How to Write an Investment Thesis You Can Actually Screen Against

Most investment theses are written for limited partners, not for sourcing. They describe a worldview: the future of work, the next billion users, AI everywhere. That is useful for raising a fund, and nearly useless for deciding which of the hundreds of companies founded this week deserve a closer look.

A thesis you can screen against is different. It is short, concrete and testable. Given a company's description and the evidence about it, you should be able to say quickly whether it fits.

The Five Parts of a Screenable Thesis

1. Market. The problem space, as narrowly as you honestly mean it. "Freight brokerage" rather than "logistics".
2. Customer. Who pays. "Mid-size freight brokers" rather than "businesses".
3. Stage. Where you invest. "Pre-seed to seed".
4. Geography. Where the company is based or sells, if that matters to you.
5. Proof. The evidence you want to see before you take a meeting: a pilot customer, early revenue, a technical founder, open-source adoption.

Put together, it reads like this:

B2B AI agents for mid-size freight brokers in emerging markets, pre-seed to seed, with at least one pilot customer.

Every word in that sentence removes companies that do not fit.

Examples You Can Adapt

Developer tools that cut cloud costs for mid-size engineering teams, with open-source traction and a commercial product in sight.

Embedded finance for small businesses in Southeast Asia: lending, payroll or treasury, with early revenue.

Software for clinics and small hospitals in Africa that replaces paper workflows, seed stage, with paying customers.

Common Mistakes

• Too broad. "AI startups" matches thousands of companies a month and screens out nothing.
• Only describing the market. Without customer, stage and proof, you will match companies you would never invest in.
• Hidden criteria. If you would pass on companies without a technical founder, write that down. A thesis is only useful if it contains the reasons you actually say no.
• Never revising it. Look at the companies you passed on each quarter. If good ones keep failing your thesis, rewrite it.

Test It Against Deals You Already Know

Before you rely on a thesis, run it against history. Take the last twenty companies you invested in or seriously considered, and the last twenty you passed on. For each one, ask: would this thesis have flagged it?

• If companies you invested in fail the thesis, it is missing something you actually value. Find the criterion and write it in.
• If companies you passed on pass the thesis, it is too loose. Find the reason you said no and add it.
• If most companies fail for the same single reason, check whether that reason really matters, or whether it is an old habit.

Twenty minutes of this exercise usually changes a thesis more than a week of debate.

From Sentences to Filters

Some parts of a thesis map directly to filters, and some need judgement. Geography, stage and sector can usually be filtered mechanically. "A technical founder" or "a pilot customer" need someone, or something, to read the evidence. A good process uses filters to cut the volume and reserves judgement for what is left.

That split is also why a precise sentence beats a long document. A screener, human or software, can hold one clear sentence in mind across hundreds of companies. It cannot do the same with ten pages.

Keep a Few, Not One

Many funds have two or three distinct theses at the same time. Keep them separate. A company can fit one strongly and another not at all, and merging them into one long paragraph makes every match weaker.

Let Software Do the First Pass

Once your thesis is precise, the first screening pass is mechanical: read each new company's description and evidence, and judge the fit. That is exactly what software is good at. In AlphaScout, you write your thesis in plain English, and every morning each new company is scored from 0 to 100 against it, with a short reason. Strong fits arrive by e-mail. See thesis matching for how it works.