The evaluation should not change after the result appears
Healthcare programs and vendor deployments generate many possible outcomes, populations, time windows, comparisons, and cost definitions. If those choices are made after results are inspected, a favorable claim may reflect analytical flexibility rather than a stable evaluation question.
A Measurement Contract records the important choices before interpretation. It is not a legal contract. It is a versioned agreement about what will be measured, how attribution will be assessed, when the analysis must refuse a claim, and how value will be calculated.
THE CORE RULE
Fix the question and decision rules before choosing the result.
What the template contains
Version 0.1 contains ten sections.
Contract
Identifier, version, status, date, decision owner, and analysis owner.
Decision
The question, deadline, and actions the evidence may inform.
Intervention
Name, version, eligibility, exposure, start date, and implementation fidelity.
Population
Unit of analysis, inclusion, exclusion, and planned subgroup questions.
Causal question
Estimand, outcome, counterfactual, comparison design, causal graph, and adjustment set.
Evidence
Baseline, observation period, required fields, timing, provenance, and quality rules.
Analysis
Primary estimator, secondary methods, diagnostics, sensitivity, and refusal conditions.
Value
Benefit unit, eligible volume, unit value, full costs, and ROI formulas.
Claim
Minimum evidence level, precision, permitted statement, limitations, and approval.
Governance
Data rights, review roles, amendments, version history, and preservation.
Begin with the decision, not the data
The first question is what action the evaluation must inform. A decision about expanding a program may require effect estimates for eligible sites. A renewal decision may require attributable value during the contract period. An internal improvement decision may need subgroup and implementation evidence even when financial return is secondary.
Once the decision is explicit, the evaluation can define the population, outcome, time horizon, and evidence threshold that fit it.
Define the counterfactual
The template requires a plain-language statement of what would likely have happened without the intervention. It also requires the design used to estimate that alternative.
Possible designs include randomized or phased rollout, matching, weighting, difference in differences, interrupted time series, regression discontinuity, or another justified approach. Naming a method is not enough. The contract should identify the assumptions and diagnostics that make its result interpretable.
Add refusal conditions before analysis
Refusal conditions define when the planned claim cannot be issued. Common examples include missing required fields, insufficient sample size, weak overlap, failed balance, lack of stable pre-trends, concurrent interventions that cannot be separated, or changes to the outcome definition.
The template treats a withheld claim as a valid result. This protects the decision from a last-minute substitution of outcome, population, or method.
Define value and full cost
The value section separates the causal effect, eligible volume, agreed unit value, attributable benefit, and full approved cost. It includes fields for vendor, implementation, labor, oversight, transition, and operating costs.
The contract should also state whether saved time represents cash savings, avoided hiring, expanded capacity, improved timeliness, or another form of value. These are different claims and should not be silently combined.
Use the contract with the Impact Passport
The Measurement Contract governs the question before analysis. The Impact Passport records what the evidence supported after analysis. The passport should reference the exact contract version and preserve any amendment history.
Together, the two artifacts make the analytical choices and resulting claim easier to inspect, reproduce, and challenge.
The structure also aligns with the CDC Program Evaluation Framework, which begins with context and program description, focuses the questions and design, gathers credible evidence, supports conclusions, and acts on findings.
Download v0.1
YAMLMeasurement Contract templateEditable structure with implementation guidance JSONValidation schemaDraft 2020-12 schema for automated structure checks
The schema validates the presence and basic shape of required sections. It cannot determine whether a comparison is credible, whether assumptions are defensible, or whether a claim is scientifically justified. Those remain evaluation responsibilities.