Estimation of Sampling Needs
How many interviews are enough? It’s the age-old question in qualitative research. Some say 12. Others say 30. A few insist on 50. Everyone cites “saturation” and hopes reviewers nod politely.
Enter BEST — Binomial Estimation of Sampling Thresholds — a simple mathematical way to put rigor behind that shrug.
Instead of chasing the vague idea of “hearing every possible theme,” BEST asks a clearer question: Have you heard all the themes that show up often enough to matter?
How it works (without the headache)
-
BEST uses a binomial model to calculate how many interviews you need to be confident you’ll capture all themes above a certain frequency.
-
You set the frequency threshold (say, any theme that appears in at least 10% of the population) and a confidence level (say, 95%).
-
Out pops a sample size. For example, 29 interviews at 95% confidence will reliably catch themes present in 10% or more of participants.
It’s a lot like power analysis for qualitative research — except instead of asking “How big a sample to detect an effect?” you’re asking “How many voices do I need to hear before I’ve really covered the landscape?”
BEST also handles:
-
Singletons — deciding whether a theme that shows up once should “count” or not.
-
Depth — estimating just how rare the themes you can capture will be, given your sample size.
-
Missing Mass — estimating how much of the thematic world is still unseen, based on the frequency of singletons in your data.
All of this lives in a free, browser-only web app. No installs, no servers, no data collection. Just open the page, paste your coded interviews (or play with the demo set), and see where your study stands.
Why it matters
BEST doesn’t replace qualitative judgment — but it gives researchers a clear, defensible way to justify sample sizes. That means:
-
Transparency reviewers will appreciate.
-
Efficiency so you don’t oversample and exhaust participants.
-
Confidence that you’ve got the broad strokes covered.
And maybe best of all, it shifts “How many interviews are enough?” from an act of faith to a design choice.
Credits
-
Randal Cox — lead author of the manuscript and the one who derived the math and algorithm at the core of BEST.
-
Theandra Brin — co-author and developer of the interactive tool, turning formulas into something usable (and a little fun).
-
Brad Sagarin — co-author, who often admits that “qualitative studies scare me.” BEST gave him the comfort of finding a little R-shaped flashlight in a landscape that usually refuses his favorite statistical rigor.
Publication status
The manuscript describing BEST is already written. We’re gathering friendly reviewer notes now and expect to submit it for publication in the coming weeks.
👉 But in the meantime, you can already start using the BEST Interview Calculator.