The Costco Business Model Fits on a Single Index Card

The Costco business model fits on one page of notes: ten headings, one or two sentences each, and next to every sentence a word saying where it came from. Run that page on Costco and it ends somewhere the shopping cart never suggests. In fiscal 2025 membership fees brought in $5.3bn, which is 2% of the company's revenue and more than half of its operating income. The rotisserie chicken is scenery. The business stands at the turnstile.

This page is not a summary of the annual report. It is a list of questions in a fixed order, and each answer is one you can either defend with a page from a filing or admit you do not have.

Ten headings and three words#

The company note template I use has ten headings, and the order is doing work: each one tests the one before it.

1. Money engine: one sentence, "the company earns when…".
2. Customer promise: what they pay for and what they would pick instead.
3. Revenue exposure: segments, geographies, concentration.
4. Demand direction: growing, flat, maturing or riding a cycle.
5. Pressure map: rivals, substitutes, customers, suppliers, regulator.
6. Moat claim: one sentence somebody could disprove.
7. Pricing evidence: what the customer did after the last increase.
8. Management: one promise, one delivery, one hard question.
9. Risks and counterargument: what would have to be true for this page to be wrong.
10. Monitoring questions: three, not thirty.

Next to every sentence goes one of three words: sourced, assumed or unclear. Sourced means you can point at the page. Assumed means you reasoned it from the mechanism. Unclear means you looked and did not find it, which is a legitimate result, because an empty box tells the truth and an invented number lies for years. If this sequence looks like a compressed version of qualitative analysis, that is because it is one: numbers arrive on the page as evidence for a sentence, never instead of one.

Where a note like this sits in the wider process, from first read to valuation, is the seven steps from a brokerage account to a first company.

How long does it take? The first pass on a company whose filings you have never opened runs an hour or two, half of it spent hunting for two numbers the highlights page does not carry. A year later the second pass takes twenty minutes, because you are comparing a page against a page rather than against a blank sheet.

I am using Costco because everybody knows the warehouse and almost nobody knows the income statement. It illustrates the method, not a pick.

The money engine of the Costco business model#

The sentence on the page reads like this: Costco earns when a member renews the card each year, and the merchandise moves at close to cost so that the renewal feels obvious. The company says something similar, only more carefully. Its fiscal 2025 annual report puts it as "the extent to which we achieve growth in our membership base, increase the penetration of Executive membership, and sustain high renewal rates materially influences our profitability". Sourced. Except that "materially" is an adverb, not a number, so the number has to be dug out.

The income statement for the year ended 31 August 2025 shows net sales of $269.9bn, membership fees of $5.3bn and operating income of $10.4bn. Which means the fees are 2% of revenue and 51% of operating income. Read nominally, that line is invisible. Read against profit, it is half the company, and only the second reading matters here, because an owner is paid out of income, not revenue.

The rest of the statement explains why. Merchandise costs were $239.9bn, so $30.0bn of gross margin survives, or 11.1 cents on every dollar of sales, against 10.9 cents a year earlier. Store, staff and administrative costs take $25.0bn of that, or 9.2 cents. Merchandise, on its own, hands back under 2 cents per dollar.

Delete the membership fee line tomorrow and operating income falls from $10.4bn to $5.1bn while not a single pallet looks different.

Those 2 cents are a decision rather than an accident, and the decision sets up the rest of the page. A retailer that wants to earn on the shelf pushes price up. A retailer that earns at the turnstile pushes it down, because every cent handed back comes home as a renewal. That is why Costco's gross margin means little read nominally or historically: 11 cents looks feeble beside any branded manufacturer, and it barely moves year to year. It starts meaning something read alongside the fee line, because the two numbers are opposite faces of one pricing decision.

The promise, and where the revenue really sits#

The second heading gets written in the language of need, not the language of the brand. A member pays before buying anything because they are handed a deal: we already negotiated the price for you, so you do not have to check three other shops. The alternatives are a hypermarket with no fee at the door and a delivery van. No filing contains that sentence, so on the page it carries the word "assumed", which is fine as long as you remember that heading seven will have to defend it.

The third heading has a source. At the end of fiscal 2025 Costco ran 914 warehouses, with 81.0 million paid members and 145.2 million cardholders. Executive members, who pay double and get 2% back on purchases, number 38.7 million, which is 48% of the paying base, and they account for 73.6% of worldwide net sales. So slightly under half the members generate close to three quarters of the sales, and that half is the customer whose departure hurts.

That ratio also rewrites the risk question. Because sales sit unevenly across 81 million cards, the page has to ask not how many members left but which ones, and the filing does not answer the second version.

What my page is missing: the company reports sales by category but not margin by the same split, so I do not know how much of those 11.1 cents comes from fuel and how much from fresh food. Unclear. The box stays empty.

Demand, pressure, and a claim you could break#

The fourth heading asks about demand direction and gets an answer dull enough to be true. Warehouse club retail in North America is a mature category that grows through warehouse count and member count rather than fashion. Merchandise sales grew 8% in 2025, from $249.6bn to $269.9bn, in the same income statement. Sourced. The counterargument is one you are not allowed to skip, because delivery takes away the visit itself, and the visit is part of the promise from heading two. Assumed.

