Eight stages, with room to go back.
Research
Understand the domain, existing approaches and whose perspective is missing.
Decision evidence: A sourced problem brief with assumptions and limits.
Problem discovery
Discuss recent real events, their cost and current workarounds. Separate an affected user from the person who can approve or pay for change.
Decision evidence: Evidence that the problem matters to identifiable people—not just praise for an idea.
Validation
Name the most consequential uncertainty. Define a small test, what would change the decision and a stop condition before running it.
Decision evidence: A dated record of observations, contrary evidence and the next decision.
Team formation
Bring together the expertise the problem requires. Agree on scope, time, responsibilities, attribution and any ownership or commercial terms.
Decision evidence: A realistic commitment and clear boundaries, not an assumed team or equity promise.
Prototype engineering
Build the smallest safe artifact that can test the uncertain part. A manual or non-software experiment may be enough.
Decision evidence: An inspectable artifact, constraints, test plan and known risks.
User testing
Observe relevant people using or assessing the prototype. Test usefulness alongside feasibility; technical correctness alone does not establish demand.
Decision evidence: Task outcomes, feedback, limitations and a decision to continue, revise or stop.
Product development
Consider support, reliability, accessibility, privacy, delivery cost and a sustainable way to reach users. A customer solution needs its own agreed scope.
Decision evidence: Evidence and available capacity for maintaining a useful solution.
Potential venture
Only consider a separate commercial venture when the opportunity, team, ownership and economics justify it. Research can also conclude, remain open or become a tool.
Decision evidence: A separate, explicit business decision—not a promised investment or automatic startup launch.
What a research record contains
The question and its author; the user and context; existing alternatives; assumptions; a method or experiment; dated source evidence; findings and counterevidence; limitations; and a next decision. Distinguish direct observations from interpretation. Obtain consent before recording or publishing someone’s information.
Download the blank research-record template. It is a writing aid, not a database or a claim of completed research.
Ideas that inform this approach
We draw on publicly described approaches to learning through experiments and customer discovery: Eric Ries’s Lean Startup principles, Rob Fitzpatrick’s customer-conversation guidance, Steve Blank’s Customer Development work, and Strategyzer’s Testing Business Ideas.
These are useful lenses, not proof of demand or guaranteed outcomes. Exploratory and public-interest research may need different success measures from a commercial product. The lab’s process should be revised as real work supplies evidence.
Next steps
Bring an organizational problem, discuss collaboration, or inspect the existing work.