When to sell: writing an investment thesis with invalidation signals

A slightly unfair thought experiment to start. Picture the worst possible version of you. Not evil-you. Compromised-you: it is a Tuesday, a position is down a third since lunch, there is a headline you have read four times and absorbed zero per cent of, and your heart is doing the thing. Now, in that exact state, answer a hard question about whether the company’s long-term prospects have actually changed. (You will not take your time.)

That, in one cursed scene, is the problem with “when do I sell a stock”: you ask it at the precise moment you are least equipped to answer it.

Why the price quietly does your thinking for you

Here is the sneaky bit. When a position is red, your brain reaches for a story that lets you keep holding: the market is wrong, this is temporary, I am the patient one here. When it is green, it reaches for the opposite story so you can take the win. Both feel like analysis. Neither is. They are vibes wearing a lab coat.

Call this the Reconstruction Problem. Mid-move, you do not really remember why you bought the thing. You rebuild the reason out of the price, like an archaeologist secretly on the payroll of the dig. The outcome has become the input, and the one variable you cannot be objective about is steering the wheel.

The fix is almost annoyingly simple: decide what would make you sell before you buy, while you still have no loss to defend and a working brain. Write it down. That document is a thesis, best pictured as a letter from calm-you to panicking-you. Calm-you sized the position on purpose and was not, at the time, being chased by a candlestick. Panicking-you just has to open the letter at the worst moment and be told what was agreed.

What goes in the letter

A thesis is not a price target and a feeling. (A price target and a feeling is a horoscope.) It is a short living document on one ticker that says why the position exists and what would prove it wrong. It has a few moving parts: the drivers, the two or three mechanisms you are betting on; the key assumptions propping those drivers up, which, fun fact, are usually where the whole thing cracks first; the position sizing, because a probe and a fat conviction bet are different theses about the same company; the invalidation rules, the part everybody skips and the only part that answers “when do I sell”; and the conviction level, revisited on a schedule rather than left to drift after the price like a duckling. Pair it with a journal of what you did and you get the discipline loop: reasoning written down, left to age, then made to face what happened.

Invalidation signals, or: deciding now what future-you may not argue with

An invalidation rule only becomes trustworthy when it stops needing your opinion. That is what an invalidation signal is: a metric condition that, when breached, flips the thesis to AT-RISK by itself. No judgment call at the moment it trips, because you made the judgment back when you were calm and the stakes were hypothetical.

The trick is to write it as a number rather than a narrative. “The story has changed” describes a mood. A threshold you set in advance is something a mood cannot wriggle out of. A couple of made-up illustrations of the shape, not claims about any real company. If the bet rests on pricing power, set a gross-margin floor: margin below some chosen level means the thing you were betting on is leaking. If you bought a trend, a price below a long moving average is the trend ending, no eulogy required. The numbers themselves do not matter; what matters is that you picked them ahead of time, tied each to an assumption, and left no wiggle room. A signal you have to interpret in the moment is one you will talk yourself out of.

This is where the workbench earns its keep. When a signal breaches in Strata, the thesis flips to AT-RISK on its own and tells you which past-you-condition just tripped, across a long list it can watch for you, from income-statement metrics through to price and valuation ratios. And to be clear, AT-RISK is not a sell order but a forced re-examination. Maybe the thesis really is broken and you exit; maybe the assumption that broke was not the only leg it stood on, so you rewrite the thesis on the record rather than quietly hold and hope. Either way the decision comes from reasoning, not the size of the red number.

A thesis you will actually write

Keep it short enough that writing it does not feel like homework, because the elaborate version is the one you never fill in. Name the ticker, the size, and why that size. List the two or three drivers that have to go right, and under each, the assumption holding it up. Then the invalidation signals as metric conditions, each bolted to an assumption above, like a margin floor or a price level. Then your conviction now and the date you will revisit it. And, because green positions deserve the same discipline as red, a line for what would make you add.

If you cannot fill in the invalidation line, you do not have a thesis yet. You have a hope with a ticker attached. Writing that line is the actual work: it makes you name, in advance, the exact evidence that would change your mind, while you are still honest enough to do it.

Let the letter do its job

So the position turns against you. Your job is not to re-decide everything with your worst brain; it is to open the letter. If nothing has tripped, the drawdown is just noise and you sit very still and feel slightly smug. If a signal has breached, you re-examine on the terms calm-you set, not the terms the price is shouting. Your mood, for once, does not get a vote.

You can start a thesis before your next buy, and let the letter, not the panic, decide when to sell.