Playbook › Playbook
Net margin above operating margin means the profit came from below the operating line
Claim
The check costs one subtraction and it fired twice in a single nine-name sweep.
| ticker | op margin | net margin | trailing P/E | forward P/E | what was actually happening |
|---|---|---|---|---|---|
| AMSC | 5% | 45% | 10.97 | 23.81 | Deferred-tax valuation allowance release |
| UMC | 22% | 33% | 20.23 | 23.49 | Non-operating income — FX and investment gains |
| FIS | 21% | 28% | 6.58 | 6.38 | Disposal gain — sale of the remaining 45% Worldpay stake |
Case 4 confirmed — the disposal gain (FIS, 2026-08-06)
The taxonomy's fourth cause was theoretical until FIS. Trailing P/E showed 6.58x. The real figure is 19.3x — the error is 3x, the largest yet recorded under this pattern.
One row of the Q1 2026 income statement carries it: pretax income $259M, "earnings from equity interest, net of tax" +$2,214M, net income $2,366M, diluted EPS $4.58. The gain came from selling the remaining 45% Worldpay stake to Global Payments on 2026-01-09. Rebuilding TTM EPS quarter by quarter — $0.50 + $0.98 + $4.58 + $0.45 — reproduces Yahoo's $6.51 exactly, confirming the statement data is sound and only the ratio misleads. Stripping the gain gives core TTM EPS of ~$2.22.
Two refinements the FIS case adds:
- The forward-P/E confirmer can fail. FIS showed forward below trailing (6.38 vs 6.58), which under the rule above reads as "clean." It was not — the vendor's forward EPS of $6.71 was an adjusted figure from a later fiscal year, so two independent errors cancelled and hid the first one. When the tripwire fires, the forward P/E cannot clear it. Find the non-operating line.
- Check the balance sheet alongside. A disposal gain of this size leaves a footprint — FIS's "long-term equity investment" line fell from $3.68B to $13M between FY2025 and Q1 2026. That is a second, independent confirmation available in thirty seconds.
AMSC is the severe case: a 9x gap. It screens as a profitable company on 11x earnings. It is a company earning a 5% operating margin, priced at 23.8x forward earnings and 53.6x EV/EBITDA with a beta of 3.28. Every value screen keyed on trailing P/E will surface it, and every one of them is wrong.
Why it works as a screen
Below the operating line there are only four things, and none of them are the business:
- Tax — a valuation allowance release, or a one-off rate benefit ([[pitfall-tax-valuation-allowance-round-trip-breaks-eps]])
- Investment and FX gains — marks on holdings or currency translation ([[pitfall-yahoo-misclassifies-strategic-investment-gains]], [[pitfall-unrealized-equity-marks-break-headline-pe]])
- Derivative marks — warrants and converts, which can also run the other way ([[pitfall-government-warrant-derivative-inverts-gaap-eps]])
- Disposal gains — a segment or asset sale
- An expiring contract booked below the operating line — added 2026-08-20 from MRAM
- Minority-interest loss allocation under HLBV — added 2026-08-26 from RUN. The first cause that is structural and permanently recurring ([[pitfall-self-selected-discount-rate-flatters-an-asset-metric]])
Causes 1-5 are non-recurring by construction. None will be there next year, which is exactly why the forward P/E is so much higher than the trailing one in both rows above. A forward P/E materially above the trailing P/E is the confirming signal, and it is free — it says consensus has already stripped out what the screen has not.
Cause 5 — an expiring contract below the operating line (MRAM, 2026-08-20)
The first cause that is neither a mark nor a one-off event, but a real recurring cash stream with an end date. It is the most deceptive variant so far, because the other four look wrong the moment you find them and this one looks fine.
Everspin Technologies, TTM to Q2 2026: operating margin -15.6%, net margin -4.2% — an 11.3pp gap worth $7.2M, of which $5.76M was "other income". The source was US Government sustainment-contract revenue ($14.6M programme, $13.3M cumulative), booked below the operating line and winding down on a schedule:
| Quarter | Other income |
|---|---|
| Q3 2025 | $1.16M |
| Q4 2025 | $2.01M |
| Q1 2026 | $2.11M |
| Q2 2026 | $0.48M |
The sharpest illustration is Q4 2025: net income +$1.20M against an operating loss of -$1.07M. A screen reads that quarter as the company turning profitable. It was the contract running off, not the business turning.
Why this variant is worse than the other four:
- It is genuine, recurring, cash revenue — so a reader who traces it stops there, satisfied it is "real". The disqualifying property is not that it is fake but that it is scheduled to end.
- It is not a mark, so no balance-sheet footprint confirms it. The FIS refinement (check the balance sheet alongside) does not fire — nothing moves in equity investments.
- The trend, not the level, is the tell. One quarter in isolation reads as ordinary other income. Only the sequence exposes it.
