Document Type: Framework
Status: Canon
Authority: HeadOffice
Applies To: Experimentation Brain, Affiliate Brain, Ads Brain, Conversion Brain, Data Brain, Ecommerce Brain, AIBS Brain
Parent: Experimentation Brain Canon
Version: v1.0
Last Reviewed: 2026-05-03
Purpose
The Optimization Target Selection Framework defines what MWMS optimizes for in any test, personalization, or campaign.
Selecting the correct optimization target is critical because:
- the wrong target produces misleading results
- the wrong metric creates false winners
- the wrong signal wastes budget
- the wrong focus prevents business impact
Optimization success is not driven by:
- better creatives
- more tests
- better tools
It is driven by:
→ selecting the correct target to optimize
Scope
This framework applies to:
- A/B testing
- personalization systems
- funnel optimization
- campaign optimization
- ecommerce optimization
- affiliate offer testing
- AIBS decision systems
It governs how MWMS selects:
- primary KPIs
- proxy metrics
- validation methods
- testing targets
Definition
Optimization Target
The primary measurable outcome used to evaluate the success or failure of a test, campaign, or personalization system.
Core Principle
Optimize for the closest measurable action that reliably leads to real business value.
The Optimization Ladder
All optimization targets sit on a continuum:
Fast Feedback → High Business Value
Level 1 — Surface Metrics (Fast, Weak)
Examples:
- clicks
- CTR
- scroll depth
- engagement
- video watch
Characteristics:
- very fast feedback
- high volume
- easy to measure
- weak correlation to revenue
Risk:
- false positives
- activity increases without profit
Level 2 — Mid Funnel Actions (Balanced)
Examples:
- landing page progression
- form starts
- email signups
- add to cart
- product views
- demo requests
Characteristics:
- balanced signal strength and speed
- strong connection to conversion
Role:
→ primary MWMS testing layer
Level 3 — Conversion Events (High Value)
Examples:
- purchases
- lead submissions
- completed applications
- paid signups
Characteristics:
- direct business impact
- lower volume
- slower feedback
Risk:
- requires traffic
- slower learning
Level 4 — Long Term Value (Slow, Strategic)
Examples:
- lifetime value
- retention
- repeat purchases
- subscription duration
Characteristics:
- highest importance
- very slow feedback
- not usable for real-time testing
Core Trade-Off
Every optimization decision balances:
| Speed of Learning | Strength of Signal |
|---|---|
| Fast | Weak |
| Slow | Strong |
MWMS Decision Rule
Step 1 — Identify Business Goal
Define:
- revenue
- qualified leads
- profit per visitor
- customer quality
Step 2 — Assess Distance To Conversion
- short funnel → optimize final conversion
- long funnel → optimize intermediate steps
Step 3 — Select Closest Reliable Proxy
If final conversion is too slow:
→ choose a proxy that strongly correlates with it
Examples:
- product view → purchase
- add to cart → revenue
- qualified lead → sales conversion
Step 4 — Validate Proxy Over Time
Proxy metrics must prove correlation.
Validation methods:
- compare proxy vs final conversion
- track downstream outcomes
- use control groups
Funnel Optimization Strategy
Strategy A — End Goal Optimization
Used when:
- short funnel
- high traffic
- clear conversion
→ optimize final conversion
Strategy B — Step Based Optimization
Used when:
- long funnel
- multi-step journeys
- signal dilution risk
→ optimize each step individually
Strategy C — Hybrid Model (MWMS Default)
- optimize step progression
- track final conversion
- validate alignment
Signal Dilution Problem
When optimizing too far downstream:
- upstream changes lose visibility
- noise hides results
- tests appear ineffective
MWMS Rule
If signal disappears:
→ move optimization target closer
Customer Lifecycle Optimization
New Visitors
Optimize for:
- engagement
- first action
- email capture
Returning Visitors
Optimize for:
- progression
- product interaction
Existing Customers
Optimize for:
- upsell
- cross-sell
- retention
High Value Users
Optimize for:
- revenue-driving actions
- high-intent behaviour
Qualification Layer
Not all conversions are equal.
Rule
Optimize for qualified outcomes, not volume.
Examples
- high-quality leads
- high AOV buyers
- target persona matches
Long Term Validation
Short-term wins must be verified.
Methods
- holdout groups
- cohort tracking
- cross-system validation
Optimization Target Selection Flow
- define business goal
- check measurability
- select proxy if needed
- validate reliability
- confirm speed of feedback
- confirm traffic sufficiency
- choose strategy
- define validation
Common Mistakes
- optimizing for clicks
- optimizing too far down funnel
- ignoring qualification
- no long-term validation
- using one metric for all users
Governance Role
This framework enforces:
- consistent KPI selection
- alignment between Brains
- reliable experimentation signals
- decision clarity
It acts as:
→ the control layer for all optimization decisions
Relationship To Other MWMS Standards
- Experimentation Brain Structured Testing Protocol
- Data Brain Measurement Integrity Framework
- Conversion Brain Behaviour Trigger Framework
- Ecommerce Brain Personalization Journey Framework
- Customer Brain Segmentation Framework
Drift Protection
The system must prevent:
- vanity metric optimization
- unclear optimization targets
- unvalidated proxy usage
- misaligned KPIs
- optimization without business relevance
- inconsistent metric selection across Brains
Architectural Intent
This framework ensures:
- all MWMS optimization aligns with real business outcomes
- testing produces meaningful insights
- signals remain reliable across the system
- performance improvements scale correctly
It transforms optimization from:
→ activity-based metrics
into:
→ outcome-driven decision systems
Final Rule
If the optimization target is unclear:
→ the test must not be launched
Change Log
Version: v1.0
Date: 2026-05-03
Author: HeadOffice
Change:
Created Optimization Target Selection Framework defining optimization ladder, proxy logic, funnel strategy, qualification rules, and validation methods based on CXL experimentation systems.
Change Impact Declaration
Pages Created:
Experimentation Brain Optimization Target Selection Framework
Pages Updated:
None
Pages Deprecated:
None
Registries Requiring Update:
Experimentation Brain Page Registry
Canon Version Update Required:
Yes
Change Log Entry Required:
Yes