An MVP is the smallest safe, credible product that can test a valuable assumption with real users and measurable evidence. It is not the first third of a feature list or a low-quality version of the final vision.
Name the Decision
Write the decision the MVP must inform: should we invest, change the proposition, replace the workflow, or stop? Then state the riskiest assumptions about desirability, usability, feasibility, viability, data, security, and operations.
If success would not change a decision, the experiment has no useful purpose.
Define One Outcome
Choose one user, context, problem, and end-to-end result. A useful statement is: “For wholesale account managers, reduce the time to produce an approved quote from two days to two hours without increasing pricing exceptions.”
Baseline the current journey. Define adoption, completion, time, quality, error, conversion, retention, or cost evidence and the minimum signal that justifies the next investment.
Map the Thin Journey
Include every step necessary to deliver the outcome: entry, identity, core action, confirmation, exception, support, and measurement. Reduce variants and automation before removing essential integrity.
Manual work behind the interface can be acceptable if it is safe, disclosed to operators, capacity-bounded, and does not invalidate the hypothesis. Hard-coded fake capability cannot prove feasibility.
Keep Non-Functional Minimums
“Minimum” does not remove security, privacy, accessibility, reliability, data integrity, observability, or legal obligations. Scale the controls to risk, but define them explicitly.
Protect production data and users. Establish support ownership, incident response, backup where needed, and a way to exit or delete pilot data.
Prioritise with Evidence
Classify scope as essential to the hypothesis, required for safety/compliance, useful later, or excluded. For disputed items, ask which assumption they test and what happens if omitted.
Use prototypes for usability, spikes for technical uncertainty, concierge tests for service demand, and production software only where necessary. Different questions need different artefacts.
Timebox Learning, Not Quality
Set a discovery and delivery window, named team, fixed outcome, and review points. When uncertainty emerges, reduce audience, use case, integration breadth, or automation—not engineering hygiene.
Maintain a decision log and prevent “while we are here” additions.
Plan the Experiment
Define recruitment, consent, rollout, analytics, qualitative research, operational monitoring, success threshold, stop condition, and owner. Segment results; an average can hide that the product works only for the easiest customers.
At the end, choose scale, iterate, pivot, or stop. Archive evidence and technical debt so a successful pilot does not silently become permanent production.
DualByte's software development service can turn a product hypothesis into a focused, measurable MVP.
Sources
Need help with implementation?
Get a free consultation with the DualByte team for your business technology needs.
