Theoretical Saturation

Thinking Tools3.1 Finding the Current FactsAlso known as Saturation

From grounded theory, developed by Barney G. Glaser and Anselm L. Strauss (The Discovery of Grounded Theory, 1967).

Sets a stopping rule for gathering cases: keep going until several new cases in a row add no new category or property.

Each interview adds less, until a run of empty ones says stop

Forty possible interviews about why dental clinics drop a software trial. Each one either turns up a theme the team has not heard or adds nothing. Set how many empty interviews in a row the team waits for, then play: the themes curve flattens, the empty run climbs, and the study stops when the run reaches the rule. The figure knows twelve themes exist; the team never does. Illustrative.

5

Themes found so far

0: 0: 02: 1: 24: 3: 46: 4: 58: 6: 610: 7: 712: 7: 714: 8: 916: 9: 918: 10: 1020: 10: 1022: 10: 1124: 11: 1126: 11: 1128: 11: 1130: 11: 1132: 11: 1134: 11: 1136: 11: 1138: 11: 1140: 11051002468101214161820222426283032343638400: 0: 0: 1: 2: 35: 4: 4: 5: 6: 610: 7: 7: 7: 7: 815: 9: 9: 9: 10: 1020: 10: 10: 10: 11: 1125: 11: 11: 11: 11: 1130: 11: 11: 11: 11: 1135: 11: 11: 11: 11: 1140: 1105100510152025303540

Interviews in a row with nothing new

0: 0: 12: 0: 04: 0: 06: 1: 08: 0: 110: 0: 112: 2: 314: 0: 016: 1: 218: 0: 120: 2: 322: 4: 024: 1: 226: 3: 428: 5: 530: 5: 532: 5: 534: 5: 536: 5: 538: 5: 540: 50246802468101214161820222426283032343638400: 0: 1: 0: 0: 05: 0: 1: 0: 0: 110: 0: 1: 2: 3: 015: 0: 1: 2: 0: 120: 2: 3: 4: 0: 125: 2: 3: 4: 5: 530: 5: 5: 5: 5: 535: 5: 5: 5: 5: 540: 5024680510152025303540

Interview 40 of 40

Themes found so far
11
Interviews in a row with nothing new
5
Interviews done when the rule stops the study
28
Themes never heard (the team cannot see this)
1

An example

A software company that sells scheduling tools to dental clinics wanted to know why trial accounts went quiet in their second week. The research lead booked calls with office managers who had dropped out, coded each transcript into themes (a confusing setup screen, a clash with the clinic's billing system, a dentist who never logged in), and kept a running tally of whether each call added a theme the team had not heard before.

Before the first call she wrote the stopping rule into the plan: stop after five calls in a row add nothing new. The first ten calls produced seven themes and the next ten added three more. One more theme came at call twenty-three. Then five calls in a row added nothing, and the study closed at twenty-eight with eleven themes, each tied to the quotes behind it. The report gave the rule, the count, and the call where the last theme appeared, so the product team could see why twenty-eight was enough.

Why it works

A stopping rule turns "we have talked to enough people" into a fact anyone can check. Saturation works because it counts categories, not cases. Early cases are dense with new material; each later case is more likely to repeat what the team has already heard, so the yield of new themes falls. When a run of cases adds nothing, the next interview has stopped paying for itself, and the run is the evidence.

Writing the rule down before the first case matters as much as the rule itself. It keeps the team from ending early because the story already feels complete, and from running on because one more interview feels safer. The run length is a dial: a longer run costs more interviews and lowers the chance of missing a rare theme, which is why the figure's rule of seven finds the twelfth.

The thinking behind it

Barney G. Glaser Barney G. Glaser, a sociologist, founded grounded theory with Anselm Strauss in The Discovery of Grounded Theory (1967). The method builds theory out of the data: each new case is compared with the categories already formed, and the next cases are chosen for what they can add to those categories, which the two called theoretical sampling. Sampling for a category ends at theoretical saturation, the point where new data stop adding properties to it. The stopping rule on this page is that idea put into an ordinary research plan.

Anselm L. Strauss Anselm L. Strauss, a medical sociologist, was Glaser's coauthor on The Discovery of Grounded Theory: Strategies for Qualitative Research (1967). The book treats theory as something built from data that are gathered and compared while the study is under way. Its answer to the question every field researcher faces, when to stop collecting, is saturation: a category is finished when fresh data no longer add to it. That gives a qualitative study a stopping point it can defend to a skeptical reader.

How to run it

  1. Write the stopping rule first. Before the first case, set the run length (five empty cases in a row is a common choice) and put it in the plan.
  2. Name the categories being built. Saturation is judged per category, so be clear whether the work collects reasons for churn, buyer roles, or objections.
  3. Code each case as it arrives. Compare it with the themes so far and mark it as adding a new theme, a new property of an old one, or nothing.
  4. Keep the tally where everyone can see it. One line per case on a shared sheet: the date, what it added, and the current run of empty cases.
  5. Choose the next cases to stretch thin themes. Pick interviewees who differ from the last few, such as larger clinics or another region, so an empty run means something.
  6. Stop when the run reaches the rule. Do not extend because one more feels safer, and do not cut early because the story is tidy.
  7. Report the rule with the findings. State the run length, the total cases, and the case where the last new theme appeared.

Where it goes wrong

The usual failure is an empty run that means nothing because the sample never changed. Ten interviews with the same kind of customer will repeat each other and look saturated while whole groups go unheard. Check who the last run came from; if they are alike, widen the sample and restart the count.

A second failure is judging saturation by how often things are mentioned. The fiftieth complaint about price adds no theme, while a single remark about a compliance rule might be the finding that matters most. Count new categories and properties, never volume. The third is setting the rule after the fact, which lets the team stop wherever the story already feels done. If no rule was written at the start, say so in the report and run a few more cases against a stated one.

When to use something else

  • Marginal Analysis. When each further case has a known cost and value, and the question is a budget: keep buying data while the next unit returns more than it costs.
  • Affinity Diagram. When the cases are already collected and the work is grouping what they said into themes.
  • Randomized Controlled Trial. When the question is the size of an effect, which needs a sample size fixed in advance.

Where it sits in SOLVE eX

Go to the source

  • The Discovery of Grounded Theory, 1967

Back to the catalog