Data saturation
Data saturation is the point in qualitative research at which new interviews or observations stop producing new information relevant to the research question. It is the most common basis for deciding that a qualitative sample is large enough.
How it works
The idea began as theoretical saturation in grounded theory, described by Glaser and Strauss in 1967, and spread to qualitative research generally. Because it is judged, not counted, researchers increasingly report how they assessed it.
One method, published by Guest and colleagues in 2020, sets a base size, the number of interviews whose unique themes form the denominator; a run length, the number of later interviews in which new themes are counted; and a threshold for new information, such as 5 percent. Runs overlap, each shifting forward one interview, and saturation is reported at the interview before the first run at or below the threshold, with the run length appended. How varied the population is and how narrow the question is decide when that happens, so no single sample size fits every study.
A worked example
Suppose a hotel group's loyalty team sets a base of 6 interviews, which yield 40 themes, a run length of 3 and a 5 percent threshold. Interviews 7, 8 and 9 add 2, 2 and 1 new themes; 10 and 11 add none. Run 7 to 9 holds 5 new themes, 12.5 percent of 40; run 8 to 10 holds 3, or 7.5 percent; run 9 to 11 holds 1, or 2.5 percent. Saturation is reported as 8+3, 11 interviews in all.
Related
Automate this in Pulse
Because Pulse interviews need no scheduling, a study can run a few interviews past the point a team expected saturation, confirming the last ones add nothing new before recruiting stops.