The Discipline of Writing Down an Investment Thesis

The most expensive habit in investing may be the unwritten thesis. In your head, the story is flexible. After a gain, you remember brilliance. After a loss, you remember bad luck. Writing is how you make the story accountable to its earlier self.

A useful thesis is short enough to revisit and specific enough to falsify. It states the claim, the time horizon, the main drivers, the risks, and the invalidation points. “I like this company” is not a thesis. “I expect operating margins to expand because of X within Y quarters, and I am wrong if Z happens” is closer.

Invalidation points are the part people skip. Without them, every dip is a gift and every rally is confirmation. With them, you have a pre-negotiated exit from ego. Price alone is a weak invalidation—regimes change—but price plus broken fundamentals is a strong one. Write both.

The thesis document is also a risk tool. It forces you to name leverage, liquidity, and crowding. It asks whether you are paid for the risks you listed. Many ideas die at this stage, which is a success. Capital saved from vague enthusiasm is capital available for clearer bets.

In team settings, writing reduces the charisma premium. A fluent speaker can move a room without moving the evidence. A shared memo lets quieter analysts puncture weak claims. Over time, culture becomes about the quality of artifacts, not the volume of confidence.

Digital-asset research needs the same discipline, perhaps more so, because narratives move faster than filings. If your thesis is “number go up,” you will not know when to leave. If your thesis names adoption metrics, fee sustainability, regulatory pathways, or catalyst calendars, you can disagree with yourself productively later.

AI tools can help draft and retrieve notes; they cannot supply accountability. I am interested in systems that attach decisions to recorded rationales so post-mortems are possible. That is part of why I care about systematic research infrastructure. Memory is a biased database.

There is an art to not over-writing. Twenty pages of decoration can hide the absence of a falsifiable core. I prefer one to two pages for most single-name or single-theme ideas, with links to deeper work. If you need a novel to justify a position, you may be constructing a cathedral around a hunch.

Review cadence closes the loop. Re-read theses on a schedule and after material events. Mark what you got right for the right reasons, right for the wrong reasons, and wrong. That practice is uncomfortable and compounding. Skip it, and experience becomes tenure without learning.

Write the thesis before you need the alibi. Future you is not as honest as present you hopes.

Position sizing should appear in the thesis, not only in a trading blotter. “Attractive” ideas still need a maximum loss you can absorb without cascading mistakes. Writing the size forces the uncomfortable arithmetic early.

Catalyst calendars help, with caution. Catalysts can be real and still already priced. Note what would constitute a true surprise versus noise. Otherwise every headline becomes a reason to fidget.

Post-mortems deserve the same template as entry theses. What was the claim? What happened? What will we change? Without symmetry, writing becomes a bureaucratic entry ritual and learning never exits.

For multi-strategy books, thesis discipline prevents silent style drift. If every idea quietly becomes a short-term momentum trade, you no longer have the portfolio you told stakeholders you run. Paper reveals drift before P&L mythology does.

I write because markets punish vague confidence. Clarity is not clairvoyance. It is respect for capital.

Shared templates reduce the cost of writing until the habit sticks. A one-page form with claim, drivers, risks, invalidation, size, and liquidity beats an essay contest. Style can vary; structure should not.

For personal accounts the same rule applies. Retail investors benefit from a notebook as much as institutions benefit from an IMS. The enemy is the same: stories that rewrite themselves after the fact.

When using AI assistants to polish notes, keep the human as author of the claim. Outsourcing the thesis is how you inherit fluent nonsense.

Time stamps on thesis revisions create a trail of how beliefs evolved. Without dates, a living document becomes a shapeless cloud. Dating changes is a small habit with large forensic value when performance reviews arrive.