Insights
A rigorous statistical model behind your Discovery study
Your Discovery study now chooses its customer segments with a statistical model built for the method, so you can see which patterns are real, how strong they are, and how far to trust them before you act.
How it works
A Discovery study segments your customers by pairing a situation that pushes them with an outcome they are pulling toward, an approach we adapted from the Jobs-to-be-Done method. The whole job rests on one judgement: which pairing reflects a real pattern in your interviews, rather than one that only looks tidy. We put an explicit statistical model behind that judgement, and it now shows its evidence on the page.
It shows how confident it is
Every map carries a confidence you can read at a glance: how often it held up when the analysis was re-run on resampled interviews (489 of 500 in the example above), and the odds the pattern is a coincidence, stated in plain words and backed by descriptive statistics, such as a p-value, for anyone who wants to check. When the evidence supports only a simpler split, the report says so and tells you roughly how many more interviews would sharpen it, so you always know how much weight the finding can bear.
It measures real coupling
The model reads every interview where a customer linked a situation to an outcome in their own words, and tests how strongly each pairing actually holds together across your whole study, correcting for the fact that it weighs many pairings at once. The map you see is the pairing the evidence backs most, so a strong link between one situation and one outcome comes through as the headline instead of being averaged away.
On the page
The report opens with the map and a short, ranked list of the sharpest findings, each linking to the segment it concerns. The methodology sits at the end for anyone who wants to audit it, and every segment is named by the real force behind it, in language a reader can act on without a glossary.
What you get
A segmentation you can question is a segmentation you can trust. The report shows its working: what it measured, how strong the pattern is, and how confident you should be before you build against it. You can hand it to your team and defend every segment on the evidence, and you spend your next round of research closing the gaps the model flags rather than guessing where they are.
Already generated a report?
Rebuild it on the new model whenever you like: open the study and select Regenerate insights. Until you do, your existing reports stay exactly as they are, so you decide when anything changes.