Paid Social Incrementality Lift Testing Framework

Role: Head of Growth Experimentation and Causal Inference

Context

• Channel environment: {{paid_social_platform}}
• Primary conversion metric: {{primary_conversion_event}}
• Minimum Detectable Effect target: {{minimum_detectable_effect}}
• Test market units: {{geo_holdout_clusters}}
• Allocated experimentation budget: {{testing_budget_allocation}}
• Statistical threshold: {{confidence_interval_target}}

Task

Author a statistical incrementality testing framework utilizing synthetic control matching and geo-holdouts to isolate the true causal lift generated by {{paid_social_platform}} on {{primary_conversion_event}}.

Method

1. Establish the statistical test design requirements based on the {{minimum_detectable_effect}} target.
2. Define the synthetic control pairing methodology across {{geo_holdout_clusters}} using historical pre-test correlation metrics.
3. Formulate the power analysis model to calculate required test duration under {{testing_budget_allocation}}.
4. Specify the test cell vs. control cell spend pacing and delivery isolation protocols on {{paid_social_platform}}.
5. Establish the mathematical equation for calculating incremental Cost Per Acquisition (iCPA) and incremental ROAS (iROAS).
6. Define standard error calculation and confidence interval validation at the {{confidence_interval_target}} level.
7. Construct decision rules for scaling, maintaining, or defunding campaigns based on verified causal incrementality.

Constraints

• MUST specify explicit synthetic control matching algorithms (e.g., Euclidean distance or Dynamic Time Warping).
• MUST NOT count non-incrementality-adjusted platform conversions toward test validation.
• Statistical significance checks must meet or exceed {{confidence_interval_target}}.
• Experimental design must include safeguards against geographical cross-contamination.

Output format

Structure the framework into four clear components:

1. Experimental Design Architecture (geo-pairing criteria and sample sizing equations)
2. Execution & Isolation Protocol (campaign flighting rules and budget management)
3. Causal Inference Calculation Engine (formulas for lift, standard errors, and iCPA)
4. Allocation Decision Matrix (threshold-based scaling policy based on results) Keep total output under 600 words.

Self-review

• Confirm that all 6 context variables appear meaningfully in the methodology and rules.
• Check that equations for incremental CPA (iCPA) and lift are explicitly written out.
• Validate that experimental power and geo-holdout contamination risks are addressed.