The numbers we refuse to print, and why that makes the rest worth trusting

A stock-comparison table with no blanks looks complete. It is also, almost always, wrong somewhere. Some figures cannot be calculated honestly for some companies, and a data provider that fills every cell anyway is printing numbers that look like facts but mean nothing. TickerVerdict takes the opposite approach: when a number cannot be interpreted for a particular business, we leave it out, show a dash, and say why. This guide lists every kind of figure we withhold, with the real companies it applies to today.

Forward P/E: never shown, for anyone

A forward P/E divides today's price by next year's expected earnings. The expectation comes from analysts' forecasts, and forecasts are not in any company filing. Every figure we publish is either filed with the SEC or a dated exchange closing price, so we have no honest source for the denominator. Rather than borrow estimates we could not show you or check, we leave forward P/E out of every page — and the scoring rules simply skip it, so its absence cannot tilt a verdict. How the rules handle that is explained in how the six-factor score works.

A P/E for a company that lost money

Price divided by a negative number gives a negative P/E. It is arithmetic, not information: a "P/E of minus 8" does not mean a stock is cheap, and a company losing very little money produces an enormous negative multiple that means even less. So when trailing earnings per share are below zero we show no P/E at all. At the 14 October 2026 close that applied to:

For these companies the useful questions are different: how large the losses are relative to revenue, how much cash is left, and whether the trend is improving. Our EV and auto stocks study works through exactly that for Ford, Rivian and Lucid.

Growth measured across a loss

A growth rate compares a starting value with an ending one. If the starting value is a loss, the result is meaningless: going from a $1 loss to a $1 profit is not "200% growth", it is a turnaround. Before we stopped this, an old data source once showed Ford's earnings growing at minus 239% a year. Today our filings layer only computes a growth rate between two positive values, so these companies show no three-year EPS growth:

• Amazon, Palantir, Shopify and Uber each reported a net loss in 2022, the base year of the three-year window.
• Micron lost money in its 2023 fiscal year, its base year.
• Ford also lost money in 2022, and Intel reported a loss in its most recent fiscal year, the other end of the window.

Revenue growth is unaffected, because revenue is never negative. That is why the growth factor is not scored when EPS growth is missing: half of its evidence is gone.

Growth across a stock split

Per-share figures change meaning when a company splits its shares. Netflix's trailing EPS is $3.18 on today's share count; its EPS from three years ago was reported on the share count before its ten-for-one split. Divide one by the other and you get a collapse in "earnings per share" that never happened. When a split falls inside the window and the older filings have not been restated, we withhold the per-share growth rate rather than adjust it ourselves. Netflix therefore shows no three-year EPS growth, while its revenue growth is shown as normal.

Metrics that mean nothing for a bank

For an industrial company, borrowing is a choice and debt is a risk. For a bank, borrowing — chiefly deposits — is the raw material of the business. So the ratios built for industrial balance sheets stop making sense:

• EV/EBITDA — a bank has no meaningful EBITDA, and its enterprise value would count customer deposits as debt.
• Gross and operating margin — banks do not report cost of sales or an operating-income line.
• Net debt to EBITDA, the current ratio and interest coverage — interest is a bank's cost of goods, not a financing burden.
• Return on invested capital — its "invested capital" would be dominated by deposits.

For JPMorgan and Bank of America we show the measures that do work — price-to-book, return on equity, net margin, P/E and dividends — and how to compare bank stocks explains how to read them.

Gross margin for insurers and lenders

Gross margin is what is left of each dollar of revenue after the direct cost of what was sold. For an insurer, the main cost is the claims it pays; for a lender, it is the interest on the money it lends out. Neither sits on a cost-of-sales line, so a gross margin computed from the filings counts the revenue and leaves out the cost that matters most. UnitedHealth's came out implausibly high for exactly this reason: the product costs it reports are a small part of what it spends, and its medical claims are the bulk.

So we print no gross margin for an insurance carrier, or for any company whose filings report both net interest income and non-interest income, which covers SoFi as well as JPMorgan and Bank of America. The blank has a consequence for the score: the quality factor needs all four of its inputs for both companies, so quality is not scored for any pair that includes one of these businesses. The "not scored" label says the factor could not be measured, not that the two businesses are equally good.

When the denominator is negative or tiny

Some ratios break when the number underneath them turns negative or shrinks to almost nothing:

• Negative shareholders' equity. Lucid ended the period with equity of −$1.06bn, so a price-to-book would be a negative number with no sensible reading, and it is blank. Its return on equity, measured against the average of its opening and closing equity, came to roughly −570% — outside our sanity limit — so that is blank too.
• Invested capital that is a sliver of the business. If equity plus debt minus cash is less than a twentieth of total assets, return on invested capital measures the sliver, not the company, so we withhold it.
• Negative EBITDA. EV/EBITDA is shown only when EBITDA is positive.

Impossible values, and wrong ones

A final safety net blanks any figure that no listed company could actually have: a margin above 100%, a dividend yield above 100%, a growth rate beyond 500% a year. These are almost always tagging errors in the source data, and the net is automatic but blunt: it catches the impossible, not the merely wrong. A figure can look plausible and still be wrong, so we also read the records themselves, and when we find one we take the whole comparison down by hand until the data is fixed. At the time of writing six comparisons are down for that reason: two where an oil major's long-term debt appears only in its annual report and was missed, two where a health insurer's "gross margin" leaves out medical claims, one where a lender's revenue is not read the way it reports it, and one where a payments network's annual dividend is tagged with a single quarter's figure. They return when they are right.

Numbers we do print, with a warning

Some figures are correct but easy to misread, so they appear with context rather than not at all:

• Very large growth rates. Salesforce's three-year EPS growth is 233.65% a year — genuine, but measured from an unusually small base. Our guide to revenue versus EPS growth explains how to read it.
• Net income above operating income. Alphabet's net income of $244.21bn over the past twelve months exceeds its operating income of $147.63bn, and Amazon's $135.28bn exceeds its $93.71bn. When that happens, a large part of profit came from outside the operating business — gains on investments, for example — and should not be valued as if it will repeat. Ratios built on net income, such as P/E and net margin, need that caveat. Operating income is the steadier base; see the Alphabet vs Amazon comparison.

Why blanks build trust

A dash in our table is not a gap in our data. It is a decision, made by a fixed rule, that a particular number would mislead you about a particular company. Every rule here applies to every company in the same way and is written into the code that builds the pages. The one thing we do by hand is take a page down when we find a figure on it that is plausible but wrong — and it stays down, with the reason recorded, until it is fixed. If you find a figure that you think should have been withheld, or one that should not have been, tell us. The full sourcing rules are in our editorial policy.