Financial Ratio Analysis: Three Questions per Number

The operating margin is 17%. So what?

On its own that number answers nothing, because you do not yet know whether it is high, which way it is moving, or against whom you are measuring it. Financial ratio analysis begins exactly here, at the second number you set beside the first. Before that you hold a reading, not a conclusion.

The sequence that turns a reading into information is short: level, trend, comparison. Then a fourth question most guides never ask, and the one that decides most often: what is happening to the growth rate itself. Four moves in a fixed order, because the third only earns its cost once you have settled whether this company deserves an hour at all.

Start with four numbers, not twenty#

A first pass does not need twenty ratios. It needs four numbers, because those four cover the whole chain: how much the company sells, how much of that it keeps, how much of it actually lands in the bank, and what the market expects next. Revenue, operating margin, cash from operations, and the two-year forecast.

That set is not a matter of taste. Each of the four comes from a different place and each can contradict the other three. Revenue and margin sit in the income statement, cash in the cash flow statement, and the forecast comes not from the company but from the people who watch it. Four independent sources are four independent chances to notice that something does not add up.

Suppose a company where the four point four different ways. Revenue grows 9% year over year. The operating margin holds at 17%, exactly where it stood a year ago. Cash from operations falls 4%. The two-year forecast calls for 6% a year, slower than the last twelve months delivered.

Four numbers, four directions, one company. Which reading is the true one? All of them, and that is the whole difficulty.

The third line is the interesting one here, because falling cash against a flat margin says something jammed between the invoice and the payment. That deserves its own treatment and has a separate piece on how profit differs from money in the account. Here one rule is enough: collect the four together, because their disagreement is the information.

A figure without a denominator settles nothing#

Is 130 million of debt a lot? There is no answer, and that is not evasion. The same amount is a footnote against 4 billion of assets and a death sentence against 200 million, and the difference appears only when both numbers stand side by side.

A ratio needs two data series because its job is to cancel scale. Without that you cannot set a company selling a billion beside a company selling a hundred million, which is the entire reason anyone reaches for ratios rather than amounts. That is why denominators usually come from the statement that holds assets and liabilities, while numerators come from the income statement or from the cash flow statement, where the movement of money is split across three separate columns.

Out of that comes the rule that saves the most time at the start. Before asking whether a number is good, check whether it is a ratio at all. With no denominator none of the three questions below has anything to grip, and the answer you invent anyway will have come from context rather than from a document.

When the reading speaks for itself#

The first question is the cheapest one to ask, which is exactly why it gets asked badly. Is this value high on its own?

Its underrated virtue is immunity. Whatever the rest of the market is doing, the plain reading keeps reporting the same thing, whereas comparison flatters a company the moment its neighbours fall into worse shape than it is. An industry in a downturn produces a page full of relative winners and not one good business.

The catch is that the plain reading only speaks at the extremes. Operating profit at 40% of revenue is high in any industry and needs no second opinion. Two percent is low just as uncontroversially. Thirteen percent says nothing, and it will keep saying nothing until three other companies stand beside it, which is to say until you know where the industry's margin ceiling sits and how much room this company keeps under it.

So the entry condition for step three writes itself, and I adopted it years ago to stop losing afternoons: reach for comparison only once the plain reading is ambiguous. At the extremes, lining a company up against its peers costs half an hour and changes no conclusion.

One limit is worth knowing before you lean on any of this. The plain reading does not transfer between ratios: a 40% margin is high everywhere, but debt at 40% of assets is neither high nor low until you know what services it. Balance-sheet ratios almost never read plainly, which is why lists of ideal thresholds fail on the first company from an industry financed by debt by design.

Direction outranks position over two years#

The second question matters more than the first, and that is the argument of this piece.

Two companies. The first runs a 10% operating margin, up from 7% over three years. The second runs 15% and is coming down from 19%. On the point, the second wins. On direction, the first is winning what the second is giving away, and an investor is buying the coming years rather than the last quarter.

None of which means trend always beats level. It means that over two or three years direction decides more often than the point, because processes inside companies are slow. If customers extracted a discount, they will extract it again next year, unless the product or the market structure changes, and such changes take quarters rather than weeks.

A margin rarely stops falling because somebody lost interest.

The same observation produces a rule that tidies up the whole reading of leverage. Debt at 35% of assets is good news in a company coming down from 60% and bad news in a company on its way up from 10%. The first is paying down, the second is accelerating, and one number is describing two opposite stories.

The symmetry therefore runs between different levels rather than within one. A company at 31% of assets and falling earns the same verdict as a company at 25% and rising, even though the second looks better in the table.

I read trend over five years, with a second pass over ten. A shorter window shows the last cycle instead of the business, and series longer than a decade describe a company that earned its living from something else back then.

One more thing shows up only in a series and never in two points: smoothness. Five years of steady 8% growth says something quite different from five years containing two jumps of 25% and two declines, even when both series end in the same place. The first company is predictable, the second depends on something you have not yet named, and that something will come back.

The growth rate has a growth rate of its own#

The fourth question is missing from the textbooks and I consider that their largest gap. It runs: is the growth rate itself falling?

