A consolidated statement is a weighted average of everything a company does. Averages tell the truth only when the things averaged are alike, and a conglomerate's aren't. So the consolidated margin may describe no business the company actually runs. You need to be able to demonstrate two consequences cold:
Consequence 1: the average is a fiction. Take ITC as it looked before the hotels demerger completed in January 2025 (illustrative, modeled on ~FY24 scale). Pull the actual segment note from the FY24 annual report.
| Segment (₹ crore) | Revenue | Result | Margin | Segment assets | Result ÷ assets |
|---|---|---|---|---|---|
| Cigarettes | 31,000 | 19,500 | 62.9% | 8,000 | 244% |
| FMCG — Others | 21,000 | 1,700 | 8.1% | 12,500 | 13.6% |
| Hotels | 3,000 | 650 | 21.7% | 7,500 | 8.7% |
| Agri business | 16,000 | 1,050 | 6.6% | 5,000 | 21.0% |
| Paperboards & packaging | 8,500 | 1,250 | 14.7% | 9,500 | 13.2% |
Why can a consolidated margin describe no real business?
The blended segment margin is ~30% on gross segment revenue, a margin that not one of the five businesses earns. The capital story is starker. Cigarettes hold ~19% of segment assets and produce ~81% of segment results, at a ~244% pre-tax return. Hotels held nearly as large an asset base as cigarettes (₹7,500 cr against ₹8,000 cr) and produced 3% of the results at ~9% pre-tax. That's below the ~16% pre-tax hurdle. So one company held one value-creation machine of historic quality, one value-consuming segment (since demerged) and three in-between businesses. For years the market argued about this very table. "ITC's ROCE" was always the wrong object of study. The segment note was the analysis.
Advertisement
Watch the eliminations when you pull the real note. Agri sells leaf to cigarettes and paperboards sells board to FMCG, so gross segment revenue exceeds external revenue by the inter-segment sales. Use external revenue for company-level ratios and gross segment revenue for segment margins.
Amazon teaches the same lesson in the US (illustrative, modeled on ~FY24 scale, so pull the 10-K segment footnote). North America did revenue of ~$387bn and operating income of ~$25bn (6.5%). International did ~$143bn and ~$4bn (2.8%). AWS did ~$108bn and ~$40bn (36.9%). AWS is ~17% of revenue and ~58% of operating income. The consolidated ~10.8% margin describes neither the ~4 to 6% retailer nor the ~37% cloud utility. Any Amazon multiple, margin trend or peer comparison run on consolidated numbers is analyzing a company that doesn't exist. You have to value the retailer against retailers and the cloud against clouds. That's sum-of-the-parts, which the valuation phase formalizes, and the segment note is the only public door into it.
What is the rate vs mix effect (Simpson's paradox)?
Consequence 2: mix shift, where the blend moves against every part (Simpson's paradox). Here's a synthetic demonstration. Work it by hand:
| Year 1 | Year 2 | |
|---|---|---|
| Segment A revenue / margin | 600 / 20.0% | 660 / 21.0% (improved) |
| Segment B revenue / margin | 400 / 8.0% | 640 / 9.0% (improved) |
| Profit | 120 + 32 = 152 | 138.6 + 57.6 = 196.2 |
| Consolidated margin | 152 ÷ 1,000 = 15.2% | 196.2 ÷ 1,300 = 15.1% (fell) |
Both segments improved, and the consolidated margin still fell. The low-margin segment grew 60% while the high-margin one grew 10%, so the weights moved. Take any "margin compression" headline about a company with fast-growing lower-margin segments: every IT company scaling BPO, every FMCG company scaling foods, Amazon whenever retail outgrows AWS. Decompose it into rate vs mix before you conclude anything about execution. It works in reverse, too. Consolidated margin expansion can hide every segment getting worse, if mix runs the other way. The consolidated number is a weighted average wearing a trench coat. Open the coat.
How do you read a segment note?
Working the segment note (Ind AS 108 / IFRS 8 / ASC 280, one architecture, all three frameworks). Segments follow the management approach. They're whatever the Chief Operating Decision Maker (CODM) reviews to allocate resources. That's why they map to real internal economics, and it's also why management controls the knife. A segment is reportable if it's ≥10% of revenue, of profit or loss, or of assets, and segments keep being reported until they cover ≥75% of external revenue. For each segment you get revenue (external plus inter-segment) and a segment result. The result is measured as the CODM measures it: often EBIT-ish, sometimes before unallocated costs, never guaranteed comparable across companies. Read its definition note first, every time. You usually get segment assets too, and sometimes segment liabilities, capex and D&A as well. Then come an unallocated column (corporate costs, treasury, tax), a reconciliation to the consolidated statements, and the entity-wide disclosures. Those cover revenue by geography and any customer ≥10% of revenue, the concentration flag that forensic work turns into a weapon.
How do you rebuild segment margins and returns?
Here's the rebuild protocol. (1) Read the segment-result definition. (2) Compute segment margin = result ÷ segment revenue. (3) Compute segment return ≈ result ÷ segment assets, or net assets if segment liabilities are given. It's pre-tax, so judge it against the pre-tax hurdle. (4) Decide how to treat the unallocated items. You can allocate them pro-rata to revenue (crude) or leave them unallocated and judge segments before corporate cost, which is cleaner as long as you say so. (5) Sanity-tie segment totals to the consolidated statements through the reconciliation. Respect the limits, too. Management sets transfer pricing between segments. Segment "assets" allocations can flatter a favored segment. And management redraws segment boundaries. A redefinition that conveniently buries a deteriorating segment inside a healthy one, right as it sours, is a first-order red flag. Restated comparatives are required, and they're your paper trail. Keep that in mind when the forensics weeks arrive.