How many customer interviews until patterns repeat?

What do researchers mean by saturation?

Saturation is the point where new interviews stop adding anything new. Hennink, Kaiser and Marconi split it in two: code saturation, when you have heard the range of themes, and meaning saturation, when you understand those themes well.

Their study, published in Qualitative Health Research in 2017, went through 25 in-depth interviews. The range of themes was there by interview nine. A richly textured understanding of them took 16 to 24 interviews.

It was a health study, not a sales one, so we read the numbers as a shape rather than a quota. By call nine you can probably list what buyers worry about, but not which worry decides a deal or how they phrase it, and those are the parts we build positioning from customers' own words on.

How many interviews does one buyer segment need?

For one fairly similar group, plan on roughly 12 to 20, knowing that published ranges run from 12 up to 30. The number moves with how alike the people are.

Koji, which sells an interview tool, cites Guest and colleagues from 2006: for a fairly homogeneous population, a sample of 12 is generally sufficient for high-level thematic analysis. eChai Ventures' page on interview counts puts the usual practice at 15 to 30 conversations with the same customer type.

Clozd, a win-loss research vendor, puts a specific B2B segment at 20 to 30 interviews. Its B2B interview sample size guide also says participation runs 15% to 30%, so a goal of 30 completed interviews can mean reaching out to 100 to 200 accounts.

Both companies sell interview work or tools, and a business that sells interviews has a reason to favor larger samples. A small team with one clear kind of buyer can start planning nearer 12 and let the calls say when to stop.

Why does each segment count separately?

Because saturation happens inside a group of similar people, and a mixed group takes far longer to get there. Twelve calls split across three kinds of buyer are closer to four calls each.

The eChai page, summarizing an article by Justin Wilcox of Customer Dev Labs, says different segments saturate at different speeds. A finance lead and an operations lead can raise different worries for weeks, and pooling their calls makes it look as if the patterns never settle.

So we count per segment, and we write down before the first call who belongs to which one. If that line is not drawn yet, the ICP Kit walks through it in one sitting.

How do you know when to stop?

Koji's guide says that when two or three interviews in a row produce no fundamentally new themes, you are likely at code saturation. The eChai summary of Wilcox adds a blunter test: you are done when you would bet a hundred dollars on what the next person says, and win.

The same summary says patterns should show up by interview five and the major questions should be answered by interview twenty. We treat five as the point to start looking, and we keep a running list of themes with the call where each first came up, because the list makes the quiet run visible.

Corvan AI counts how often each question comes back across the calls you upload, which keeps that list without a spreadsheet. Hitting the stopping rule still only gives you the range of themes, and in Hennink's study the depth took 16 to 24 calls, not nine.

What none of this settles is what counts as new. A buyer who says setup is too hard and another who says it needs an engineer for a week may be raising one theme or two, and we have not found a source that says how to tell.

How many wins and losses do you need to trust a pattern?

Koji's win-loss guide says to hold structural changes until you have 10 to 15 interviews. It treats a factor that appears in 40% or more of losses but under 10% of wins as systemic.

The same win-loss analysis guide suggests starting with 5 to 8 interviews a month, roughly two a week. At five a month, a small B2B team reaches 15 in three months and can review the themes after each batch.

The 40% test works because it compares wins against losses. A complaint that turns up in half your losses and half your wins is something buyers say either way. One that sits in the losses and almost never in the wins is worth acting on, and our post on finding recurring objections in sales calls covers how to spot it.