House Ways & Means just advanced an MLR Transparency Act, 42–0 — forcing Medicare Advantage plans to show how much revenue actually reaches patient care. The ledger it wants to open already has a public v0: CMS's contract-level medical-loss-ratio file. Here is every MA and Part D contract from the 2023 filings — 635 contracts, one dot each, plotted against the 85% floor. Hover any dot. Then drag the enrollment floor and watch the wildest ratios dissolve.
$537.8B flowed through these filings. $52.4M came back.
Every MA and Part D contract must file how much of its revenue went to claims and quality improvement — the medical loss ratio. Fall below 85% and the plan owes CMS a remittance. In contract year 2023, that floor caught 23 of 635 contracts, and the money returned was 0.01% of the revenue in the file. The floor exists. It almost never bites.
635
MA + Part D contracts, CY2023
$537.8B
total revenue in the filings
90.7%
enrollment-weighted MLR
$52.4M
remittances owed (23 contracts)
The explorer
635 contracts, one dot each
Horizontal: average enrollment (log scale — a 220-member plan and SilverScript's 6.1M share one chart). Vertical: adjusted MLR — the share of revenue that went to claims + quality improvement. Dot size: total revenue. Red = below the 85% floor (owes a remittance). Navy = 85–100%. Green = over 100% — claims exceeded revenue, the plan lost money on care.
Minimum enrollment0
Segment
MLRs above 150% are pinned to the top edge (hover for true value — one tiny PDP filed at 337.5%). Drag the enrollment floor to ~10,000 and nearly every extreme ratio disappears: they're small-contract artifacts, not business models.
Try it: set the enrollment floor to 10,000. The 337% MLR vanishes, the 75% MLRs mostly vanish, and what's left is the actual industry: a tight band between 85 and 100, hugging the floor from above. Plans manage to the line — precisely, and legally. That precision is the tell that the number is managed, not emergent.
The built-in blind spot
What this ledger structurally cannot see
A denied claim is invisible here. The MLR numerator counts claims paid — a denied SNF stay never enters the math. Per KFF's July brief, MA plans deny 65% of long-term-care-hospital and 54% of inpatient-rehab prior-auth requests — and a plan can do that while sitting comfortably above the 85% floor. Denial-heavy and MLR-compliant are not opposites; they can be the same strategy. That's why this week's other fight — plan-level denial and overturn disclosure — is the dataset that would actually change behavior, and why who owns that benchmark is an open question worth arguing about.
Mind the vintage and the grain. 2023 is the most recent year with full detail, and MLR filings run ~18 months behind reality. The data is contract-level, not plan-level — one contract can hide dozens of plan designs. And ratios from tiny contracts are statistical noise: the 18 contracts filing above 115% have a median enrollment of ~1,700. The slider above is the honesty control.
The 80/20: the 85% floor returned $52.4M on $537.8B — 0.01%. The enforcement power of the MLR was never the remittance check; it's the disclosure itself. If the MLR Transparency Act survives the Senate, the buildable thing is the layer that joins this file to denial and overturn rates — the Bloomberg terminal of the denial economy. This page is v0.1 of one column of it.