Product discovery
The vision named the buyer. The research found another user.
FlowDesk had a detailed operations-manager persona, but fourteen churn interviews showed that dispatchers were central to daily use of the product.
A persona can be 95% complete and still have no research behind it.
Steps 01–02
The persona was complete on paper, but unsupported by research
FlowDesk sells shift-scheduling software to mid-sized logistics operators. It has 400 customers, mostly in the Benelux.
The product vision named one target group: operations managers. The discovery workspace therefore had one persona, with five sections filled in and a 95% completeness score.
No interviews were attached to it. The content came from people who had met operations managers during sales conversations, which gave the team useful information about buyers but little direct evidence about daily users.
Completeness measures how much has been written. It does not measure how well the content is supported.
Step 03
The interviews revealed a user the vision had missed
Personas inherit the limits of the research and target groups used to create them.
Fourteen churn interviews had been sitting in a folder since February. Eleven of them mentioned dispatchers, a role that did not appear anywhere in the existing discovery model.
Dispatchers did not sign the contract or attend sales demos. They opened the schedule every morning before the depot started work. FlowDesk had no explicit model of this user, and dispatchers could not access the product unless someone bought them a seat.
Requiring evidence on persona claims exposed the gap. The company had described the buyer in detail while leaving a frequent user almost entirely out of its product model.
Step 04
Draft jobs, pains and gains from the research
AI can produce a first draft across the interviews. A product manager still decides which claims are useful and sufficiently supported.
The useful part is traceability. Each job, pain and gain keeps a link to the interview evidence behind it, so the team can inspect or challenge the claim later.
One pain appeared in eleven of the fourteen interviews, expressed in different ways: the schedule lived in somebody else's account. Each comment looked small in isolation, but together they described a recurring access problem.
WHY IT MATTERS DOWNSTREAM
Specific wording makes later testing easier. A gain such as “stops being the last to find out” describes observable behaviour. A phrase such as “improved visibility” is harder to test because it leaves the expected change unclear.
Step 05
Prioritise assumptions by importance and uncertainty
Importance tells you how much the assumption matters to the decision. Uncertainty tells you how much evidence you still need.
The dispatcher persona produced nine assumptions. Scoring them by importance and uncertainty reduced that list to two assumptions that were important enough, and uncertain enough, to justify an experiment.
The team rated “setup effort is the main blocker” as highly important and low in uncertainty because everyone believed it. That confidence was not based on direct evidence, so the uncertainty rating itself needed to be challenged.
A useful discovery process should end with a smaller number of important questions to answer next.
Step 06
The new persona changed the business model
Customer segments can stay connected to personas so changes in research are reflected in the business model.
FlowDesk's lean canvas originally had one customer segment. Adding dispatchers produced two distinct groups: the operations manager who pays and the dispatcher who uses the product every day.
The value proposition also changed. The earlier wording focused on publishing schedules without spreadsheets. The revised version included access for the wider depot team.
That raised a separate commercial question. The company had assumed that growth would come partly from adding paid seats. If dispatchers needed free access, that revenue assumption also required testing.
The highest-risk assumption then moved into a two-week experiment, while the dispatcher persona remained connected to the opportunity it had revealed.
What discovery is for
Discovery keeps changes connected to the rest of the product model
Vision, personas and the business model are connected so that new research can change the parts of the product model that depend on it. The team can see which assumptions and decisions need to be revisited instead of updating each artefact separately.
Fair questions
What people ask about product discovery
- We already have a product vision. Do we start over?
No. Keep the vision if it still describes where you want to go. In the FlowDesk example, the vision stayed in place while the target groups and personas underneath it changed as new evidence appeared.
- How much of the persona does AI write?
AI can draft a persona from the transcripts you provide, with claims linked back to their source. You decide what to keep, edit or remove. Claims without supporting evidence remain visible as unevidenced.
- Do we need the business model canvas if we're not a startup?
Use it when research could change assumptions about customers, value or revenue. In the FlowDesk example, free dispatcher access conflicted with an existing assumption about paid seat expansion, which made the commercial implication visible early.
- What makes an assumption high-risk?
An assumption becomes a strong candidate for testing when it is important to the decision and still uncertain. Assumptions that matter less, or already have strong evidence, can wait.
- How long does this take?
It depends on how fast you can connect the interviews you already have. For a ten-person team that has typically taken about five days. In the demo, the reframe took two weeks and used research that had been sitting in a folder since February.
Start with the interviews you already have.
Connect the research to your personas and assumptions. The first useful result may be finding a user or problem that your current product model does not describe.
FlowDesk is the Shorter Loop demo workspace. You can open it to inspect the canvases, evidence links and persona relationships used in this example.