Document Type: Reference
Status: Working Document
Version: v1.2
Authority: Experimentation Brain (Operational Layer)
Applies To: Pre-promotion testing insights, temporary experimentation learnings, and protocol-candidate signals
Parent: Experimentation Brain Architecture
Promotion Rule: Insight must remain valid for 7 days before Canon mutation
Last Reviewed: 2026-04-25
Purpose
This log exists to capture structured testing insights before promotion.
It prevents loss of signal without mutating Canon.
This document:
• does not change authority
• does not authorise capital
• does not override Canon
• does not enforce governance
It is a temporary compression buffer.
Insights recorded here may later be:
• promoted to a Testing Protocol
• integrated into the Experimentation Brain Operating Guide
• fed back into Affiliate Brain or Research Brain (NEW)
• discarded if invalidated
Scope
This reference applies to:
• temporary testing insights awaiting validation
• structured experimentation learnings not yet promoted into canon
• pre-promotion hypothesis and test-method intelligence
• protocol-candidate signals inside Experimentation Brain
• working storage for later review, validation, or discard
🔴 Extended Scope (NEW)
This page also captures:
• test result learnings from completed lifecycle stages
• interpretation insights not yet stable
• forecast vs actual variance patterns
• repeatable behavioural signal observations
This page governs temporary insight preservation only.
It does not govern:
• live canon authority
• experiment approval
• capital allocation
• direct protocol enforcement
• formal experimentation governance
• permanent operating rules
Core Principle
Not all learning is ready for canon.
Unstructured learning is easily lost.
Unvalidated learning is dangerous.
This log ensures:
👉 learning is captured early
👉 but promoted slowly
Definition / Rules
Working Document Rule
This page exists to preserve useful signals before they are strong enough for formal promotion.
It is not a canon page.
It is not an authority page.
It is a governed holding layer for potentially valuable experimentation intelligence.
Promotion Rule
An insight may be considered for promotion only if:
• it remains valid for at least 7 days
• it continues to appear structurally useful
• it does not conflict with canon or higher governance
• it survives review without material contradiction
• it demonstrates repeatability or stability (NEW)
Until then, the insight remains provisional.
Use Rule
Insights here may inform:
• future protocol drafting
• experimentation operating guides
• review discussions inside HeadOffice or Experimentation Brain
• refinement of hypotheses (NEW)
• improvement of measurement planning (NEW)
Insights here may not:
• override existing canon
• authorise live testing decisions by themselves
• justify capital movement by themselves
• be treated as formal experiment law
🔴 Learning Capture Standard (NEW)
All new entries must include:
• What was expected (forecast)
• What occurred (actual result)
• Variance explanation
• Signal reliability assessment
• Lifecycle stage source (where learning came from)
Incomplete learning must not be logged.
🔴 Lifecycle Link Rule (NEW)
All insights must originate from:
👉 Experimentation Brain Test Lifecycle
Entries must reference:
• test stage (Interpretation, Decision, etc.)
• test classification (Winner, Weak, Failure, etc.)
This prevents:
→ random, untraceable insights
🔴 Data Integrity Dependency (NEW)
Insights must only be logged if:
• underlying data passed validation
• signal integrity is acceptable
• measurement integrity is confirmed
If not:
→ insight must be flagged as low trust
Log Entries
Insight ID: TST-001
Source: CXL – What to Test
Signal: Tests are solutions to identified problems
Implication: No problem → No test
Category: Hypothesis Discipline
Promotion Candidate: Yes
Canon Impact: No
Protocol Impact: Yes
Insight ID: TST-002
Source: CXL – How Many Changes Per Test
Signal: Multiple changes allowed if tied to one hypothesis
Implication: No stacked random changes
Category: Hypothesis Coherence
Promotion Candidate: Yes
Canon Impact: No
Protocol Impact: Yes
Insight ID: TST-003
Source: CXL – A/B vs MVT
Signal: Avoid MVT at low traffic levels
Implication: Default to A/B or A/B/N
Category: Traffic Sensitivity
Promotion Candidate: Yes
Canon Impact: No
Protocol Impact: Yes
Insight ID: TST-004
Source: CXL – Bandits
Signal: Bandits maximise revenue, not learning
Implication: Classify test type before execution
Category: Test Classification
Promotion Candidate: Yes
Canon Impact: No
Protocol Impact: Yes
Insight ID: TST-005
Source: CXL – Existence Testing
Signal: Remove elements to test necessity
Implication: Identify persuasion noise
Category: Friction Reduction
Promotion Candidate: Yes
Canon Impact: No
Protocol Impact: Yes
Insight ID: TST-006
Source: CXL – Iterative Testing
Signal: Small tests compound learning
Implication: Avoid abandoning supported hypotheses prematurely
Category: Iterative Discipline
Promotion Candidate: Yes
Canon Impact: No
Protocol Impact: Yes
Drift Protection
The system must prevent:
• working insights being mistaken for canon
• temporary signals being treated as formal test authority
• capital or execution decisions being justified from this page alone
• unvalidated learnings mutating protocol pages too early
• useful signals being lost due to lack of temporary storage
• insights being logged without lifecycle context (NEW)
• insights being logged from invalid data (NEW)
This page preserves signal.
It does not promote signal automatically.
Architectural Intent
Testing Intelligence Log exists to give Experimentation Brain a controlled place to hold useful testing learnings before they are strong enough to become permanent structure.
Its role is to:
• reduce signal loss
• improve promotion discipline
• connect lifecycle → learning → future decisions (NEW)
• stop canon from being mutated too quickly by fresh but unproven insights
Final Rule
If an insight is useful but not yet proven stable:
→ it belongs here — not in canon
Temporary intelligence must stay temporary until it earns promotion.
Change Log
Version: v1.2
Date: 2026-04-25
Author: MWMS HeadOffice / Experimentation Brain
Change
Upgraded log to include:
• lifecycle linkage requirement
• forecast vs actual learning structure
• data validation dependency
• improved integration with Experimentation Brain system
Version: v1.1
Date: 2026-03-15
(UNCHANGED)
Version: v1.0
Date: 2026-03-03
(UNCHANGED)