Lesson 9 of 9

The one-page investment memo

Step 3 of 6

Worked example

Costco, written 20 Nov 2026.

MEMO — Costco Wholesale, 20 Nov 2026, price $1,093.83 (19 Nov 2026 close), still including the $30 special dividend payable 4 Dec. Sources: FY2026 10-K (52 weeks to 30 Aug 2026); 8-K of 10 Nov 2026; Module 10 lesson 10's model.

Heading Content
DECISION Research more. My model can't explain the price, so I don't yet know what I'd be betting on.
THESIS Members keep renewing and paying fees, which fund low prices that bring in more members.
FACTS Revenue $303.2B (+10.1%). Operating margin 3.85%. Membership fees $5,907M, equal to 51% of operating income. 84.1M paid members (+3.8%); renewal 92.3% in the US and Canada, 89.8% worldwide. ROIC about 23% (Module 6 lesson 03).
VALUATION P/E 52.7 on diluted EPS of $20.76; FCF yield 1.9%, below the Treasury's 5.07%. My DCF: $426.47, 39% of the price ($396.60 after the dividend, 37% of $1,063.83): a gap of about 60%, against the template's own 15% (Module 10 lesson 07).
WHAT THE PRICE ASSUMES In my model: an operating margin of about 8.9% every year (more than twice today's), or a 5.7% cost of capital (a beta near 0.15), or 52 years of 9% growth. My opinion: it assumes much more than Costco's history, perhaps more fee income or growth for far longer. Untested.
VARIANT VIEW None yet. "Great company, loyal members" is what the market already believes, so there is nothing to act on.
CATALYSTS The first-quarter FY2027 10-Q (12 weeks to 22 Nov 2026): fee growth, two years after the September 2024 fee rise. The FY2027 10-K: renewal rates.
RISKS (1) At 52.7 times earnings, slower growth could hit the price hard even if the business does well. (2) The US and Canada: 86% of net sales. (3) Member trust; tariffs.
WHAT WOULD MAKE ME WRONG US and Canada renewal falls below 90% two years running; fee growth from new members and upgrades stalls once the fee rise has passed through; operating margin falls back below FY2024's 3.65%.
SIZE None: not acting. If I act: a 1% risk limit and a review point set in advance, under Module 16 lesson 05's example 8% cap.
OPEN QUESTIONS (1) In a ten-year model such as the Ginzu (Module 10 lesson 10), what growth, for how long, does $1,063.83 need at today's margin? (2) How fast do fees grow from new members and upgrades alone? (3) Is there evidence the margin can rise much above 3.85%?
REVIEW DATE 31 Jan 2027, after the first-quarter 10-Q; sooner if an 8-K changes a number above. Then act only if the answers give me a variant view; otherwise do nothing, or research more only with new questions. Log it in the journal (lesson 04).

What each of the three decisions needs

So far every worked decision is "research more". Here is what each one needs. Harbour Tools is invented: its numbers are no view on a real company.

Decision Example decision line The memo must also hold
Do nothing "Do nothing: the price needs a margin I find no evidence for." The specific reason, and what would reopen it (a fact, a number, a source)
Research more "Research more: the three questions below, by 31 Jan 2027." The open questions, a review date, and what evidence would move it
Act "Act: buy $2,000 of Harbour Tools, 5% of a $40,000 portfolio." SIZE from your risk limit and exit or review point (16-03), under your caps (16-05); the sell card (16-07); a journal entry and pre-mortem, done first (17-04, 17-05)
SIZE  1% of $40,000 = $400 at risk; buy $50, review point $40 (20% below)
      $400 ÷ ($50 − $40) = 40 shares = $2,000: 5%, under an 8% cap

"Research more" is honest only if it has an end: a date, and the evidence that would move it to one of the other two. Without both, it's "do nothing" that won't say so.

Keeping it up to date. Don't rewrite the memo when news arrives; add a dated update under it.

UPDATE 12 Mar 2027, Harbour Tools (invented)
Changed:   gross margin 31% last quarter, down from 34%
Touches:   WHAT WOULD MAKE ME WRONG ("below 32% two quarters running"): one so far
Decision:  unchanged, holding; check next quarter's report

Module 18 carries a full worked update on Nvidia.