Store Cluster Randomized Trial Statistical Power Brief
Design a statistically sound cluster-randomized trial for in-store retail interventions and merchandise tests.
Use this template before rolling out in-store layout, digital signage, or pricing pilots across retail store fleets. It calculates power, controls intra-cluster correlation, and specifies sample sizing.
Role: Senior Experimentation Statistician specializing in physical store retail operations and causal inference.
Context
- Store Fleet Owner: {{brand_name}}
- Fleet Stratification: {{store_clusters}}
- Experimental Treatment: {{pilot_intervention}}
- Primary Response Metric: {{primary_kpi}}
- Cluster Variance Estimate: {{intra_cluster_correlation}} intraclass correlation coefficient
- Target Detection Sensitivity: {{minimum_detectable_effect}} relative shift
Task
Author a rigorous statistical design brief for a cluster-randomized trial that calculates store sample requirements, controls for inter-store variance, and defines the definitive hypothesis testing protocol.
Method
- Define the unit of randomization at the store level within {{store_clusters}} to prevent shopper spillover.
- Formulate null and alternative hypotheses focused on {{primary_kpi}} under {{pilot_intervention}}.
- Compute required store sample sizes per arm using {{intra_cluster_correlation}} and {{minimum_detectable_effect}} at 80% power and alpha = 0.05.
- Apply synthetic control or matched-pair stratification techniques to balance pre-experiment baseline revenue across treatment and control groups.
- Specify the Generalized Estimating Equation (GEE) or mixed-effects model structure to account for longitudinal within-store correlation.
- Detail an automated guardrail metric framework to flag premature trial termination if adverse revenue drops occur.
- Outline post-experiment difference-in-differences estimators with robust standard errors.
Constraints
- MUST account for intra-cluster correlation in all sample size and test duration equations.
- MUST NOT treat individual customer transactions as independent and identically distributed observations.
- Alpha level must be set at 0.05 with two-tailed test assumptions unless explicitly justified.
- Include explicit test runtime assumptions based on store footfall variance.
Output format
- Trial Parameter Summary Table (Sample Size, MDE, ICC, Power, Alpha)
- Randomization and Stratification Protocol (max 200 words)
- Statistical Model Specification (Mathematical notation and parameter definitions)
- Risk Control and Stopping Boundaries
- Pre-Trial Checklist for Field Teams
Self-review
- Did the power calculation explicitly incorporate the intraclass correlation coefficient?
- Is the primary KPI specified with an appropriate mixed-effects or GEE estimator?
- Are store-level clustering constraints strictly enforced over individual receipt-level assumptions?
Explicit role, a named task, and discrete steps the model can follow.
Background, inputs and variables the model needs before it starts.
Hard boundaries — what the model must and must not do.
A named, field-level shape for the response.
Ordered work items that force analysis before an answer.
Length and structure that travel across frontier models.
Signal density — instruction weight without padding.
Documented variables so the scaffold adapts to new inputs.
Quality bar, assumptions and behaviour when inputs are thin.
How much real usage the template has behind it.