One Ratio, Three Stories: The DuPont Decomposition
Knowledge Bank · Shelf: Financial Analytics
Return on equity is net income divided by shareholders' equity. One ratio, one number. It is the figure people tend to quote when they want to say a business is doing well, and it is the figure that most reliably conceals why.
The DuPont decomposition rewrites that single ratio as a product of three: net margin (net income / revenue) × asset turnover (revenue / total assets) × leverage (total assets / equity). The identity is exact — the three legs multiply back to ROE. In our quant engine (ra-quant) this identity is not assumed; it is asserted by an invariant test that must hold on any data. That is a statement about arithmetic, not about insight, and we will come back to the difference.
We are drawing here on two documents: the house's "RA Analytical Engine: Master Plan" (implementation record, 26 September 2026), and the owner's study index, "Frameworks & Concepts Index," whose ratio-analysis and DuPont entries come from his financial-accounting coursework — measuring a business from its statements rather than from its headlines.
What the engine computes
The engine computes the decomposition per period — margin, turnover, and leverage, with ROE, for each period in a statement series. It also computes an attribution: how much of a change in ROE between two periods came from each leg, expressed in percentage points. Three legs, three contributions, one arithmetic sum.
The data discipline is the same as elsewhere in the house. Insufficient data returns null rather than a filled number, and a period missing any leg yields no ROE for that period. We would rather show a gap than a guess.
Story one: margin
Net margin asks how much of each sales dollar survives to net income. It is the leg people intuitively trust, because cost discipline is a real thing and margin is where it shows up.
But a rising margin has a second reading. It can mean a shrinking business whose remaining sales are its most profitable ones — the low-margin lines gone, the good ones left, and the ratio improving as the enterprise narrows. The leg alone cannot say which story is true. Both produce the same movement in the same number.
Story two: turnover
Asset turnover asks how much revenue each dollar of assets produces. Rising turnover reads naturally as efficiency: the same asset base, more sales.
It can also mean under-investment — the asset base aging while sales hold for a while. For a period, that looks like improvement. The turnover leg cannot distinguish sweating assets harder from letting them wear out.
Story three: leverage
Leverage asks how many dollars of assets stand on each dollar of equity. It is the leg that behaves differently from the other two.
Hold margin and turnover constant, add debt, and ROE rises. Nothing operational has improved. Leverage lifts the headline arithmetically. It is the only leg that can raise the number while making the company more fragile — and it is the leg a single ROE figure hides completely. A reader looking only at ROE sees the result of leverage without seeing leverage.
What a decomposition can and cannot say
The decomposition reorganizes; it does not explain causes. Attribution is arithmetic — it says which leg moved, not why it moved. If margin contributed two percentage points of a change in ROE, the arithmetic is settled; the reason is not. Reading the "why" requires the statements behind the legs: the revenue lines, the cost structure, the asset register, the financing.
This is the honest use of the tool. It tells you where to look. It does not tell you what you will find, and it does not tell you what to do with what you find.
A worked shape, without numbers
Suppose ROE rises between two periods. The attribution splits the rise across the three legs. If margin carried most of it, the questions are about revenue mix and cost. If turnover carried it, the questions are about asset intensity and whether investment kept pace. If leverage carried it, the questions are about the balance sheet — and about whether the improvement in the headline is an improvement in the business at all.
Each of those is a question, not an answer. The engine produces the split; the reader produces the reading.
The limits, stated plainly
The identity is arithmetic, not causal. The three legs multiply to ROE exactly, and that exactness says nothing about why any leg moved.
Three legs are a lens, not the whole statement. Tax, one-off items, and buybacks move equity and net income in ways the three legs compress. What the decomposition shows is a reorganization of the ratio, not a full account of the business.
A shrinking equity base can lift ROE through the leverage leg without leverage in the borrowing sense changing. Large buybacks are the obvious case: equity falls, the assets-per-equity ratio rises, and the headline improves while the company has not borrowed anything. The leg is doing its arithmetic correctly; the label is doing more work than the arithmetic supports.
Period-end balances are snapshots and can differ from averages. The decomposition as computed uses the balances in the statement series, and a period-end figure is a point in time, not a typical level.
On status: the DuPont decomposition runs live in Research Analyst today, in the product's own reports. The function-level detail here is drawn from ra-quant, the house's standalone package of the same machinery — test-verified on its development branch; the package itself is what remains unshipped. The full two-part status is the subject of the companion piece, "What a Quant Engine Actually Computes."
We publish what we produce. Here that is a decomposition, an invariant test, and an attribution — pointers, not conclusions. The three legs are a way of asking better questions of a set of statements. They are not a verdict on a business, and nothing in this piece is one.