TL;DR
The short version
There is no clean "X% you, Y% your environment" split for workplace performance, and anyone selling one is overselling. Stable individual differences and situational/systemic factors are roughly comparable in magnitude, both modest (effect sizes cluster around r = .10 to .40), and they multiply rather than compete for a fixed share.
The person side was revised sharply downward in the last five years: the cognitive-ability number every textbook quoted fell from .51 to about .23 once a long-standing statistical error was fixed. The operator's takeaway isn't a percentage. It's that situation strength is the lever you control: the same person's traits pay off about twice as much when you give them discretion.
Synthesised from a fact-checked sweep of the personnel-selection and person-situation literature (Sackett et al., Funder, Judge & Zapata, Bloom & Van Reenen).
The person side is weaker than the textbooks claimed
For two decades the received wisdom was that general mental ability correlates about .51 with job performance, the single best predictor in the selection toolkit. That number has been corrected to roughly .31 in the older data and only .23 in 21st-century samples. The downward revision is driven mostly by fixing the math, not by worse workers: the range-restriction corrections used in meta-analysis "were often inappropriate" and inflated the old estimates.
Sackett, Zhang, Berry & Lievens (2022); Cambridge / Industrial and Organizational Psychology
Complexity rescues the person side at the top end: cognitive-ability validity for job proficiency runs .50 / .44 / .32 for high / medium / low-complexity work, highest exactly where knowledge work lives. Personality is comparably modest. Stable traits predict occupational attainment with an average standardised beta around .23, on par with IQ and childhood socioeconomic status, not dwarfing them.
The environment is real but no larger
The instinct to swing the other way, to "blame the system," doesn't survive the numbers either. The classic situationist studies (Milgram, Darley's bystander work, Festinger-Carlsmith) produce effects of r = .30 to .40, the same magnitude as the personality coefficient situationists used to dismiss traits. Across social-psychology experiments the average situational effect is only r = .21.
The whole person-versus-situation debate rested on a false premise: that the two compete, so the more one matters the less the other can. People adjust to situations while keeping their individual differences. Where environment does show up as separable, measurable variance is the firm: management quality behaves like a technology that causally moves productivity, and the employer is a distinct component of individual pay, though naive estimates of that firm effect are biased upward.
The interaction is the lever
The most useful finding isn't a main effect, it's the interaction. In a meta-analysis of personality and performance, all five major traits predicted performance better in "weak" situations (unstructured work, high discretion) than in "strong," constrained ones. The effect is large: the right context roughly doubles a trait's predictive validity.
This is the empirical spine under "remove yourself as the bottleneck" and the empowered-team model. It also reframes the question: "what percent is environment versus you" has no clean answer because a large chunk of the variance lives in the interaction, and much more is within-person, the same employee performing differently across time, which is routinely (and wrongly) treated as noise.
A statistical warning, and what it means for you
Be skeptical of any dramatic single number. Stop squaring correlations: reporting "r = .30 explains only 9% of variance" is a rhetorical move used to belittle real effects, and an effect of r = .40 or greater is genuinely rare in psychology. A predictor at .30 isn't weak, it's near the achievable ceiling for human behaviour.
- Selection has a low ceiling. The best single individual predictor caps around r = .3 to .5 even for complex roles. Hire well, but don't expect it to be where the leverage lives.
- Situation strength is the controllable lever. Traits pay off about 2x in high-discretion contexts. Remove constraints, grant discretion.
- Management practice is an installable productivity input. Treat it as technology, not personality.
- Don't attribute one quarter's output to the person. Within-person variance over time is large and easy to misread as a stable signal.