Document Type: Framework
Status: Canon
Version: v1.6
Authority: Data Brain
Applies To: All MWMS environments where performance credit is assigned to channels, assets, messages, or decision environments
Parent: Data Brain Canon
Last Reviewed: 2026-04-26
Purpose
Attribution Reliability Framework defines how MWMS evaluates, interprets, and uses performance credit assigned to marketing activities, environments, and behavioural influences.
Attribution does not reveal absolute truth.
Attribution provides directional interpretation of influence.
Incorrect attribution produces incorrect optimisation decisions.
Incorrect optimisation decisions distort resource allocation.
Distorted resource allocation weakens growth efficiency.
Attribution Reliability Framework ensures MWMS:
- interprets contribution signals cautiously
- applies attribution consistently across all Brains
- uses attribution as a controlled decision system
- understands attribution as behavioural insight, not just reporting
Core Principle
Attribution indicates influence, not certainty.
Attribution must never be treated as:
- exact truth
- causal proof
- complete behavioural explanation
Attribution is:
→ directional
→ partial
→ model-dependent
→ visibility-limited
🔴 Attribution Decision System Rule
Attribution is not a reporting metric.
Attribution is a decision input layer.
All optimisation decisions that depend on performance credit must:
- explicitly define attribution logic used
- apply consistent attribution interpretation
- align with system-wide attribution rules
If attribution is used inconsistently:
→ decision quality collapses
→ optimisation becomes unstable
Attribution must only drive decisions when:
- signal reliability is acceptable
- confidence level is defined
- model limitations are understood
🔴 Cross Brain Attribution Alignment Rule
All Brains must operate on aligned attribution logic.
The system must prevent:
- different attribution models used across channels
- conflicting performance interpretations
- isolated channel-level attribution logic
If:
- Ads Brain uses one attribution logic
- Affiliate Brain uses another
- Conversion Brain uses another
→ MWMS produces conflicting optimisation signals
HeadOffice must enforce:
→ shared attribution interpretation layer
→ shared decision logic
→ shared confidence understanding
🔴 Channel Position Bias Rule
Attribution models inherently bias certain positions in the journey.
Typical bias patterns:
End-of-journey bias:
- brand PPC
- affiliates
- direct traffic
Early/mid-journey undervaluation:
- SEO
- generic PPC
- content
- awareness channels
Consequences:
- over-investment in closing channels
- under-investment in discovery channels
- distorted growth strategy
Attribution must be interpreted with position awareness.
🔴 Journey Interpretation Rule
Attribution exists to understand behavioural journeys, not just assign value.
Attribution must be used to:
- analyse full journey structure
- identify key behavioural steps
- understand conversion pathways
- identify drop-off points
- detect lag between interactions
Attribution must not be reduced to:
- single-number reporting
- channel comparison only
Attribution is a behavioural intelligence layer.
🔴 Signal Reliability Dependency Rule
Attribution reliability depends on the reliability of underlying signals.
If events are:
- missed
- delayed
- partially captured
- unreliable under real conditions
→ attribution accuracy is reduced
Attribution cannot be more reliable than the signals it is built on.
🔴 Visibility Gap Impact Rule
Attribution must account for unobservable behaviour.
Examples:
- iframe interactions
- third-party tools
- cross-device journeys
- external conversion environments
Where visibility gaps exist:
→ attribution represents only partial reality
Confidence must be reduced accordingly.
🔴 Partial Journey Rule
User journeys are often incomplete.
Examples:
- missing touchpoints
- untracked interactions
- delayed conversions
Incomplete journeys result in:
- over-crediting visible touchpoints
- under-crediting hidden influence
Attribution must be interpreted as partial journey mapping.
🔴 Event Loss Impact Rule
Missed events distort attribution.
Examples:
- clicks lost before navigation
- tracking blocked by timing
- missing interaction signals
Consequences:
- incorrect channel credit
- distorted funnel contribution
- false optimisation signals
🔴 Cross-System Fragmentation Rule
Signals may exist in different systems without full alignment.
Examples:
- Ads platform sees conversion
- GA4 does not
- backend shows different totals
Fragmentation results in:
→ multiple incomplete views of reality
No single system provides complete attribution.
🔴 Attribution Trust Rule
Attribution must be treated as probabilistic interpretation.
Trust must be based on:
- signal quality
- data completeness
- system alignment
- validation outcomes
Attribution Model Dependency Rule
Attribution outputs depend on model design.
Different models produce different interpretations.
Model selection influences:
- channel valuation
- optimisation decisions
- budget allocation
No model represents absolute truth.
Attribution Visibility Limitation Principle
Attribution only reflects observable interactions.
Hidden influence cannot be measured directly.
All attribution outputs must be interpreted as:
→ observable subset of behaviour
🔴 Cross Platform Attribution Validation Rule
Attribution must be compared across:
- analytics platforms
- ad platforms
- backend systems
Differences must be investigated.
