THE SERVICE
Independent evaluation for healthcare interventions and vendor deployments.
Usage, time saved, completed activities, and changes in a business metric are useful observations. They do not by themselves establish that an intervention caused the change. Causalytics specifies the decision, outcome, comparison, assumptions, and value model before the result is interpreted.
01 / DEFINE
Measurement Contract
Fix the intervention version, population, outcome, causal diagram, comparison, estimand, diagnostics, cost model, and refusal conditions before analysis.
02 / ESTIMATE
Causal analysis
Use the design the evidence can support, such as a randomized rollout, difference-in-differences, matching, weighting, or an interrupted time series. Report uncertainty and sensitivity.
03 / STATE
Impact Passport
Issue a portable record of the estimate, attributable value, limitations, approvals, lineage, and exact claim permitted by the evidence. Withhold the claim when the design is inadequate.
WHAT YOU PROVIDE
Evidence needed to begin
- The intervention, program, or vendor version under evaluation
- The eligible population and unit of analysis
- Adoption timing, exposure, and relevant rollout variation
- The outcome that matters to the buyer
- A plausible baseline, comparison group, or discontinuity
- A customer data owner who can authorize and explain the data
- Vendor, implementation, labor, oversight, transition, and operational cost inputs
- The business or procurement decision the evaluation must inform
WHAT YOU RECEIVE
Inspectable deliverables
- An evidence-readiness assessment
- A pre-deployment or pre-analysis Measurement Contract
- An executable causal diagram and identification record
- Reproducible estimates, diagnostics, and sensitivity results
- An attributable ROI model with stated assumptions
- An Impact Passport with a supported, limited, or withheld claim
INDEPENDENCE
Payment does not depend on a favorable result.
The commercial structure uses a fixed scope for evaluation design and implementation, followed by a defined measurement phase when evidence must accumulate. A positive finding is never guaranteed. A limited or withheld claim is a valid output when the data or identification strategy cannot support attribution.
WORKING METHOD
See the causal system, not a results claim.
The demonstration uses synthetic data to show the method: define the causal graph, derive the adjustment set, compare estimates, test diagnostics, calculate attributable value, and carry the result into an Impact Passport. It is evidence that the evaluation workflow exists. It is not evidence of production impact.
COMMON QUESTIONS
What an organization or vendor should know.
Do we need to replace the existing program or vendor?
No. Causalytics evaluates the intervention and its evidence. Existing operations, vendors, and source systems remain in place.
Can an intervention that is already live be evaluated?
Sometimes. The answer depends on baseline data, rollout timing, exposure records, a credible comparison, and whether important confounders were measured. Causalytics begins with an evidence-readiness assessment.
Does an evaluation always produce an ROI number?
No. Attributable ROI is issued only when the causal effect and cost inputs meet the Measurement Contract. Otherwise, the Impact Passport limits or withholds the claim and records why.
What kinds of interventions can be evaluated?
Examples include care delivery changes, access and equity programs, member interventions, operational redesign, and AI or technology deployments. The intervention must be bounded, measurable, and connected to a decision.
Can the result support a procurement, renewal, or program decision?
Yes, when the evidence meets the agreed claim threshold. The Impact Passport gives decision makers an inspectable record of the estimate, attributable value, uncertainty, limitations, and data lineage.
START WITH ONE INTERVENTION
Bring the intervention, the evidence, and the decision it must inform.
The first conversation determines whether the available evidence can support a bounded evaluation and what must be captured next.