Discovery advice is usually written for teams with a dedicated researcher and a quarter to spend. Most companies bringing in part-time product leadership have neither. The question is not how to run ideal discovery; it is what to keep when you have a few hours a week and a decision that cannot wait.
The good news is that the ratio of value to effort across discovery activities is extremely uneven. A small subset does most of the work.
Keep: direct customer conversations
Five or six real conversations with people who have the problem will move your understanding more than anything else available to you. Not a survey, not a summary from sales, not a synthesis someone else did. Direct contact, with follow-up questions you did not plan.
They do not need to be long. Thirty minutes, asking what they were trying to do, what they did instead, and what it cost them. Ask about the last time it happened rather than what they generally do — recollection of specific events is far more reliable than self-reported patterns.
Keep: what customers already do in the product
Behavior beats stated preference, and you already have the data. Where do people stop. What do they do repeatedly that the product does not support well. Which accounts expanded and what did they use first. An hour in the analytics tool with a specific question is worth more than a week of it as a browsing habit.
Keep: a written problem statement
One paragraph: who has the problem, what it costs them, what they do today, and how we would know if we had solved it. It takes twenty minutes and it is the single artifact that most reliably prevents a team from building a well-executed answer to the wrong question.
Cut: the competitive matrix
The fifteen-column comparison grid is almost always produced to feel thorough and almost never changes a decision. Know what the two or three real alternatives do and where they are weak. That fits on half a page.
Cut: persona documents nobody opens
If the team can already describe the customer accurately in conversation, the document adds nothing. If they cannot, the document will not fix it — customer contact will.
Cut: research for decisions already made
This is the most common waste and the hardest to see, because it looks exactly like real discovery. If the founder has decided, the contract is signed, or the direction is not actually open, the research is theater. Either reopen the decision honestly or skip the study.
Cut: large-sample surveys when you need direction
Surveys are good at measuring how widespread something is once you know what to ask. They are bad at telling you what to ask. Early on you need direction, and direction comes from conversations.
The sizing rule
Match research effort to how reversible the decision is.
- Cheap and reversible — a copy change, a default, a flow reorder. Ship it and watch. Research is more expensive than the mistake.
- Expensive but reversible — a new surface, a significant feature. A handful of conversations plus existing behavioral data.
- Expensive and hard to reverse — pricing structure, data model, a platform commitment, anything that touches customer contracts or migration. Slow down, and spend real time here. These are the decisions where a week of research pays for itself many times over.
A sample of one is an anecdote. A sample of six saying the same unprompted thing is direction — not proof, but enough to act on when the action is reversible. Most teams over-research the reversible decisions and under-research the permanent ones.
The trap
The failure mode of constrained discovery is validation rather than learning: asking questions shaped so the answer confirms the plan, then reporting back that customers loved it. The tell is that the research never changes anything. If six months of discovery has not killed a single idea, it is not discovery.
Related reading: what fits into ten hours a week, and why the roadmap is usually not the problem.