CX Strategic Knowledge · Quality & Research
Participant numbers (qual & quant)
Sample size is a discipline, not a vibe: hundreds for quant, 8-12 per segment for generative qual.
Participant numbers are calculable, not negotiable. A 20,000-person population needs 377 responses for 95 percent confidence at 5 percent error. Generative research needs 8 to 12 participants per segment; evaluative needs at least 5 per segment, fresh each round.
Quantitative confidence has math behind it. For a population of 20,000: 95 percent confidence at 5 percent error takes 377 responses; 99 percent confidence at 5 percent error takes 645; 99 percent confidence at 2 percent error takes 9,084. Tightening confidence and error is expensive, and pretending a small survey carries that weight is self-deception.
- Generative: 8 to 12 participants per segment
- Add an accessibility group
- Four segments means roughly 32 to 48 people
- Evaluative: minimum 5 participants per segment
- Recruit new participants each round
Watch the double standard. When other functions decide from anecdote, nobody blinks. When CX professionals talk to a small, correctly chosen sample, the work gets dismissed as just a few people. That standard deserves to be challenged, in both directions.
Apply this
Reading about participant numbers (qual & quant) is one thing. Seeing where it applies in your journey is the useful part.