In short
A focused discovery for a known problem may take 1–2 weeks; a new product with several user groups, complex data, integrations, or operational risk often needs 3–6 weeks. Timebox the work around a specific investment decision and stop when critical assumptions, feasibility, scope, risks, and the next test are clear enough.
Planning ranges
| Context | Typical range |
|---|---|
| Known workflow improvement | 1–2 weeks |
| Focused new feature or service | 2–4 weeks |
| New product or unfamiliar market | 3–6 weeks |
| Complex regulated or multi-party system | 6+ weeks, usually in bounded stages |
Discovery length should follow uncertainty and consequence, not a standard workshop package.
What extends discovery
- several user groups with conflicting needs;
- difficult access to representative users;
- unreliable analytics or missing operational data;
- uncertain business model;
- complex permissions and workflows;
- legacy systems and integrations;
- security, privacy, or regulatory review;
- technical feasibility risk;
- distributed decision-makers;
- a broad problem with no priority.
A three-week example
Week 1: frame and collect
Align on the decision, audience, evidence, constraints, current workflow, and assumptions. Begin interviews, analytics review, and technical investigation.
Week 2: test the riskiest points
Map the opportunity, prototype uncertain interaction, investigate data and integrations, and observe representative users or operators.
Week 3: synthesise and decide
State evidence and uncertainty, narrow scope, record risks, recommend the next test or MVP, and define success and stop conditions.
Activities can overlap. The sequence matters more than ceremony: frame, investigate, test, decide.
Exit criteria
Discovery can end when:
- the audience and problem are bounded;
- current alternatives and consequences are understood;
- critical assumptions are visible;
- the highest risks have evidence or a test plan;
- data, integration, and operational constraints are credible;
- the next release has scope and non-objectives;
- decision-makers know what remains uncertain;
- the next investment has an owner and evaluation criteria.
It does not need certainty about every future feature.
Signs discovery is too short
The feature list existed before evidence, no representative user or operator was observed, integrations remain names in a slide, or scope depends on contradictory stakeholder assumptions.
Signs discovery is too long
Research repeats known themes, new artefacts do not change decisions, the team avoids testing, or “more confidence” has no defined threshold.
Common questions
Can discovery take only a few days?
Yes, for a narrow decision with accessible evidence and low consequence. Be explicit about assumptions that remain untested.
Is discovery included in development time?
Sometimes. Ask whether it produces a separate decision and scope, and whether later estimates change based on its findings.
Should engineers participate?
Yes when data, integrations, feasibility, security, performance, or operations influence the product decision.
Can discovery happen while building?
Continuous discovery can run alongside delivery, but major commitments benefit from focused evidence before expensive implementation.
What if discovery recommends stopping?
That is a valuable outcome when the evidence prevents a larger unsupported investment.
