Sizing a study, fast

Sizing a study, fasthave a defensible rule of thumb ready

For qual, you size to saturation, not to a magic number. For quant, you size to power. Be able to say both cleanly and to flex the number to the decision's stakes.

• Qual: roughly 5 sessions per distinct segment to surface major themes, extending until new sessions stop producing new codes (saturation). A higher-stakes decision buys more sessions, not a different rule.
• Survey / quant: size to detect the smallest effect that would change the decision (the minimum detectable effect, MDE) at a fixed confidence and power, then report the precision you actually achieved, not just the point estimate.
• The honest move: when the recruit is hard or the clock is short, run fewer and flag the confidence level rather than pretending small-n is conclusive.