What Are Kill Criteria in Stock Research? Falsifiable Investing Explained

The definition

Kill criteria (thesis-breakers) are the pre-stated conditions under which an investment thesis must be considered broken. Each one names something concrete — a metric crossing a threshold, a competitive event, a disclosure in a filing — that, if it happens, means the original reasoning was wrong.

Two things separate kill criteria from ordinary risk commentary. First, they are decided in advance, so they cannot be quietly rewritten to fit whatever happened. Second, they are checkable: a future earnings call, filing, or disclosed number can settle whether each criterion has been met. A risk section says “this could go wrong.” A kill criterion says “if this specific thing goes wrong, the thesis is dead.”

Why a thesis must be falsifiable

A thesis that cannot be proven wrong is not a thesis — it is a narrative flexible enough to explain any result. The stock falls 40%? “The market hasn’t caught up yet.” Growth stalls? “A pause before the real inflection.” Every outcome gets absorbed, and the position is never actually re-examined.

This is the failure mode kill criteria exist to prevent. Without pre-stated invalidation conditions, investors tend to redefine the thesis to fit the evidence rather than test it against the evidence — confirmation bias doing its quiet work. Writing down the thesis-breakers in advance removes the room to rationalize: either the named conditions occurred or they did not.

Falsifiability also sharpens the thesis itself. Asking “what would prove me wrong?” exposes which assumptions the bull case actually depends on. A thesis with no identifiable kill criteria is not a strong thesis — it is an untested one.

What good kill criteria look like

Good kill criteria share three properties, and each has a vague counterfeit that looks similar but does no work.

Specific. A good criterion names a metric, threshold, or event: “gross margin falls below a stated floor for two consecutive quarters.” Its counterfeit is “if fundamentals deteriorate” — elastic enough that no quarter ever quite triggers it.

Observable. A good criterion can be checked against a future filing, earnings call, or disclosed number by anyone reading it. Its counterfeit is the unverifiable — “if management loses its edge” — which turns the test into a judgment call the investor will resolve in their own favor.

Decided in advance. A good criterion is committed to before the outcome is known, so meeting it forces a re-evaluation. Its counterfeit is the retrofit: deciding after a bad quarter that this miss “doesn’t really count” because the criterion was never pinned down to begin with.

A useful self-test: could a skeptical reader, twelve months from now, say definitively from public information whether each criterion was met? If not, it is a mood, not a kill criterion.

A concrete example

An example Monsaic analysis of NVIDIA Corporation (NVDA) — an illustration of the format, not advice — carried this one-line thesis: a system-level AI factory supplier, still earnings-revision driven, with valuation sensitive to hyperscaler capital spending and gross-margin normalization. Its kill criteria were:

• Hyperscaler and AI cloud capex plans roll over for two consecutive quarters.
• Gross margin structurally falls below 68% without a clear mix-transition explanation.
• Customer silicon or competitor accelerators take enough share to flatten NVIDIA’s data center growth.

Notice what each one does. “Capex could slow” would be a generic risk; “capex plans roll over for two consecutive quarters” is a bar checkable against the next two quarters of hyperscaler disclosures. The margin criterion names a number — 68% — and anticipates the exception that would not count (a mix transition with a clear explanation). The same analysis stated, in its valuation scenarios, what would break the bear case too: demand staying sold out and earnings revisions continuing to move up. Falsifiability cuts both ways.

Kill criteria vs. a stop loss

A stop loss reacts to price; kill criteria react to evidence about the business. A stop loss says “sell if the stock falls to X,” regardless of why. Kill criteria say “the thesis is wrong if these conditions occur,” regardless of what the price is doing.

The two protect against different failures. Price can fall sharply while the thesis holds — a market-wide drawdown can trigger a stop loss on a business whose fundamentals never wavered. And the thesis can break while the price holds — a kill criterion can be met in a filing months before the market reprices it, while a stop loss stays silent because the chart looks fine. A stop loss manages the size of a loss; kill criteria test whether the reasoning behind the position is still true.

How Monsaic uses kill criteria

Monsaic is an AI stock research platform, and every Monsaic report defines its kill criteria before the grade is assigned — so the verdict is falsifiable by construction, not graded first and rationalized after. The kill criteria sit alongside the report’s valuation scenarios, claim-level citations, and an explicit “as of” date, so a reader can check the thesis-breakers against future evidence. The full discipline is laid out in the Monsaic methodology, and the broader case for this standard in how AI stock research should work. Kill criteria are only as good as the public sources behind them, and they are analysis inputs, not personalized guidance.

FAQ

See a kill criterion on a live thesis

Kill criteria only mean something attached to a real position. Every covered stock’s excerpt states the condition that would make Monsaic change its verdict — read one and judge whether it’s specific enough to actually fail.

Browse covered stocksSee pricing

Keep reading

• AI stock research, done rigorously — the Monsaic pillar
• The Monsaic methodology
• What is a valuation scenario?
• Generate your first Monsaic