. You _have_ to inflate them.

Your broken clock isn't free.

[lowbloodsugar]

Amazing. You just proved her point with data and then drew the exact opposite conclusion.

[aswerty]

Qualitative and quantitative approaches together inform us best. Probably not the most eye opening of statements.

But I think as you indicate, qualitative generally paints the picture, and quantitative validates it.