Document Type: Protocol
Status: Active
Version: v1.1
Authority: Experimentation Brain
Applies To: Experimentation Brain test result handling, decision-making, and post-test workflow
Parent: Experimentation Brain Canon
Last Reviewed: 2026-04-25
Purpose
This protocol defines how MWMS must handle test results after a test has been executed.
Its purpose is to ensure that every test:
• produces a structured result
• leads to a clear next action
• generates reusable learning
• prevents uncontrolled scaling
• improves future decision quality
• is evaluated against expected performance (NEW)
This protocol transforms testing from isolated activity into a continuous learning and improvement system.
Scope
This protocol applies to:
• all tests executed within Experimentation Brain
• post-test evaluation
• result classification
• decision-making after test completion
• feedback into Affiliate Brain and Research Brain
It governs what must happen after a test produces results.
It does not govern:
• test setup
• test execution
• statistical calculation methods
• budget allocation
Core Principle
A test is not complete when it stops running.
A test is only complete when:
• its result is classified
• a decision is made
• learning is captured
• next action is defined
• actual performance is compared to expected performance (NEW)
🔴 Decision Discipline Extension (NEW)
All test results must be evaluated using:
👉 Forecast → Actual → Variance → Decision
Without this:
→ optimisation becomes guess-based
Workflow Overview
Test Execution
↓
Test Completion
↓
Result Classification
↓
Decision
↓
Action
↓
Learning Capture
↓
Feedback Loop
Step 1 — Test Completion
A test enters completion when:
• planned duration is reached
• success criteria are met
• failure criteria are met
• test is manually stopped
At this point, the test must move into structured evaluation.
Step 2 — Result Classification
Every test must be classified into one of the following categories:
Clear Winner
Definition:
• meets or exceeds success criteria
• strong signal quality
• consistent performance
• aligns with or exceeds forecast expectations (NEW)
Weak Positive
Definition:
• shows some positive signal
• inconsistent or marginal performance
• requires refinement or retesting
• variance from forecast is unstable (NEW)
Inconclusive
Definition:
• insufficient data
• unclear signal
• test conditions not adequate
• unable to validate against forecast (NEW)
Structured Failure
Definition:
• fails to meet success criteria
• provides clear negative signal
• result is interpretable
• underperforms forecast expectations (NEW)
Invalid Test
Definition:
• test setup was flawed
• execution issues occurred
• data is unreliable
• data fails Data Brain validation (NEW)
🔴 Data Validation Gate (NEW)
Before classification:
Data must pass:
• Signal Integrity
• Measurement Integrity
• Data Trust
If not:
→ test must be classified as Invalid Test
Step 3 — Decision Mapping
Each classification must map to a defined action.
Clear Winner → Scale Or Expand
Actions:
• escalate to Finance Brain for capital readiness
• compare performance against forecast expectations (NEW)
• consider scaling strategy
• expand testing dimensions
Weak Positive → Refine And Retest
Actions:
• adjust angle
• refine creative
• modify targeting
• run follow-up test
• adjust hypothesis based on variance (NEW)
Inconclusive → Extend Or Redesign
Actions:
• extend duration
• increase sample size
• redesign test
• review measurement conditions (NEW)
Structured Failure → Kill And Record
Actions:
• stop further testing
• record failure
• capture learning
• validate that failure is signal-valid (NEW)
Invalid Test → Reset And Rebuild
Actions:
• fix test design
• fix measurement issues (NEW)
• relaunch with corrected structure
Step 4 — Learning Capture
Every test must produce structured learning.
Required Learning Fields
• What was tested
• What worked
• What did not work
• Why the result occurred (best interpretation)
• Market insight
• Angle insight
• Platform insight
• Funnel insight
• Recommendation for future
• Variance vs expected outcome (NEW)
Step 5 — Feedback Loop
Learning must be routed back to:
Affiliate Brain
Improves:
• opportunity selection
• evaluation quality
• decision accuracy
Research Brain
Improves:
• signal classification
• pattern detection
• market understanding
Step 6 — Scaling Control
Scaling must not occur automatically.
Scaling Conditions
Before scaling:
• result must be classified as Clear Winner
• confidence must be sufficient
• risk must be understood
• Finance Brain must approve capital
• performance must align with forecast expectations (NEW)
Scaling Block Conditions
Scaling must be blocked if:
• result is weak or inconsistent
• signal is unclear
• risk is high
• insufficient data
• data integrity is compromised (NEW)
• variance from forecast is unstable (NEW)
Step 7 — Record Final Outcome
Each test must end with:
• final classification
• final decision
• final action
• learning captured
• comparison to forecast (NEW)
Controlled Loss Alignment
This workflow enforces the MWMS Controlled Loss Principle by:
• ensuring failures are interpretable
• preventing uncontrolled scaling
• converting losses into learning
• reducing repeated mistakes
• maintaining capital discipline
• preventing decisions based on weak or invalid data (NEW)
Governance Role
This protocol ensures:
• testing produces structured outcomes
• decisions are consistent
• learning is captured
• system improves over time
• decisions are based on validated data and expected performance (NEW)
Relationship To Other MWMS Pages
This protocol operates alongside:
• Experimentation Brain Test Candidate Screen Specification
• Affiliate Brain To Experimentation Brain Handoff Specification
• Experimentation Brain Structured Testing Protocol
• Finance Brain Capital Allocation Ladder
• MWMS Controlled Loss Principle
• Data Brain Measurement Planning Framework (NEW)
• Data Brain Data Trust Framework (NEW)
Drift Protection
The system must prevent:
• tests ending without classification
• results without decisions
• scaling without validation
• ignoring failed tests
• repeating mistakes without learning
• treating all results as equal
• decisions without forecast comparison (NEW)
• decisions based on unvalidated data (NEW)
Architectural Intent
This protocol completes the MWMS testing loop.
It ensures that:
• every test produces value
• every result leads to action
• every action improves the system
• decisions are grounded in both data and expectation (NEW)
It transforms MWMS into a learning system that compounds over time.
Change Log
Version: v1.1
Date: 2026-04-25
Author: Experimentation Brain / HeadOffice
Change
Upgraded workflow to include:
• forecast vs actual comparison layer
• Data Brain validation gate
• variance-aware decision logic
• stronger scaling discipline
• alignment with Measurement Planning system
Change Impact Declaration
Pages Created:
None
Pages Updated:
Experimentation Brain Test Result And Decision Workflow
Pages Deprecated:
None
Registries Requiring Update:
No
Canon Version Update Required:
No
Change Log Entry Required:
Yes