Listen
The defining limit of qualitative research has always been scale.
A typical qualitative paper stands on 25 to 100 interviews, because that is what a coding team and a timeline can hold. So the richest kind of evidence there is, what the people inside the system actually know, gets sampled instead of measured, and the sample is defended in the limitations section every time, because it has to be.
Why
The thing I got wrong
I spent my career building instruments that measure an organization from the outside. First getting the data at all. Then benchmarking, which is putting one organization’s number next to another’s and asking what the difference is made of. Then matching practice against performance, to find out which practices drive the result and which are merely present in the binder. Each one answered something the one before it could not. And all three are blind in exactly the same place.
None of them can tell you what the person operating the system knows about it. The practices that separate a good operation from a bad one are mostly not in any document, because they were invented on a shift and never written down, and you cannot inventory what was never recorded. The only way to get it is to ask the people doing the work.
I have been asking them for most of my career, and for most of that time I could reach about ten at a time. A cross functional group, or a shift with pizza and nothing in my hands to write with. The research literature does not do much better: 25 to 100 interviews is a publishable study. The method works better than anything else available, and it does not scale. That limit shaped my work so completely that I stopped noticing it was a limit. About three years ago it stopped being one. This instrument takes 1,000 or 10,000 interviews, every one in the person’s own words, and codes all of them.
I built my career on the idea that you measure the system, and I was right, and it was not enough. What the people inside it know is the last unmeasured thing. The 8M Institute exists because it is finally measurable, and because the right number of people to ask was never ten. It was everyone.
What it does
Three years ago the limit came off. This is what 1,000 or 10,000 interviews look like.
It asks everyone in the population, in their own words. Not 25 to 100 interviews. 1,000 or 10,000: every person in scope gets a prompt and a box, and types or dictates the way they would a text message.
It codes occasions, never opinions. The prompts ask what happened, what you did, who decided, and how often. Nobody is asked to rate anything, so what comes back is evidence rather than sentiment.
Eight a priori categories sort the causes. Man, Machine, Material, Method, Measurement, Mother Nature, Maintenance, Management. Every occasion an organization produces lands in one of the eight.
The finding is the weakest category, never an average. No composite is computed anywhere, which is itself a methods claim: nothing in the instrument can blur the weakest category into an acceptable mean.
It parallels GenAI augmented content analysis. The same qualitative coding discipline used in the doctoral work behind OrbisDiscovery, run at census scale instead of sample scale.
Proof
Firms with a wide gap between how management and staff see the business show 16% lower return on assets, 10% lower market value against asset value, and a 48% higher chance of CEO turnover. The gap this instrument measures is the gap those findings price.
Stop the Line. The book is free. It is how most people meet this work. Get Stop the Line at 8minstitute.org (opens in a new tab)
Eight a priori categories, and the finding is the weakest one
Man
Qualification and competence
Machine
Equipment and the systems driving it
Material
Inputs and inbound information
Method
Procedure and standard work
Measurement
Data, definition, and abstraction
Mother Nature
The physical environment of the work
Maintenance
Asset upkeep and recall
Management
Resourcing, targets, and governance
Eight boxes, eight owners, no exits. Every cause an organization produces lands in one of them, and the finding is the weakest of the eight, never the average. No composite is computed anywhere.
The method was never the limit. The interview count was.
Thirty minutes. Bring the population you have been sampling, and the question you would ask if you could ask all of them.