Which measurement year is a projection based on?

Every projection states the measurement year and the Star Year it maps to, because the two are not the same and the difference decides whether a change affects the current scope or a later one.

How do you handle outlier deletion and guardrails?

Both are implemented from the published CMS technical notes rather than inferred, and the applied version is recorded on every output alongside the measure registry version.

Can I reproduce your numbers?

Yes, and you should. Every figure ships with its computation trail and its source. Back test it against a Star Year where you already know the answer, and tell us where we are wrong.

Do you guarantee a Star rating?

No. We influence a number we do not control, so we decline rating guarantees and never tie fees to a rating outcome.

For the person who has to defend the number

Here is the methodology. Break it, and tell us where it fails.

This page exists because you should not trust a projection you cannot inspect. It is public and ungated on purpose. If our method does not survive your reading of it, we would rather you find that out now than in October.

What the projection is built on

Published CMS Star Ratings data, and nothing else, for the entry product. The data is public and it is public for everyone, so the value is not secret inputs. It is specification discipline and traceability: a maintained measure registry with versioned specifications, so a methodology change is a version bump rather than a silent break.

Every figure traces to its CMS source. Where CMS published no value for a contract, domain, or measure in a given year, the field stays blank rather than imputed. A blank is information. An imputed value is a guess wearing a number's clothes.

Where the math is deterministic

All financial and measure math is deterministic. Sensitive calculations never run through a generative path. A language layer reports validated results in plain English and does not compute, infer, or estimate any value. In a regulated setting a single incorrect figure is a compliance event, not a rounding error, so the analytic source of truth is always a validated computation and the model never originates a number on its own.

Common questions

Which measurement year is a projection based on?: Every projection states the measurement year and the Star Year it maps to, because the two are not the same and the difference decides whether a change affects the current scope or a later one.
How do you handle outlier deletion and guardrails?: Both are implemented from the published CMS technical notes rather than inferred, and the applied version is recorded on every output alongside the measure registry version.
Can I reproduce your numbers?: Yes, and you should. Every figure ships with its computation trail and its source. Back test it against a Star Year where you already know the answer, and tell us where we are wrong.
Do you guarantee a Star rating?: No. We influence a number we do not control, so we decline rating guarantees and never tie fees to a rating outcome.

What we will not claim

We are not a certified measure engine and we are not an accredited measure vendor. We will never imply otherwise. We do not guarantee a rating. We do not publish an accuracy percentage without stating the holdout that produced it, because a number without its definition is marketing rather than evidence.

Found a hole? Tell us.

Talk to us