Sign in

Investor behaviour

What should make you sell: writing invalidation conditions

An invalidation condition is a fact about a business, decided in advance, that would tell you your reasoning was wrong — not a price level. "If gross margin falls below 40% for two consecutive quarters" is an invalidation condition. "If it drops 20%" is a stop-loss, which manages risk but says nothing about whether you were right. Writing invalidation conditions before you buy does two things: it forces the thesis to be specific enough to be falsifiable, which often reveals there wasn't one, and it commits you to an exit while you can still think clearly.

· 5 min read · Synoptiv

The asymmetry nobody plans for

Most investors can explain why they bought something. Very few can state what would make them sell — beyond a vague sense that they would know it when they saw it.

This asymmetry is where a large share of avoidable losses live. The buy decision is made calmly, with time and research. The sell decision arrives during a drawdown, with incomplete information, under stress, and against a position you are emotionally invested in defending. Those are the worst possible conditions for clear thinking, and it is precisely when the decision gets made.

The fix is unglamorous: decide in advance, in writing, while you are still calm.

What an invalidation condition actually is

An invalidation condition is a specific, observable fact about the business that would demonstrate your reasoning was wrong.

Three properties, all necessary:

Specific. "If the competitive position weakens" is not a condition — it is a feeling with a date on it. "If gross margin falls below 40% for two consecutive quarters" is a condition.

Observable. You have to be able to check it from public information on a known schedule. A condition that depends on knowing something you cannot know is decoration.

Decided in advance. A condition written after the fact will always be written to accommodate what has already happened.

Note what is absent from all of that: the share price.

Why a price stop is not an invalidation condition

Both are worth having. They are not the same tool and substituting one for the other is the central error here.

A stop-loss answers: how much am I willing to lose on this position? It is position-sizing and risk management. It is agnostic about whether you were right.

An invalidation condition answers: was my reasoning correct?

These come apart constantly, in both directions:

  • A stock falls 30% in a sector-wide drawdown while the company executes exactly as you expected. The stop fires. The thesis was never invalidated — you were right and sold anyway.
  • A stock rises 20% while the largest customer quietly leaves and margins compress. Nothing fires. The thesis broke while the position was profitable, which is the more dangerous case because profit suppresses scrutiny.

The second is the expensive one. Gains are the most effective disguise a broken thesis has.

What good conditions look like

The test: could two people disagree about whether it had happened? If yes, it is not specific enough.

Weak — unfalsifiable:

  • "If the growth story doesn't play out"
  • "If management stops executing"
  • "If the competitive moat erodes"

Strong — checkable:

  • "If revenue growth falls below 15% year over year for two consecutive quarters"
  • "If the largest customer, at 22% of revenue, does not renew in the FY27 cycle"
  • "If free cash flow remains negative through FY27 after guidance said it would turn positive"
  • "If the CFO and CEO both depart within twelve months"
  • "If gross margin compresses below 55% while revenue growth is under 20%"

Notice the last one is a conjunction. Margin compression during rapid growth may be a deliberate trade; margin compression without growth is a different situation entirely. Real theses usually depend on combinations, and stating them that way is more honest than pretending a single metric carries the argument.

How to find yours

The reliable method is to write your thesis as a short argument, then attack it.

  1. State the thesis in three sentences. What the company does, why it should be worth more than it is now, and over what period.
  2. Underline the assumptions. Every "because" and every implicit forecast. There are usually four or five.
  3. Rank them by load. Which assumptions, if false, break the whole argument — as opposed to merely denting it? Typically two or three.
  4. Turn each into an observable event. What would you see, in a filing or a release, if that assumption were failing?
  5. Set a review schedule. Most conditions are checkable quarterly. Put it in a calendar, because an unchecked condition is the same as no condition.

Step 2 is where the surprises happen. It is common to discover that a thesis rests on an assumption you never consciously made — usually about a growth rate continuing, or a margin holding.

The part that is genuinely hard

Writing conditions is easy. Honouring them is not.

When a condition triggers, you will have a reason why this instance does not count. Sometimes that reason will be legitimate — new information genuinely changes things, and refusing to update is its own failure mode. But the base rate strongly favours motivated reasoning, and everyone believes they are the exception.

Two things help.

Write the condition and the consequence together. Not "if margin falls below 40% I'll reassess" — reassessing is what you would do anyway. "If margin falls below 40% for two quarters, I sell half." A pre-committed action is harder to rationalise around than a pre-committed feeling.

Keep the original document, dated, and version it. When you revise a thesis, record what changed and why, rather than editing in place. Over a few years this becomes the most valuable investing document you own: a record of what you believed, what actually happened, and how you tended to be wrong. Nobody's memory preserves that honestly — memory rewrites the thesis to fit the outcome.

Why this is the most useful habit on the list

Invalidation conditions do not improve your stock picking directly. What they do is make your reasoning auditable — by your future self, who will otherwise remember a much more coherent version of what you thought.

They also make "I don't know" a legitimate output. If you cannot construct a condition that would change your mind, the position is conviction rather than analysis. That is allowed — but it should be a choice you make knowingly, not one you discover in a drawdown.

Synoptiv publishes invalidation conditions alongside every bull and bear case for exactly this reason, and versions each analysis as the story changes rather than overwriting it. An analysis you cannot hold accountable is not much of an analysis — the same standard applies to AI-written ones.

Common questions

What is an investment thesis invalidation condition?

A specific, observable fact that would demonstrate the reasoning behind a position was wrong. It is stated before buying, describes the business rather than the share price, and is concrete enough that you could not argue about whether it had happened. For example: 'the largest customer, at 22% of revenue, does not renew.'

Is a stop-loss the same as an invalidation condition?

No, and conflating them is a common and expensive error. A stop-loss caps the money at risk in a position. An invalidation condition tells you whether your analysis was correct. A stock can fall 30% while the thesis remains entirely intact, and it can rise 30% while the thesis quietly breaks. Both tools are useful; they answer different questions.

How many invalidation conditions should a thesis have?

Two to four. One is usually not enough to cover the ways a thesis can fail. More than four generally means they are not the real load-bearing assumptions, and a long list is easy to ignore — which defeats the purpose.

What if I cannot think of anything that would change my mind?

That is a finding, and an important one. It almost always means the thesis is a general impression rather than an argument — 'it's a great company' cannot be falsified. Either the reasoning needs to be made specific enough to be wrong, or the position is being taken on conviction rather than analysis, which is worth knowing before the money is committed.

Terms used here

See this applied to a real company

Synoptiv runs this kind of analysis on US stocks and publishes the reasoning — including what would change our mind. Browse analyzed stocks or read how the analysis is produced.

Analysis and education, not investment advice. Nothing here is a recommendation to buy or sell any security.