Start from what it costs, because that part is concrete. A price multiple built on 20% and then fed 11% unwinds faster than almost anybody expects, and in a company of real size the distance between those two readings is several years of profit.

Now the series that produces it. Revenue rising 20%, then 16%, then 13%, and 11% in the latest year is positive in all four readings and climbing throughout, and every year still delivers less than the one before it. A five-year total will look excellent, because it compares the end against the beginning and never notices the shape in between.

Where the braking comes from, and whether it reverses, is settled by taking apart what revenue growth is actually made of: price, volume, an acquisition, or a currency. In a large company a fading rate is unavoidable anyway, because the base grows faster than the market, so on its own it disqualifies nothing. It still has to be measured and named before anyone sits down to a valuation.

The symmetry works here too. A growth rate that stopped falling and has held flat for two years is better news than a higher rate still braking, because the first company has found its floor and the second does not yet know where the floor is.

Peers, own history, or the whole market#

The third question comes last because it is the most expensive. Against whom?

The first answer is free and most people skip it: against the same company five years ago. Its own history removes the argument about comparability entirely, since the business model, the accounting standard and the cost structure are identical on both sides. Only when its own history fails to settle the matter do I reach for somebody else's numbers.

Most data services answer with a list of similar firms, and those automatic lists mix businesses of wildly different size: a global manufacturer stands next to a regional distributor because they share an industry code. Two or three companies chosen by hand make better background than twenty chosen by an algorithm.

The second principle is less obvious and it took me years to accept. The goal is not the best company in an industry but a good business anywhere in the market, because the strongest player in a declining industry usually loses to an average player in a growing one. Comparison inside an industry tells you who manages better. It does not tell you whether that industry is worth searching at all.

That leaves background wider than the industry, and there a number with a date beats an impression. The analyst forecast round-up for the broad US index published on 4 September 2026 puts expected revenue growth for 2026 at 12.0%. So our company at 9% is growing slower than the market, even though its curve points upward and looks healthy.

Background like that ages with the market, so every use of it needs a fresh reading and a date. A ratio is compared against something that is itself moving, which is why nobody can write down a permanent table of what counts as good revenue growth.

Where financial ratio analysis breaks#

Three situations look identical on a chart, and in each of them the number means something other than it suggests.

The first is a negative or near-zero base. Earnings per share go from 1 to 2, growth reads 100%, which is arithmetically true and informationally empty. An algorithm awards full marks, a person glances at the absolute value and moves on. The same mechanism runs in reverse: a company climbing out of a loss produces growth rates that cannot be written down sensibly, because dividing by a negative number flips the sign of the result.

The second is a change of fiscal year end. Move the point where a year stops and a step appears on the chart, and that step looks like an event while it is a change in the definition of the period. A fiscal year need not match the calendar, and in some industries it systematically does not, so before comparing two companies it is worth checking that both are measuring the same slice of time.

The third is the sneakiest, because it involves forecasts. A reported figure includes one-off items and analyst forecasts usually exclude them, so side by side the forecast sits below the reported number for no reason. Somebody then reads that as a warning of decline and sells something that never fell.

The cure for all three is the same and it is dull: look at the shape of the series, not at the last number. A single reading can be explained ten ways. A ten-year curve narrows that list sharply, and where it does not, something genuinely unusual is happening and the evening is worth spending on it.

The grid that replaces the score#

Four moves do not add up to a score, and they were never meant to. What they produce is a coordinate, with one axis for where the company stands and one for where it is going.

Read the grid as a list of next questions rather than as nine verdicts. That is its entire job.

Eight of those cells can be settled without ever leaving one company's own filings. The ninth, the middle one with an average level and a flat trend, carries no information whatsoever, and it is the only place where peers genuinely change the conclusion. It is also the cell where I stop most often.

The question the grid cannot answer#

Four moves organise the reading and say nothing about why the numbers look the way they do. A company whose margin rose because it sold off a loss-making segment shows exactly the same pattern as a company that raised prices and kept its customers. The first did it once, the second can repeat it, and the grid cannot see that difference.

I think this is the boundary of all work on ratios, and no larger pile of ratios moves it. What moves it is a document: the segment note, the management commentary, the risk section. Numbers say where to look. The reason always sits somewhere else.

I do not know whether a company in the middle-and-rising cell will hold that direction for two years, and neither does anyone else. I do know that a company in the high-and-falling cell rarely stops on its own, and there is one place to check that: whether the forecasts are coming down along with the margin, or not yet.

Forty minutes and a sheet of paper#

Take one company you own and write four numbers on a sheet of paper: revenue growth, operating margin, cash from operations, and the two-year forecast. Beside each write three answers: high or not, which way, against whom. Then add the fourth, about the pace. That takes forty minutes and beats an hour spent reading other people's commentary.

The statements are free, incidentally. For US companies the series sit in the 10-K and 10-Q filings in the EDGAR full-text search, next to the regulator's own guide to reading financial statements.

The same work then repeats for every next company, which is precisely why it pays to automate it. At Taufolio that is the job of the Full report: it pulls the series straight from the filings, arranges them into these same four moves, and leaves a link next to every number pointing at the document it came from. What that looks like on a real company is in the sample reports.

And if you have time for only one of the four numbers, take cash from operations.

It is the one that least often says what management would like to hear.