> If we as an industry are going to move away from headcount as a professional growth metric, we need to have some other, more direct way of measuring the business impact a manager delivers so we can reward that.
A manager’s job, fundamentally, is to ensure that their reports are delivering appropriate value to the company, so the most direct measure of a manager’s performance is the performance of the ICs that report to them (possibly indirectly). If you’re going actually to use this, however, you need a way to fairly evaluate the performance of ICs that doesn’t rely on input from their management chain which is itself a hard problem.
> If you’re going actually to use this, however, you need a way to fairly evaluate the performance of ICs that doesn’t rely on input from their management chain which is itself a hard problem.
Why would you need to decouple that to evaluate managers? TBF at manager level this can be a rather simple and robust “how did the org subtree rooted at this person fare at contributing to the stated business objectives?” Along team performance, this depends on staffing matching the expectations. Negotiating which are core responsibilities of a manager.
> Why would you need to decouple that to evaluate managers?
If a manager’s performance evaluation is based primarily on their reports’ evaluations, then there’s a conflict of interest: The manager is incentivized to inflate their reports’ evaluations because doing so directly improves their own evaluation.
You're approaching your stated problem wrong. See:
> A manager’s job, fundamentally, is to ensure that their reports are delivering appropriate value to the company, so the most direct measure of a manager’s performance is the performance of the ICs that report to them (possibly indirectly).
Now, you're arguing that `sum(evaluate(impact) for each IC)` is absurdly bad. Which it is, because you're compounding errors in individual evaluations. These are hard to do, because you're trying to distribute team's achievements between contributions from particular people. I say that `evaluate(sum(impact for each IC))` is easy and robust. In fact simpler than an evaluation for a single IC, because you don't have to distribute anything and a lot of things that are external distractions for ICs fall into responsibility of managers.
> Now, you're arguing that `sum(evaluate(impact) for each IC)` is absurdly bad.
No, I’m not. I’m arguing that it’s unwise to trust someone to do an evaluation who has a personal stake in the result of that evaluation.
Looking at how well a team works together to advance company goals is an important factor in any evaluation system, but using that as the only factor in a manager’s evaluation is as shortsighted as evaluating a software engineer solely on how many LOC they produced.
In practice, overall team performance should be a factor in IC evaluations as it’s reflective of various hard-to-quantify contributions of the team members. It should also be a somewhat larger factor in a manager’s evaluation, with the weight between IC and team performance shifting from mostly-IC at the bottom of the hierarchy to mostly-team at the C-suite level.
You don't need to use the individual evaluation of ICs, you use the overall success of the team in delivering whatever it's their job to deliver. If nobody but the manager is capable of evaluating that, somebody has really messed up.
A manager’s job, fundamentally, is to ensure that their reports are delivering appropriate value to the company, so the most direct measure of a manager’s performance is the performance of the ICs that report to them (possibly indirectly). If you’re going actually to use this, however, you need a way to fairly evaluate the performance of ICs that doesn’t rely on input from their management chain which is itself a hard problem.