Minimally Viable Experiment
Test the decisive assumption with the fastest useful feedback.
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 6
- Confidence
- 99%
A minimally viable experiment is the smallest controlled action capable of testing the assumption behind a decision. Instead of asking someone to commit to a permanent change, frame the action as a temporary experiment with a clear hypothesis, bounded exposure, and rapid feedback. Specify what you believe, the evidence that would confirm or weaken that belief, and the fastest realistic way to obtain it. This reduces fear, accelerates action, and tightens the causal link between behavior and result. It goes one step earlier than a minimum viable product because the assumption may be testable before anything is built. Learning occurs through the designed experiment and its feedback, not merely through accumulating unexamined experience.
Origin
Barnard learned the power of experimental framing while helping recover a $5 billion SAP implementation. An executive said he acted because Barnard proposed a one-month experiment rather than a permanent change. Extracted from The Jasmine Star Show.
Core principles
- 01Calling a change an experiment lowers the emotional barrier to action.
- 02Experience becomes learning only when assumptions are deliberately tested.
- 03Fast feedback strengthens the connection between action and result.
- 04The smallest valid test should precede a full product or rollout.
- 05Every experiment needs a clear assumption and observable result.
How to run it
- 1
Identify the blocked action
Select a decision that has been made or considered but is not being executed because the outcome is uncertain. Define the feared commitment.
Pro tip Choose an action where one unknown is driving most of the hesitation.
Watch out Do not experiment casually with safety, legal, or irreversible risks.
- 2
State the assumption
Write the belief the action depends on in testable language. Specify what you expect to happen and for whom.
Pro tip Use an if-then statement with an observable result.
Watch out A vague assumption produces ambiguous learning.
- 3
Design the minimum test
Find the smallest action capable of producing credible evidence. Test demand, behavior, or feasibility before building the complete solution.
Pro tip Ask whether a mock offer, manual service, prototype, or short trial would suffice.
Watch out A test that is too artificial may not predict real behavior.
- 4
Define feedback
Choose the metric, threshold, and observation period before running the experiment. Decide what success, failure, and ambiguity mean.
Pro tip Prefer behavioral evidence over stated intentions.
Watch out Changing the success criterion after seeing results destroys the test's value.
- 5
Bound the exposure
Set explicit limits on time, cost, users, and operational risk. Make it easy to stop or reverse the experiment.
Pro tip Use a fixed end date and budget ceiling.
Watch out A supposedly temporary test can become permanent through inertia.
- 6
Run and review
Execute promptly, collect the defined feedback, and compare it with the assumption. Decide whether to scale, revise, or abandon the idea.
Pro tip Schedule the review before launching the experiment.
Watch out Do not interpret weak or mixed evidence as automatic validation.
In the wild
A major SAP implementation was six months behind schedule. Rather than asking executives to adopt a permanent operating change, Barnard proposed trying an experiment for one month. The temporary framing gave the executive enough confidence to act and generate feedback.
→ The organization implemented the proposed approach, caught up, and delivered the project on time.
An entrepreneur assumes customers will pay for a new service. Before building software, the entrepreneur presents a concrete offer to qualified prospects and tracks deposits or signed commitments under a fixed test period.
→ Real buying behavior validates or rejects the pricing assumption before substantial development cost is incurred.
Common mistakes
Building before testing the assumption
A complete product can consume months even when the central demand or pricing assumption could have been tested in days.
Running an experiment without a hypothesis
Activity alone does not generate useful learning when no assumption or success threshold was defined.
Seeking slow feedback
Long delays weaken the connection between action and result and postpone the next decision.
Is it for you?
Best for
It is best for product ideas, offers, workflow changes, and strategic decisions that can be tested safely.
Not ideal for
It is not ideal for irreversible actions or tests whose risks cannot be ethically and safely bounded.
From the transcript
“we don't learn from experience. We learn from experiments.”
“think about every decision you make as an experiment”
“what's the fastest feedback I can get.”
From the episode
Make Better Decisions Faster