Productivity was never one number

When leadership asks for "the developer productivity number," the kindest thing an engineering manager can do is hand them the SPACE framework paper, published in ACM Queue by Forsgren, Storey and colleagues. Its argument fits in a sentence: productivity has five dimensions, satisfaction, performance, activity, communication and efficiency, and any single metric is a distortion of the whole.

The paper's sharpest insight is about the dimension everyone reaches for first. Activity, the countable stuff, commits, pull requests, story points, is the easiest to measure and the least meaningful alone, because activity without outcomes is motion, and activity under measurement becomes performance of motion. The authors recommend metrics drawn from at least three dimensions, at least one of them perceptual, which is the polite academic phrasing of: you must actually ask the humans how the work feels.

Managers sometimes hear that as unrigorous. Our experience runs the other way. Survey data about friction (where do you wait, what do you dread, what would you fix first) predicts the diagnostic's findings better than any single telemetry stream we pull, and the telemetry then confirms and sizes what the humans already knew. The perceptual data is not the soft complement to the hard numbers. It is the map that tells you which hard numbers to read.

What we would take from SPACE in practice: pick one question per dimension, keep the set small enough to fit on one page, and hold the set stable for a year so trends mean something. The tradeoff is that a five-dimensional answer will never fit in the single cell of a board slide, and someone will keep asking for the cell. The reversal condition does not exist for the framework itself, only for its ceremony: a five-person team that talks daily needs the paper's humility, not its instrumentation.