How do you know when to sell a stock you have held for years?

Why the question is hard after years rather than months

A position held for six months is still attached to the argument that created it. A position held for six years usually is not. The company has reported a dozen times since, the story has been retold in your head with the benefit of what happened, and the original reasoning has quietly been replaced by something much weaker: the fact that you own it, and that so far it has worked.

This is why the sell decision feels so much harder than the buy decision, and why it is so often made on price instead. Price is the only number still visible. The argument is gone.

So the useful version of the question is not “should I sell this?” It is “what did I believe, and is it still true?” Everything below is about getting back to a position where that second question can actually be answered.

Step one: reconstruct the thesis you cannot remember

Before deciding anything, write down — in the past tense, honestly — what you thought when you bought. Three or four sentences is enough. If you genuinely cannot recall, reconstruct it from what you can find: the price you paid tells you what valuation you accepted, and the accounts published just before that date tell you what you were looking at.

The test of whether you have written a real thesis is whether it could be wrong. These are not theses:

• “It is a great company.”
• “They have a moat.”
• “Everyone uses their products.”

Each of those is compatible with every possible future, which means no piece of news can ever contradict them. These are theses, because each one names something that could stop happening:

• “Revenue compounds above 15% because they keep taking share in a market that is itself growing.”
• “Gross margin stays above 60% because switching away costs customers more than the price increase.”
• “Capital allocation is disciplined: they buy back stock below intrinsic value rather than making large acquisitions.”

Write the version you actually held, not the version that would look best now. A reconstructed thesis that quietly incorporates what you have since learned is not a benchmark, it is a justification.

Step two: name what would make each part false

For each sentence, write the condition that would tell you it has stopped being true. This is a kill criterion, and there is a fuller guide to writing them. Two properties make one worth having: it is observable in a filing or a set of accounts, and you could tell whether it had happened without knowing the share price.

“The stock falls 30%” fails both. “Gross margin below 60% for two consecutive quarters” passes both. So does “the founder-CEO leaves and the replacement comes from outside the industry”, which is qualitative but perfectly checkable.

Writing these years late feels artificial, and it is — you are setting a rule knowing some of the outcome. Do it anyway, and date it. From today forward it is a real benchmark, which is more than the position had this morning.

Step three: check the evidence, in the right order

Not all sources answer the question equally well, and the order matters more than people expect.

• Annual and quarterly reports are the only evidence that can confirm a thesis, because a thesis is a claim about how the business performs over a period and these are the record of a period. Read the management discussion, not the headline number.
• Earnings calls are next, mostly for what management is asked and declines to answer. The questions analysts keep returning to are usually where the doubt is.
• Material events — an acquisition, a resignation, a restatement — can break a thesis in a single document, but they can never confirm one.
• News can raise a concern and should never settle one. One journalist’s account of one event is not evidence about how a business performs over years, however well written it is.

Notice what is absent: the share price, analyst ratings, and how the position has performed. None of them is evidence about whether your reasoning still holds. They are evidence about what other people currently think, which is a different question and one you were not trying to answer.

The four answers, and what each one means

Done properly, this produces one of four conclusions, and only one of them is “sell”.

• The thesis holds. The evidence is consistent with what you believed. Do nothing — and record that you checked, so next year you know when it was last tested.
• Something has shifted but nothing has broken. A number is moving the wrong way, or a criterion is close to its threshold. This is a reason to watch and to write down what you are watching for, not a reason to act.
• A stated condition has been met. The thing you said would mean you were wrong has happened. This is the case you wrote the criterion for. It does not sell for you, but it obliges you to either sell or write down, explicitly, why the criterion was wrong rather than the company.
• The thesis has been replaced without you noticing. The reasons you would give today are not the reasons you gave then, and the new ones arrived after the facts did. This is the most common outcome and the least comfortable. Treat it as a fresh buy decision: at today’s price, with today’s information, would you open this position? If not, you are holding out of habit.

Four reasons people hold anyway

The analysis above is not usually what stops people. These are.

• The tax bill. A real cost and a real input, but it belongs in the arithmetic rather than in the decision about whether the thesis holds. Answer that question first, then work out what the tax changes. Done the other way round, the tax bill becomes a reason never to sell anything.
• The gain so far. What you paid is not a fact about the company. A position that has tripled and one that has halved should be judged by the same evidence, and the strong pull to treat them differently is the single most reliable way to hold a broken thesis and sell a working one.
• Identity. After enough years, owning something becomes part of how you see yourself as an investor. This is worth naming out loud, because it is invisible while it is operating.
• Not wanting to be wrong in public. Including in front of yourself. Writing the reasoning down at the time is what makes this survivable later: a recorded thesis that failed is a data point, while an unrecorded one that failed just feels like foolishness.

How often to actually do this

Every set of annual results, and any time something material happens — an acquisition, a change of chief executive, a restatement, a large change in how the business makes money. That is roughly two to five occasions a year for a typical holding.

Not on price moves. A stock that falls 30% while every reason you own it stays intact deserves no review at all; one that rises 30% on nothing may deserve one. Reviewing on price guarantees that you look hardest exactly when you are least able to think clearly.

In short

You know it is time to sell when a condition you named in advance has been met, or when you would not buy the position today at today’s price with what you now know. Everything else — the drawdown, the run-up, the boredom, the headline — is noise dressed as a signal.

The reason this is hard is not that the rule is complicated. It is that following it requires a written record of what you believed, and almost nobody keeps one. That is the specific problem this product exists to solve: you write the thesis and the conditions once, and every filing, earnings call and material news item that follows is read against them, so the question “does this still hold?” has an answer you did not have to remember.

See what an audit looks like · Write your first thesis

Related

• Kill criteria: deciding in advance what would make you sell
• Keeping an investment decision journal that is worth re-reading
• How the engine decides