Back to blog
Rate Data

What 'Confidence-Scored' Rate Data Actually Means

Raw MRF rows can't be taken at face value. Confidence scoring is an honest signal of how much weight a rate can bear before you build a negotiation on it — real rates made trustworthy, never modeled estimates.

TierBench

"Confidence-scored" is a phrase we use a lot, and it's worth being precise about what it means — because it's easy to hear it as a hedge, and it's the opposite. Confidence scoring isn't us being unsure of the data. It's us being honest about something every rate carries whether or not anyone labels it: how much weight it can actually bear.

Why a raw row can't be taken at face value

By now the reasons should be familiar if you've followed this series. A single machine-readable-file row looks like a fact, but several things sit between that row and a number you can act on:

  • Entity ambiguity — the provider reference has to be resolved to the real organization behind it, and that's genuinely hard. (See what an MRF is.)
  • Carve-outs — the rate you're reading may not be the one that governs behavioral health, if a carve-out network holds that contract.
  • Contract-structure variation — the same code can be a per-diem at one facility and a case rate at another, which changes what the number means entirely. (See the billing grammar of behavioral health.)
  • Ghost rates — the row might be terminated, duplicate, or legacy, and dead in practice. (See ghost rates.)

Any one of these can make a row mean less — or something entirely different — than it appears. Stack them and "just look up the rate" stops being reliable. That's not a reason to distrust the data. It's a reason to measure how much to trust each piece of it.

What confidence scoring is for

Confidence scoring exists to answer one question before you build on a number: how firmly can I lean on this?

Instead of handing you a rate and walking away, TierBench scores how much to trust each rate using real signals in the source data — for instance, whether it comes from a single facility's own schedule or from a broader multi-group arrangement — and filters or flags what can't be trusted: terminated, duplicate, ghost, and legacy postings.

The output isn't a vaguer number. It's the same number plus an honest read on its footing — so a rate you'd stake a negotiation on and a rate you'd treat as a loose indication don't look identical on the page. That distinction is exactly what a raw file erases.

What it is — and isn't

A few things confidence scoring is deliberately not, because the misreads matter:

  • It is not a modeled or estimated rate. The rates stay real — actual negotiated rates from the source. The score sits on top; it never replaces the number with a projection. TierBench works from actual rates, never estimates.
  • It is not a promise of a specific payment. A high-confidence rate is a well-supported signal, not a guarantee of what a given claim will pay.
  • It is not a black box you have to take on faith. The point of a confidence signal is transparency — telling you when to lean in and when to be careful — not asking you to trust blindly.

We keep the public description at exactly this level on purpose; you can read the same framing on the Coverage & Methodology page, which lays out the pipeline from extraction through scoring without dressing it up as something it isn't.

Real rates, made trustworthy

The phrase we keep coming back to is "actual rates, made trustworthy." That's the whole idea in four words. The rates are real — pulled from what payers actually publish. Trustworthy is the layer we add: resolving them to the real facility, reading them in context, filtering the dead ones, and scoring how firmly to lean on each survivor.

The alternative — a modeled estimate — starts from a formula's guess and dresses it up as precision. Confidence-scored actual rates start from something real and are honest about its footing. For a facility about to sit across from a payer, that difference is the difference between a position you can defend and a number you're hoping holds.

See it on your rates

Confidence scoring is most useful when it's pointed at your own contracts and your own peers. See which payers we hold rate data on, read the rate-data FAQ for the short answers, or book a 20-minute call to walk through where your rates stand.


Back to all posts