Discovery is the type that returns understanding rather than direction, and the report is shaped around that. This article covers the parts specific to discovery. For the sections every report shares, plus exports, judging how much weight a finding carries, and the thin-run banner, start with Reading a synthesis report.
The learnings block in a discovery report is titled Key Learnings, and it deliberately carries no per-learning action. It sits under the report’s headline, which is shown in a dark blue box, and each learning is numbered.
That is a design decision, not a gap. Attaching an action to each learning in a discovery study would manufacture direction you haven’t earned. The actions live one level down, in Findings and Next Steps. Each learning does carry links to the findings it was drawn from, so you can drop straight to the evidence.
This is the last block inside What Matters Most, and it is usually the highest-value one in a discovery report. It holds the expectations the interviews didn’t bear out, each with what it changes about what to do next, and a closing line labelled The quiet surprise on what your participants have quietly resigned themselves to. An expectation that traces back to an item in the assumptions or framing check links to that item. When nothing genuinely surprised, the block doesn’t appear.
Older reports build this block from the rest of the report instead: the checked items that came back Not supported or Contested, under Expectations that didn’t hold, and Problems participants have quietly accepted.
Read it before the findings. The findings tell you what you went looking for and found. This tells you what you would have missed, which is the entire reason to run discovery rather than jumping to validation.
Discovery puts two things in one list: the themes it clustered, and the problems that surfaced on their own. They are ordered by how broadly each one showed up across your participants.
Each finding carries:
The list is ordered by reach, but reach is not the same as importance. A problem five participants raised without being asked is stronger evidence than one eight participants agreed to after you brought it up, because the second result is partly a measure of how reasonable your question sounded.
When you’re choosing what to validate next, scan the provenance markers before you trust the ordering.
The evidence appendix carries one of two sections, depending on whether you filled in the optional assumptions field when you set the study up.
If you did, you get the assumptions check. It picks the three to five assumptions that matter most, drawn from the ones you wrote, your learning goals, and what running a discovery study implies, and grades each against the interviews. So an assumption you wrote can be left out, and one you never wrote can appear.
If you left it empty, you get the framing check instead. Nothing there is something you told us you believed. Each item is a reading the analysis took from your questions before it looked at the interviews, and the report will never say you expected it. This matters because a learning goal is a question, not a belief: asking what blocks adoption is not the same as believing you know the answer.
Either way each item gets a verdict of Supported, Partially supported, Contested, or Not supported.
It sits in the appendix on purpose. In discovery, the framing is not what the study is about. Reading it first would frame everything that follows as a scorecard on your predictions rather than a picture of the audience.
Because discovery is the one interview type that deliberately withholds direction. You went in without knowing which problems matter, so a report that handed you a build recommendation would be inventing confidence you haven't earned yet. What you get instead is an ordered picture of what people actually raised and how widely. The move after a discovery report is usually a problem validation study on the two or three findings that reached furthest, which is the type that does return verdicts.
Unprompted means a synthetic participant raised it themselves before you asked. Probed means it only came up because your question went looking for it. Mixed means both happened across different participants. This is the highest-value distinction in a discovery report: a pain raised unprompted is evidence the problem is live in someone's head, while the same pain agreed to after prompting may only mean your question was reasonable. When you're deciding what to validate next, weight the unprompted findings first even if a probed one reached more participants.
In the evidence appendix. If you filled in the optional assumptions field when you set the study up, the section is called the assumptions check: it grades the three to five assumptions that matter most, drawn from what you wrote, your learning goals, and what the study type implies. If you left that field empty, it is called the framing check instead, grading the reading the analysis took from your questions before it saw the interviews, and the report will not tell you that you expected anything. Either way each item gets a verdict of Supported, Partially supported, Contested, or Not supported. It sits in the appendix because in discovery your assumptions aren't the point of the study, the audience's own framing is. Read the findings first and this section second, so you see what people raised before you see how well the framing held.
Start with a free project, or walk through Candor with us first.