Detection step this adds: when the tripwire fires and the source turns out to be ordinary operating-ish income rather than a mark, plot it by quarter and ask whether it has a contractual end. Government contracts, transition-services agreements after a divestiture, and expiring licence or royalty streams all behave this way. Read the contract value against cumulative recognised-to-date — the ratio tells you how much runway is left.
How to apply
- Run the subtraction on every screen output before ranking. If
net margin > operating margin, the row does not qualify on its trailing multiple, full stop. - Check the forward-vs-trailing P/E direction as the second gate. Forward above trailing = consensus expects EPS to fall = the trailing number contains something that is leaving.
- Re-rank on operating income or free cash flow. These sit above the contaminated line. For AMSC, EV/EBITDA of 53.6 told the truth that a P/E of 11.0 hid.
- The inverse case is equally common and equally misleading: a large charge below the operating line makes a healthy company screen as a collapse. Same check, same fix.
- Related: [[principle-down-a-lot-is-not-cheap]] · [[pattern-buyback-manufactured-eps-screens-as-growth]] · [[pitfall-gate-cleared-by-adjective-not-arithmetic]].
Cause 6 — HLBV minority-interest loss allocation (RUN, 2026-08-26)
The first cause that cannot be stripped out, because it never stops. Causes 1-5 are all one-off events you identify, remove, and move past. This one recurs every quarter for as long as the company raises tax equity, and it breaks P/E and ROE simultaneously and in opposite directions inside the same vendor feed.
Sunrun, TTM to Q2 2026: operating margin 2.66%, net margin 11.6%.
| $M | |
|---|---|
| Consolidated net income incl. NCI | −958.5 |
| Loss allocated to noncontrolling / redeemable NCI | +1,361.4 |
| Net income to common | +403.0 |
| Diluted EPS | $1.48 |
Under HLBV (Hypothetical Liquidation at Book Value), tax-equity investors are allocated ~99% of taxable losses and the ITC in each fund's early years. GAAP follows that allocation: the loss lands on the NCI line and the residual left for common turns positive. Consolidated pretax income was negative in all four quarters. Retained earnings are −$3,547M — the "earnings" have never accreted to book value.
The feed contradicts itself, which is the detection step:
| Vendor field | Value | Built on |
|---|---|---|
| EPS(ttm) / P/E 6.10 | $1.47 | Net income to common ($403.0M / 274.0M) |
| ROE −0.20 | −20% | Consolidated loss over total equity (−$958.5M / ~$4,827M) |
The same feed reports a 6x P/E and a −20% ROE on the same company in the same period. Neither is usable, and the contradiction is free to spot.
What this cause adds to the detection routine: when the tripwire fires, check whether net income attributable to common differs from consolidated net income. If it does, the wedge is a minority-interest allocation, and the questions are "is it structural?" and "which numerator is each vendor ratio using?" — not "what one-off do I strip?" Common in tax-equity partnerships, yieldcos, Up-C structures, and any consolidated-but-shared-economics vehicle.
Known blind spot — contamination that sits ABOVE the operating line (2026-08-17)
The check has a hole, and the same company found it. All four causes above are below the operating line, which is why the subtraction works. An acquisition earnout revaluation is booked in operating expense, so it moves operating income and net income together and the tripwire stays silent.
AMSC's quarter ended 30 June 2026: GAAP operating margin 10.5%, net margin 10.1% — the check passes cleanly. True operating margin was 1.8%. The difference was an $8,115K credit from writing down the Comtrafo earnout, sitting inside total operating expenses. The same ticker that supplied the severest case of this pattern (a 9x gap, one quarter earlier) also supplies its first documented evasion.
A passing tripwire is not a clean bill of health on an acquirer. Add one step: if the company has bought anything in the last three years, read the operating-expense detail and use segment operating income. Full mechanism, detection steps and the inverted-sign problem: [[pitfall-earnout-revaluation-inflates-operating-income]].
History
- 2026-08-26 — Cause 6 added (RUN): HLBV minority-interest loss allocation. The first cause that is structural and permanently recurring rather than one-off, and the first where the vendor feed contradicts itself — P/E built on income-to-common, ROE on consolidated loss.
- 2026-08-04 — Established from the tech sweep (AMSC, UMC).
- 2026-08-06 — Case 4 (disposal gain) confirmed at FIS; added the two refinements above.
- 2026-08-20 — Cause 5 added (MRAM): an expiring contract booked below the operating line. First variant that is real recurring revenue rather than a mark or one-off, and the first where the balance-sheet cross-check does not confirm. Detection now requires plotting the line by quarter and asking whether it has a contractual end date.
- 2026-08-17 — Blind spot documented. The check cannot see contamination booked above
the operating line. Pattern held at
static/high— the four below-the-line causes are unchanged and the subtraction still works; it is now known to be necessary but not sufficient on serial acquirers.