Profit margins

Picture ₹100 of revenue entering a toll road. At the first gate it pays for the product itself: materials, factory labor, freight. At the second gate it pays for the organization: salespeople, advertising, head office, R&D. At the third it pays for the machines wearing out. At the fourth it pays the lenders and the government. Whatever exits the road belongs to shareholders. The margin stack is simply the reading at each tollgate:

What does each gap between margins tell you?

The layers matter because different costs live between them, and each inter-layer gap is a separate diagnosis:

• Between revenue and gross profit lives COGS, the cost of the thing sold. A falling gross margin means input costs are rising faster than prices, or mix is shifting to worse products, or discounting has begun. This is the pricing power layer. Moats I will show you that durably high, stable gross margins are one of the cleanest numerical fingerprints of a moat.
• Between gross and EBITDA live the operating expenses excluding D&A: selling and distribution, advertising, G&A, R&D. This is the organization layer. If revenue grows 15% and these costs grow 15%, the model has no operating leverage. If they grow 8%, scale is doing real work.
• Between EBITDA and EBIT lives depreciation and amortization, the accounting charge for capital consumed. The wider this wedge, the more capital-hungry the model. A software firm loses 1 to 2 points of margin here. A telecom or steel plant loses 10 to 20.
• Between EBIT and net income live interest, other income, taxes and one-offs. These are the financing and state layers, plus whatever exceptional noise the year contained. Nothing here is about the business model. It's all about the balance sheet, the tax code and luck, so experts diagnose at the operating layers and only summarize at net.

Which EBIT should you use?

There are two EBITs, and you need to settle which one you're using before the algebra starts. Practitioners build EBIT two ways that differ by one line:

• Bottom-up (operating EBIT): Revenue − COGS − operating expenses − D&A. It excludes other income. Use this one for operating margin, because interest earned on a cash pile isn't the business model.
• Top-down: EBIT = Profit before tax + Interest expense. It sweeps other income in automatically. The 5-step DuPont uses this construction, because only then does the algebra multiply out with nothing left over.

The difference between the two is other income. For Hindustan Unilever below, operating EBIT is ₹13,200 cr and top-down EBIT is ₹14,200 cr. The ₹1,000 cr gap is interest and gains on the cash pile. Neither number is wrong. Using them interchangeably without noticing is, so state your construction every time.

How do the reporting standards present costs?

Where the numbers come from, the presentation split (economics once, standards side by side). The economics of the stack are universal. The three reporting regimes just hand you different pieces of it.

• US GAAP (function-of-expense): the P&L is organized by what the cost was for, such as cost of sales, SG&A and R&D. Gross profit is on the face or one subtraction away. But D&A is embedded inside COGS and SG&A and often isn't shown on the P&L at all, so pull it from the cash-flow statement, the way you did when linking the three statements. Read the COGS definition footnote too. Several large US retailers present "cost of sales" excluding depreciation or occupancy shown elsewhere, so two retailers' gross margins can differ by points on presentation alone.
• Ind AS / Schedule III (nature-of-expense): Indian P&Ls are organized by what the cost was: cost of materials consumed, purchases of stock-in-trade, changes in inventories, employee benefits, finance costs, depreciation and amortisation, other expenses. There's no gross-profit line, so you assemble COGS yourself: COGS = cost of materials consumed + purchases of stock-in-trade + change in inventories of FG/WIP/stock-in-trade. The change line is positive when inventory fell, because costs incurred earlier are charged now, and negative when it rose. The consolation prize is that D&A and finance costs appear on the face, so EBITDA is one subtraction away. That's the mirror image of the US situation.
• IFRS: IAS 1 permits either presentation. A filer presenting by function must disclose D&A and employee costs in the notes.

What should you watch for with EBITDA?

One India-specific trap is worth disarming now. Many Indian data vendors and some managements quote "EBITDA" including other income, because the Schedule III P&L presents "Total income" (revenue from operations + other income) at the top. Always compute EBITDA off revenue from operations only, and strip other income. A treasury desk earning 7% on idle cash isn't operating profit. When a screener's EBITDA and yours disagree, look here first.

EBITDA is good for three jobs. It compares operating engines across different capital structures and asset vintages. It supplies the credit market's covenant yardstick, net debt/EBITDA, which you'll meet properly in Ratio Toolkit II. And it anchors the EV/EBITDA multiple in valuation. It isn't cash flow, though. Depreciation is a real cost that comes back as capex, working capital eats cash EBITDA never sees, and taxes and interest must actually be paid. Buffett's standing question in the Berkshire letters (does management think the tooth fairy pays for capital expenditures?) and Munger's ruder suggestion for what the acronym stands for are the correct level of suspicion. Carry the lease trap forward as well. An Ind AS 116 / IFRS 16 reporter shows no rent expense, so its EBITDA is mechanically higher than a US GAAP twin whose operating-lease rent is booked in opex. Never compare EBITDA margins across that border without the repair. Practice problem 8 works it, and Ratio Toolkit II carries it through in full.

Margin signatures: reading the business model off the stack

Set four illustrative stacks side by side. All figures are illustrative, modeled on ~FY24/FY25 scale, so pull the actual filings (₹ cr for India, $bn for US):

*Infosys presents by nature. Its "gross" layer here treats direct delivery costs (mostly engineer salaries) as cost of services, a reminder that for people businesses the gross line is a modeling choice.

Read the signatures. HUL loses only ~30 points of margin between gross and net. Brands mean price exceeds input cost enormously (53% gross), and the toll road after that is short, because an asset-light FMCG pays modest D&A and almost no interest. Walmart starts at 24% gross, because retail buys finished goods and resells them, and it delivers 2.9% to the bottom line. The whole model runs inside a 21-point corridor of cost discipline. Give away one point of gross margin and you've given away a third of net profit. Infosys has almost no material costs (its "COGS" is salaries), so the informative layers are EBITDA and EBIT, and the gap to net is mostly tax. The steel maker's stack is meaningless as a point estimate. Its net margin can double at the cycle's top and go negative at the bottom, so read every layer against mid-cycle norms. Benchmarking and the commodities playbook both come back to those norms.

What do margin trends tell you?

Then read the trends. They matter more than the levels.

• Gross margin drifting down, year after year, means pricing power eroding, input inflation unpassed, or mix degrading. For an FMCG this is strategic bad news even when net margin is held. Ask what was cut to hold it. It's usually advertising, which mortgages the brand.
• An opex ratio (opex ÷ revenue) falling as revenue grows means operating leverage is working and the organization scales.
• A widening wedge between EBITDA and EBIT means capital intensity is creeping up. Check capex against depreciation, and expect to come back to it when you test whether profits are backed by cash.
• Net margin rising while operating margin is flat means the improvement came from below the line: cheaper debt, a tax break, or other income. It's real, but it isn't the business model. The 5-step DuPont exists to catch this.