Consistency increases confidence.
🔴 Conversion Source Conflict Rule
Different systems may assign credit differently.
Conflicts must be:
- identified
- documented
- understood
Unresolved conflict reduces decision confidence.
🔴 Model Bias Awareness Rule
All attribution models introduce bias.
Bias types include:
- recency bias
- position bias
- interaction weighting bias
Bias must be:
- acknowledged
- documented
- considered in decision making
Attribution Influence Categories
Attribution signals may represent:
- direct influence
- indirect influence
- assistive contribution
- behavioural reinforcement
Interpretation must consider influence type.
Multi Touch Behaviour Principle
Conversions result from multiple interactions.
Single-touch attribution is structurally incomplete.
Multi-touch behaviour must be assumed as default.
🔴 Attribution Validation Process
Attribution must be validated through:
- cross-system comparison
- historical consistency checks
- anomaly detection
- segmentation review
Unvalidated attribution must not drive decisions.
🔴 Attribution Confidence Levels
Attribution outputs must be classified:
- high confidence
- medium confidence
- low confidence
Confidence determines:
→ decision weight
🔴 Confidence Adjustment Rule
Confidence must be reduced when:
- event reliability is low
- signal flow is incomplete
- visibility gaps are significant
- journey capture is partial
- data integrity conditions are not met
- segmentation context is inconsistent
Confidence increases when:
- signals are reliable
- systems are aligned
- segmentation is consistent
- data integrity is validated
🔴 Attribution Stability Rule
Attribution outputs must remain stable under:
- time variation
- segmentation changes
- system updates
Instability indicates weak attribution reliability.
Behavioural Signal Hierarchy Compatibility
Attribution must align with behavioural signal hierarchy.
Higher-quality signals must take precedence.
Attribution Interpretation Model
Attribution must be interpreted within:
- behavioural context
- journey structure
- segmentation layer
- model limitations
Attribution Sensitivity Rule
Small changes in input data may produce large attribution changes.
Sensitivity must be monitored.
High sensitivity reduces reliability.
🔴 Attribution Failure Conditions
Attribution must not be trusted when:
- signal loss is high
- fragmentation is unresolved
- model bias is unaccounted for
- validation has not occurred
Interaction with Signal Integrity Framework
Attribution depends on signal integrity.
Signal Integrity Framework governs event reliability.
Interaction with Measurement Strategy Framework
Measurement Strategy defines what is captured.
Attribution depends on measurement completeness.
Interaction with Signal Design Specification Framework
Signal design determines attribution input quality.
Poor design produces unreliable attribution.
Interaction with Data Layer Architecture Framework
Data layer defines how signals are structured and transported.
Attribution depends on consistent data architecture.
Interaction with Behavioural Event Analysis Framework
Behavioural Event Analysis provides context to attribution signals.
Interaction with Experimentation Brain
Attribution supports experiment interpretation.
Experimentation Brain must consider attribution limitations.
🔴 Decision Usage Rule
Attribution must only influence decisions when:
- confidence level is defined
- validation is complete
- model limitations are understood
If not:
→ decisions must be treated as high risk
Failure Modes Prevented
- misinterpreting incomplete journeys
- over-crediting visible channels
- ignoring hidden influence
- trusting unreliable attribution
- conflicting cross-brain decisions
- channel bias distortion
- treating attribution as absolute truth
Drift Protection
The system must prevent:
- event loss affecting attribution unnoticed
- signal fragmentation increasing over time
- visibility gaps expanding due to system changes
- segmentation inconsistency affecting attribution
- data integrity degradation affecting attribution
- cross-brain attribution divergence
Architectural Intent
Attribution Reliability Framework ensures MWMS:
→ understands attribution is imperfect
→ uses attribution as controlled decision input
→ aligns attribution across all Brains
→ interprets behaviour, not just numbers
Final Rule
Attribution explains influence patterns, not causal certainty.
If attribution is not validated:
→ it must not drive decisions
🔴 Final Extension
If attribution is:
- incomplete
- inconsistent
- misaligned across system
- based on weak signals
→ decisions must be treated as high risk
Change Log
Version: v1.6
Date: 2026-04-26
Author: Data Brain
Change:
Major upgrade integrating attribution as a decision system:
- added Attribution Decision System Rule
- added Cross Brain Attribution Alignment Rule
- added Channel Position Bias Rule
- added Journey Interpretation Rule
- expanded framework from reliability-only to decision intelligence layer
Change Impact Declaration
Pages Created:
None
Pages Updated:
Data Brain Attribution Reliability Framework
Pages Deprecated:
None
Registries Requiring Update:
No
Canon Version Update Required:
No
Change Log Entry Required:
Yes
End of Framework