Complete firstDecision-relevant hypothesis
One falsifiable assumption and the investment depending on it are named.
Without decision relevance, the team measures activity rather than risk reduction.
Method guideProduct Discovery
Experiment Canvas
A practical method guide: clarify prerequisites and materials, understand the complete process, and look up outputs, pitfalls, and stop criteria.
ComplexityLow
Time30-60 min
Participants1-5
FormatWorkshop + async
MaturityEstablished
01
Prerequisite
What needs to be finished first
Complete firstMeasurement and ethics access
Population, data capture, privacy, and stop authority are clear.
Without access or guardrails, the test is not responsible.
02
Preparation
What needs to be ready before start
Materials
Canvas, hypothesis source, baseline, audience access, event definitions, privacy review, and decision log.
People / roles
Experiment owner · Product · Research/Analytics · risk or compliance expert
Pre-read
Prepare the hypothesis, investment decision, available population, baseline, and relevant guardrails.
Time needed
30–45 minutes to design; execution is separate
Setup
Record hypothesis, test, primary metric, and threshold in the Experiment Canvas; verify data capture and stop boundary before launch. The Test Card offers the same four core fields for simple tests. Maintain one authoritative plan.
03
Core question
The one question this method answers
Which critical assumption will we test through what observable behaviour, and which pre-set threshold decides?
04
Flow
Phases with goal, actions, output and pitfall
An Experiment Canvas makes the team state what must be true, how it will test the claim, what it will measure, and which threshold triggers a decision. Activity is not evidence; the measure must fit the hypothesis.
Experiment Canvas
Fictional worked example
Eight fields connect assumption, test, measurement, safeguards, and decision.
Hypothesis
What must be true?
At least 30% of exposed active project leads start a CSV export.
Decision
Which investment depends on this?
Build an interactive prototype above threshold; otherwise revise the value proposition.
Test design
Who sees what, for how long?
Fake door for 200 eligible active accounts over 7 days.
Primary metric
Numerator, denominator, segment, window
Unique export starts / exposed active accounts in 7 days.
Threshold
Which boundary applies before launch?
Success ≥30% with at least 150 exposed active accounts.
Guardrail
When to pause or stop?
Stop above 3% support complaints about the unavailable feature.
Evidence
Which raw data and quality checks?
Exposure and click events, account segment, bot filter, deviation log.
Learning and next step
What did the rule yield and what follows?
32% at n=181; guardrail 1.1%. Build prototype and test usability.
- 1
Hypothesis
Write a discrete assumption
Goal
Define one decision-relevant, falsifiable assumption.
Inputs for this step
- Product or business model
- Known uncertainties
- Upcoming investment decision
Actions
- State what must be true for the idea to work.
- Specify audience, behaviour or outcome, and context.
- Choose one assumption per experiment.
Output
Testable hypothesis
Audience · behaviour · context · expected outcome
Quality check
Can an outcome clearly weaken the claim?
Pitfall
Several assumptions are bundled into one sentence, making any result ambiguous. Split the claim and select the most consequential assumption.
Timing: 5–10 minutes
- 2
Design
Design the smallest useful test
Goal
Choose the smallest ethical test that yields enough evidence for the decision.
Inputs for this step
- Testable hypothesis
- Available audience and channels
- Time, cost, and compliance constraints
Actions
- Observe behaviour rather than opinion where the claim requires it.
- Define population, exposure, procedure, and data capture.
- Check deception, privacy, and operational side effects.
Output
Test design
Population · intervention · duration · data capture · ethics
Quality check
Does the test directly address the hypothesis?
Pitfall
The team selects a fast opinion test for a hypothesis about actual behaviour. Match evidence strength to the decision and collect behavioural data where needed.
Timing: 10–15 minutes
- 3
Measure
Set metric and threshold before testing
Goal
Define a computable primary metric, data basis, and decision threshold.
Inputs for this step
- Test design
- Baseline or reference value
- Decision need
Actions
- Define numerator, denominator, segment, and measurement window.
- Set threshold and minimum data before launch.
- Add guardrails and stop limits.
Output
Measurement rule
Metric · formula · threshold · minimum data · guardrail
Quality check
Can raw data determine success or failure unambiguously?
Pitfall
The threshold is chosen after seeing results, or activity is confused with effect. Freeze and version threshold, denominator, and analysis rule before launch.
Timing: 10–15 minutes
- 4
Learn
Run, analyse, and decide
Goal
Evaluate observed data against the pre-set rule and take the next decision.
Inputs for this step
- Approved design
- Measurement data
- Pre-set threshold
Actions
- Run the test to protocol and log deviations.
- Check data quality and guardrails before the primary metric.
- Decide persevere, adapt, retest, or stop with rationale.
Output
Learning card
Data · deviations · result · decision · new assumption
Quality check
Does the decision follow the pre-set rule and evidence strength?
Pitfall
A surprising result is made to fit by changing segments or thresholds after the fact. Label exploratory analysis separately and create a new test for new hypotheses.
Timing: Execution separate; 15–30 minutes analysis
05
Artifact
What comes out at the end
Sketch: what the artifact looks like
Form
Versioned Experiment Canvas as the authoritative test plan; record evaluation and next action with a raw-data link in the canvas or in a linked Learning Card.
