The Subtraction Framework
The index guarantees exposure to rare right-tail winners. If predicting those winners in advance is hard, a parallel claim may be more tractable: remove probable wealth destroyers while preserving enough breadth to capture them.
The base rate case for subtraction
The S&P 500 works partly because it guarantees exposure to rare right-tail winners. An investor who held the index from 1990 to 2020 captured the returns of Microsoft, Apple, Amazon, and the other handful of companies that accounted for most net wealth creation — without having to identify them in advance.
If right-tail selection is genuinely hard, an alternative claim becomes more interesting: can you modestly improve long-run outcomes by removing companies that are likely to destroy capital, without pretending to predict every future winner? The Subtraction Framework is the attempt to answer that question seriously.
This is not a claim that negative selection always beats concentration. It is a claim that removing high-probability losers is a more tractable task than predicting high-probability winners, and that the two approaches are compatible.
What the subtraction framework is not
It is not classic stock picking in reverse. It is not a factor strategy dressed in new language. It is not market timing or sector rotation. And it is not a mechanical screen applied without judgment.
The screens identify candidates for removal. Judgment still determines whether the screen result reflects a genuine deterioration in the business or a temporary accounting artifact. The framework is a starting filter, not a final answer.
The negative screens
Eight screens define the initial removal candidates: deteriorating net income trajectory on a trailing multi-year basis; weak organic growth quality, where revenue growth is driven by acquisition or currency rather than unit economics; declining return on assets over a three-to-five-year window; negative spread between ROIC and estimated WACC; worsening leverage, particularly in cyclical businesses; weak free-cash-flow conversion relative to reported earnings; customer concentration above 30% in a single payer or counterparty; and capital-infusion dependence — businesses that require ongoing external capital to fund operations.
A business that fails two or more screens without a clear temporary explanation moves onto the removal list. A single screen failure with an identifiable cause may be watched rather than removed.
The Bessemer Converter
Inversion produces a sharper starting point than selection. Before asking which businesses deserve more work, ask which businesses should stop the analysis immediately.
The 15-minute rule: if any of the following appear in the first pass, research ends. Debt brittleness — a balance sheet that cannot survive a two-year revenue contraction without covenant breach or equity issuance. Cyclicality mistaken for secular growth — revenue that tracks end-market volumes but is framed as a structural compounder. Substrate obsolescence — a physical or regulatory substrate that has a visible, better-capitalized replacement already scaling.
Capital-allocation red flags stop work at the next stage: high asset growth that outpaces revenue growth over five years; acquisition camouflage, where EBITDA grows but free cash flow does not; ROIC below WACC on a through-cycle basis.
Epistemological no-go zones are businesses where the most important variables are genuinely unknowable in a 25-year frame: businesses fully exposed to AI commoditization of their core product, regulatory businesses where the next rate case or policy ruling is unpredictable and material, and businesses in jurisdictions with high political expropriation risk.