UNH

The band is the interquartile range of every estimator that could run on UNH’s filings: the median P/E, P/S, P/B and P/FCF this stock has actually traded at over 19years of history applied to today’s fundamentals, a two-stage discounted-cash-flow estimate. With four or more estimators we take the interquartile range, so a single wild reading cannot define the band; with two or three we show the full spread, because trimming the tails off three points does not remove noise, it manufactures confidence. We publish a range rather than one number because the estimators genuinely disagree, and averaging that disagreement away would hide the most useful part.

Valuation labels describe only where the price sits against that band:

• Significant Discount — price is more than 30% below the bottom of the band.
• Discount — price is below the bottom of the band.
• Fair Value — price sits inside the band, i.e. about where this company’s own fundamentals put it.
• Premium — price is above the top of the band.
• Significant Premium — price is more than 30% above the top of the band.

These labels are positional, not predictive. Premium means the current share price sits above the band computed from UNH’s own filings and trading history. It does not mean the stock cannot keep rising, that a fall is expected, or that the market has it wrong. Equally, Discount means the price sits below that band — it does not imply the stock is cheap in any meaningful sense, that it is worth buying, or that the market has mispriced it. A company can trade above or below its own historical range for years, and often does, for reasons the range knows nothing about.

The labels describe where today’s price falls against a calculated band. Nothing more than that.

Measured disagreement is our answer to a locked “uncertainty rating”: it is simply how far the estimators spread as a share of the midpoint, tiered Low / Moderate / High / Extreme, with the raw spread shown beside it. A wide band earns a high tier because the answer then depends heavily on which yardstick you pick — that is a finding, not a defect.

Filing health is computed separately and never looks at price. It reads the Piotroski F-Score, the Altman Z-Score and the Beneish M-Score straight from the filings: means none of them flagged, means one moderate flag, means a distress or manipulation flag, or several softer ones. For banks and insurers it reads : all three scores were built on non-financial companies and their authors excluded financials, so for a balance-sheet business they measure the shape of the business rather than its health. Keeping the two axes apart is deliberate — a cheap company with weak filings is a very different object from a cheap company with sound ones, and a single blended verdict would hide which you are looking at.

The score is passed over computable, not passed out of five. If only three tests could be run on UNH’s filings, the score is out of three and the other two are listed with the reason they were skipped. Scoring a skipped test as a failure would punish a company for how its industry reports, and scoring it as a pass would flatter one.

Labels come from that ratio:

• Well Covered — 90% or more of the computable tests passed.
• Covered — 65% or more.
• Thin — 40% or more.
• Strained — below 40%.

Below three computable tests we publish no label at all and show the individual results instead. A summary drawn from one or two tests is not a summary.

Free cash flow coverage carries the most weight in practice, because dividends are paid in cash and free cash flow is the cash actually available to pay them. The earnings payout ratio is the better-known test but the weaker one — earnings are an accounting figure, dividends are not. Where the two disagree, believe the cash.

Why tests get skipped.A REIT gets no earnings payout ratio: GAAP depreciation on appreciating property understates a REIT’s distributable earnings, which is exactly why the industry reports FFO and why REITs are required to distribute most of their taxable income. Banks and insurers get no leverage test, because deposits are debt and structurally high leverage is what a balance-sheet business is, not a warning sign.

The biggest limitation, stated plainly. The coverage tests read a single fiscal year, and one year of cash flow is far more volatile than a dividend record. A legal settlement, a tax payment or an acquisition can sink coverage in a year that says nothing about the dividend. Where a weak year contradicts a long unbroken record, we say so above rather than letting the harsher number stand alone — but you should still check what happened in that specific year.

The streak is the weakest test here, and it is deliberately listed last. Every dividend that was ever cut had an unbroken streak right up until the year it was cut. A freeze does not break the run — a company that holds its dividend flat through a bad year has not cut it — so raises are counted separately, and a long flat run cannot pass itself off as dividend growth.