What is an investment thesis (and how to actually write one)

Most retail investors can tell you what they bought. Few can tell you why, in a way specific enough to be proven wrong. That gap is the whole problem. If you can’t say what would change your mind about a position, you don’t have a thesis. You have a feeling with a ticker attached to it.

An investment thesis is a written statement of why you own a position, broken into claims you can check, the drivers that keep those claims true, and the invalidators that would prove them wrong. This guide covers what that looks like in practice, a template you can copy, and a worked example on a real ticker.

What is an investment thesis?

An investment thesis is a written statement of why you believe a position will work, broken into the specific and observable things that have to be true for that belief to hold. It is a set of claims you’re making about a business, plus a way to know when the world starts touching those claims. It is not a price target. It is not a prediction.

A real thesis has three parts.

• The claim. What you believe about the business. “Margins expand as the company scales its logistics network.” Not “this stock is going up.”
• The drivers. The mechanism that has to hold for the claim to stay true. Revenue growth, a product cycle, a cost structure, a regulatory outcome. Something you can track.
• The invalidators. What would prove you wrong. If you can’t answer this, the claim isn’t falsifiable, and an unfalsifiable claim can’t be monitored. It can only be believed.

Three parts, one page. If it doesn’t fit on a page, it isn’t a thesis yet. It’s research that hasn’t been forced into a decision.

Why do most investors skip it?

Writing a thesis takes ten minutes and forces you to admit what you don’t know. Buying a stock takes a few seconds and asks nothing of you. Most platforms are built around the second experience, because the first one is friction and friction looks like a bad metric on a dashboard.

The deeper reason is less comfortable. Writing an invalidator means naming the event that would make you wrong, and part of you doesn’t want it named. An unwritten thesis can quietly survive anything. A written one can’t. Most investors aren’t avoiding the ten minutes. They’re avoiding the exposure.

The ten minutes is the point. The friction is where the thinking happens.

What does a good claim look like?

The difference between a thesis and a narrative is checkability. Compare:

Every claim in your thesis should survive this test: a stranger could read it, watch the next two earnings reports, and tell you whether it’s holding.

One honest exception: some claims resolve slowly. Management quality, a culture, an option on a market that doesn’t exist yet. You can’t falsify those with a quarterly print, and pretending you can produces fake precision. The fix isn’t to drop the invalidator. It’s to convert it into a milestone with a date: what should be observable, by when. If the date passes and it isn’t, that’s the trip.

The template

Copy this template. Fill it in for one position you already hold.

TICKER:
Why I own it, in one sentence:
What the market believes instead:

Claim 1:
  Driver:
  Invalidator:

Claim 2:
  Driver:
  Invalidator:

Time horizon:
What I'd need to see to add to the position:

Or write it in Thesis, where the same structure gets monitored against filings and news for you.

Rules for filling it in:

• Start with the one sentence. If you can’t write it, you don’t own a position, you own a hunch.
• Then write what the market believes instead. If you can’t name a belief you’re diverging from, you don’t have an edge, you have exposure. A thesis that agrees with everyone is a description of the price.
• Break the sentence into claims. Most positions have one to three real drivers. If you’re listing six, you have one thesis and five hopes.
• Make every driver a number or an event, not an adjective. “Strong execution” is not a driver. “Cloud revenue growth above 25% year over year” is.
• Make every invalidator specific enough to trip. “The market turns bearish” is not an invalidator. “Gross margin below 40% for two consecutive quarters” is.
• Write it somewhere you’ll actually revisit. Not a note buried in your phone. Somewhere that comes back to you when the business reports earnings, files with the SEC, or shows up in the news. That last part is the exact gap Thesis was built to close.

A worked example: NVDA

Numbers as of August 2026 (Nvidia’s fiscal Q1 2027). They will age. That’s the point of a thesis: it’s checkable later. NVDA is deliberately the example: you already know the story, so the only new thing to see is the structure.

Now watch how the thesis earns its keep. Two months later, two things happen.

A hyperscaler announces expanded internal silicon capacity for inference workloads. That doesn’t trip the invalidator as written, which was scoped to training. But it touches Claim 1, so it gets logged against that claim with a date, and the next print gets read with it in mind.

The stock drops 6% in a broad tech selloff. Nothing gets logged. No claim mentions the tape. This is most days, and the correct response to most days is nothing.

Why write it down?

A written thesis does two things a gut feeling can’t.

It gives you a way to know when to update your thinking, instead of reacting to every headline that mentions the ticker. Most news doesn’t touch your actual reasoning. A written thesis tells you which pieces do.

It gives you a record. Six months from now you can look back at what you believed and why, and check it against what happened. That’s the only way to get better at this. Not by being right or wrong on a single trade, but by seeing whether your reasoning holds up over time.

How do you keep a thesis alive?

A thesis is not a document you write once. It’s a position you maintain.

• On every earnings report, score each driver: confirmed, challenged, or broken. Use the numbers, not the market’s reaction. A stock can drop 8% on a print that confirmed every driver you wrote.
• When a filing or a news event touches a claim, log it against that claim, with a date. Most news will touch nothing. That’s information too.
• Once a quarter, reread the whole thing. If you catch yourself rewriting claims to fit what the price did, stop. That’s thesis drift, the quiet failure mode where the story mutates so it can never be wrong. The original claim either held or it didn’t.
• When an invalidator trips, act on the rule you wrote when you were calm. The entire value of writing it in advance is that the decision was made by the version of you that wasn’t watching the position bleed.

Where Thesis helps

This is what we built Thesis around. You write the claims. We read the filings, the earnings transcripts, and the news, and we tell you only when something touches one of your drivers or invalidators. Nothing otherwise.

Concretely: the hyperscaler event from the example above reaches you looking like this.

Silence is a feature. Most days, nothing has happened to your thesis, and you should know that too.

Frequently asked questions