The three questions I ask Claude before I touch a single number

Most investors open a spreadsheet before they understand the business.

The numbers feel like objcetivity

But a model built on a misunderstood business is just a precise way to be wrong.

Here's how I changed the sequence.

There’s a trap that catches even experienced investors.

You find a company, the revenue growth looks interesting, the margins are expanding, and within ten minutes you’re running scenarios in a DCF.

You feel prroductive.

You’re not doing analysis yet.

You’re doing arithmetic on a story you haven’t actually read.

Numbers describe what happened.

They don’t explain why it happened, whether management actually drove it or just happened to be in the right place, or whether it will keep happening as the business scales and competition intensifies.

Without that context, the model answers the wrong question with a lot of decimal places….

Before I open a spreadsheet on any company, I run three questions through Claude.

I give it the relevant sections of the 10-K or 20-F, sometimes the latest earnings transcript, and I work through each question in order.

The whole sequence takes about fifteen minutes.

What I get out of it changes everything I look at afterward.

Not the product description.

Not the investor relations summary that says “we are a leading provider of innovative solutions.”

The actual economic mechanism.

Where does cash come from, on what cadence, and what does the customer have to believe is true to keep paying for it year after year?

This sounds obvious.

It almost never is.

Companies describe their businesses in ways that obscure the unit economics.

A SaaS company might report subscription revenue while the real value driver is professional services that lock customers in.

A retailer might look like a margin story when it’s actually a working capital story.

The 10-K business description is written by lawyers and IR teams, not by people trying to help you understand the business.

I paste the “Business” and “Management’s Discussion and Analysis” sections and ask Claude to do two things: explain the actual revenue mechanism in plain terms, then identify the single biggest assumption the entire model depends on.

That assumption becomes the center of every research question I ask next.

Most competitive analysis in analyst reports is cosmetic.

They list the obvious competitors, note that competition is intense, and move on.

The real threat almost always comes from somewhere the company isn’t watching closely, either because it’s in a different industry, operating at a different price point, or building a capability that won’t matter for two years and then suddenly matters a lot.

I ask Claude to think about competitive threat in three categories.

Direct competitors are easy to name and usually already priced in.

Adjacent players are companies in related markets that could expand into this one if margins are attractive enough.

Non-obvious threats are the interesting ones: businesses from entirely different industries that could undercut the core value proposition with a different delivery mechanism entirely.

The third category requires Claude to think harder, which is why the prompt forces it to be specific about the mechanism of disruption, not just identify who the threat might be.

A name without a mechanism is not useful research.

This is the most important question in the sequence. It forces you to name the conditions before you’ve already convinced yourself they’re met.

Most investors work backward. They get interested in a company, build a model, and then generate a buy thesis to explain the model. The conditions for success get identified after the conclusion. That’s backwards. You end up finding evidence for what you already believe.

This prompt forces Claude to derive the required conditions from the business model itself, before I’ve expressed any opinion. It then tests each condition against the company’s own historical behavior and the industry’s base rates. What you’re left with is a list of specific bets, each with a rough honest probability attached. That’s what an investment thesis actually is.