how-to · evidence-based answer
How many participants do I need to detect an effect?
The short answer
Run a power analysis before the study, not a p-value autopsy after. The sample you need depends on the effect size you expect, the power you want (conventionally 80%), and your alpha; small effects in noisy organizational data need far larger samples than most pilots have. An underpowered pilot that finds 'no significant effect' has not shown the program failed — it has shown the test could not detect it.
The problem underneath
Underpowered pilots are the default in HR: teams run an intervention on thirty people, find 'no significant effect,' and kill programs that were never testable at that sample size. Power analysis before the study is the fix.
The evidence
- 205+ profiled and searchable works
- Original research bridging the literature to practice
- The source-graded, citation-verified organizational-science registry
- Effect-size / validity findings
Every claim on this site traces to a graded source — see the proof graph.
Go deeper
Related questions
- What sample size does my HR study or pilot need to be conclusive?
- How to run a power analysis for an organizational experiment