Versioning / ownership
Retain date, author, source state, and decision status for each revision; do not overwrite earlier results.
Tool alternatives
- Experiment Canvas or Test Card
- Analytics or observation protocol
- Learning Card and decision log
experiment-canvas-working-template.md
Plan Experiment Canvas with real scope, suitable roles, and a clear decision objective.
Experiment Canvas: Worksheet
Question
To be confirmed
Desired outcome
To be confirmed
Scope
Full learning cycle Time: 30–45 minutes to design; execution is separate
Complete this worksheet on paper or in your own document during the work.
Preparation for this scope
Method setup
- Materials: Canvas, hypothesis source, baseline, audience access, event definitions, privacy review, and decision log.
- Roles: Experiment owner · Product · Research/Analytics · risk or compliance expert
- Advance information: Prepare the hypothesis, investment decision, available population, baseline, and relevant guardrails.
- Overall time needed: 30–45 minutes to design; execution is separate
- Setup: Record hypothesis, test, primary metric, and threshold in the Experiment Canvas; verify data capture and stop boundary before launch. The Test Card offers the same four core fields for simple tests. Maintain one authoritative plan.
Prepare for the work steps
Hypothesis: Write a discrete assumption
- Product or business model
- Known uncertainties
- Upcoming investment decision
Design: Design the smallest useful test
- Testable hypothesis
- Available audience and channels
- Time, cost, and compliance constraints
Measure: Set metric and threshold before testing
- Test design
- Baseline or reference value
- Decision need
Learn: Run, analyse, and decide
- Approved design
- Measurement data
- Pre-set threshold
Experiment Canvas
Hypothesis
What must be true?
...
Decision
Which investment depends on this?
...
Test design
Who sees what, for how long?
...
Primary metric
Numerator, denominator, segment, window
...
Threshold
Which boundary applies before launch?
...
Guardrail
When to pause or stop?
...
Evidence
Which raw data and quality checks?
...
Learning and next step
What did the rule yield and what follows?
...
Strategyzer: Validate Your Ideas with the Test Card
Hypothesis: Write a discrete assumption
Expected artifact: Testable hypothesis
Audience · behaviour · context · expected outcome
Entry:
...
- Can an outcome clearly weaken the claim?
Design: Design the smallest useful test
Expected artifact: Test design
Population · intervention · duration · data capture · ethics
Entry:
...
- Does the test directly address the hypothesis?
Measure: Set metric and threshold before testing
Expected artifact: Measurement rule
Metric · formula · threshold · minimum data · guardrail
Entry:
...
- Can raw data determine success or failure unambiguously?
Learn: Run, analyse, and decide
Expected artifact: Learning card
Data · deviations · result · decision · new assumption
Entry:
...
- Does the decision follow the pre-set rule and evidence strength?
Open questions and next steps
...
Method guide: https://methodatlas.meierhoff-systems.de/en/methods/experiment-canvas/run-sheet
06
Example output
Concrete filled scenario, fictional example
Fictional example for orientation. Counts and timings describe this case and are not universal requirements.
experiment-canvas-beispiel.md
Concrete filled scenario, fictional example
Fictional example: we believe at least 30% of weekly active project leads will start a CSV export when offered in the results view. Test: clickable fake door for 200 eligible accounts over seven days. Primary metric: unique export starts / exposed active accounts. Success ≥30%; stop if support complaints exceed 3%. Then build a prototype; otherwise revise the value proposition.
07
Pitfalls
Recognize symptoms and steer against them
Quick overview, the full text is in the matching phase in the flow
- Write a discrete assumptionSeveral assumptions are bundled into one sentence, making any result ambiguous.View in phase
- Design the smallest useful testThe team selects a fast opinion test for a hypothesis about actual behaviour.View in phase
- Set metric and threshold before testingThe threshold is chosen after seeing results, or activity is confused with effect.View in phase
- Run, analyse, and decideA surprising result is made to fit by changing segments or thresholds after the fact.View in phase
Trap
Write a discrete assumption
Symptom
Several assumptions are bundled into one sentence, making any result ambiguous.
What to do
Split the claim and select the most consequential assumption.
Trap
Design the smallest useful test
Symptom
The team selects a fast opinion test for a hypothesis about actual behaviour.
What to do
Match evidence strength to the decision and collect behavioural data where needed.
Trap
Set metric and threshold before testing
Symptom
The threshold is chosen after seeing results, or activity is confused with effect.
What to do
Freeze and version threshold, denominator, and analysis rule before launch.
Trap
Run, analyse, and decide
Symptom
A surprising result is made to fit by changing segments or thresholds after the fact.
What to do
Label exploratory analysis separately and create a new test for new hypotheses.
08
Stop criteria
When to pause, resolve prerequisites, or choose another approach
Every entry follows the method and links to evidence or explicitly marked uncertainty.
The next decision, owner, and review date are visible where required.
Where to go next?
Product Discovery
Opportunity Solution Tree
Updates the solution tree with the test's learning and decision outcome.
Product Discovery
Assumption Mapping
Reorders the next risky assumption based on the test result.
Product Discovery
Learning Card
Condenses the completed test evaluation into observation, insight, and next action while referencing the existing plan.
Finished the method guide?
Go to the profile for purpose, similar methods, and sources or continue to the next method in the catalog.