The fifth heading is the pressure map, and most of its boxes can be closed at Costco in a single line. There are two rivals plus discounters who charge nothing at the door, and on the customer side there are 145 million cards and not one unit of bargaining power. Two pressures do the work. At $270bn of sales the leverage against suppliers sits with the buyer, and that leverage is what funds the low shelf price, which is the whole promise from heading two. Delivery is the more dangerous one, because it does not push the price up, it takes the visit the member already paid for out of that promise. Only the scale and the card count are sourced here. The rest is assumption.

Heading six is the moat claim, and this is where most notes collapse into the word "brand".

The pricing test, and what renewals did#

In a membership model the evidence of pricing power is not the shelf price. It is what the customer does after the fee goes up, and a fee increase comes with a date attached. Effective 1 September 2024 the basic card in the US and Canada costs $65 and the Executive tier $130 instead of $120, roughly 8% more, with the annual 2% reward cap lifted from $1,000 to $1,250.

The result sits in two consecutive annual filings. Fee revenue rose from $4.8bn to $5.3bn, or 10%, more than the increase itself, because new members joined too. Renewals went the other way: in the US and Canada from 92.9% at the end of 2024 to 92.3% a year later, and worldwide from 90.5% to 89.8%. Which means the year of the first full increase is also the year renewals in the largest market gave back 0.6 of a percentage point.

That the increase is what took those 0.6 points is the reader's inference, not the company's finding. Costco does not explain the decline. The filing offers only that "the timing of renewal after expiration is impacted by a variety of factors, such as warehouse openings and promotional activity". That is a list of candidate causes, not a cause. So both readings, mine and the company's, carry the same word on the page: assumed.

What is 0.6 of a point worth? Say, in round numbers, 80 million members at $65 a card. One point of renewals is 800,000 cards, or roughly $50m of fees a year, under 1% of operating income. On its own, close to nothing. The catch is that renewal is not a revenue line at all. It is a thermometer for the promise under heading two, and if it falls because members stopped believing the price, then $270bn of sales follows it down, slower and without an announcement.

Only now has the moat claim from heading six stopped being a nice sentence. The loop predicts something checkable: after a fee increase people stay, because the alternative is dearer even once you add the $65 back. Two filings say almost all of them stayed, though not all, and that fee revenue grew faster than renewals slipped. That is the whole result, and rounding it either way is how a note turns into a story.

I do not know whether those 0.6 points are promotional timing or the first sign of fee fatigue. The next two annual filings will settle it: back above 92.5% in the US and Canada and the technical explanation was right, below 92% and the moat claim under heading six has its first crack.

Management, risks, and an honest counterargument#

Heading eight sets one promise against one delivery. Management promises the member the lowest possible price, and in the year of the fee increase gross margin moved from 10.9 to 11.1 cents, a shift of 0.2 of a cent. The increase went into the fee rather than the shelf price, which is exactly how that promise should look in an income statement. On the earnings call I would ask for something the filing does not carry: if some renewals depend on promotional pricing, how much of a promotionally acquired card is still there in year two?

Heading nine stands before the conclusion, not after it. The heaviest risk is arithmetic, because 51% of operating income comes from one line a customer cancels with a click, with no court, no negotiation and no notice period. Close behind it is that the loop from heading six turns both ways, since fewer members means less scale, higher prices and fewer members still. The third carries the word "unclear": the filing does not split margin by category, so I cannot say which shelf funds which.

The counterargument to the whole page is a strong one, and it goes like this: the chicken is not scenery, because the chicken is what walks the customer to the turnstile, and without the shelves nobody pays $65. That is true and the page does not dispute it. Splitting the fee from the merchandise is not a claim that merchandise does not matter. It is how you see where the profit is made and which number needs watching. My own view is that Costco is one of the few retailers where a falling merchandise margin can be good news, because it is often paid for in renewals. That is an opinion, and it carries the word "assumed".

What the page deliberately leaves out#

There is no share price on it, no P/E, no valuation. Not because valuation does not matter, but because the page answers the earlier question: do I actually understand where the profit comes from and what holds it there. Valuing a company whose money engine you cannot state in one sentence produces a well-formatted spreadsheet with an arbitrary number at the bottom.

There is no verdict on it either. The tenth heading ends in questions with thresholds rather than a sentence starting "and therefore". I catch myself wanting to add a conclusion after an hour with a filing, because a note without a punchline feels unfinished. It is finished. The punchline of the page is the list of things to look for in the next filing.

One more limit, stated honestly: the page is only as good as the disclosure. Costco reports renewal rates, member counts and the Executive share of sales, so heading seven has something to work with. A company that publishes no number about customer behaviour leaves you nothing but "unclear" there, and then the honest output of an hour's work is to set it aside.

The same page on a bank asks differently#

Run the ten headings on a bank. Money engine: a bank earns on the gap between what it pays for deposits and what it charges for loans, so the customer does not pay at the turnstile, they pay by leaving money cheaply. Promise: safety and convenience. Pricing evidence: what depositors did when the bank cut its rate below the competition. Risks: credit quality, capital, the regulator. The same sequence produces questions about deposit costs and loan losses, and not one about renewals.

Heading seven changes most in that move, even though its wording stays put. At Costco the pricing evidence was customer behaviour after a fee increase, a number published once a year in a filing. At a bank it is how much deposit balance walked out after a rate cut, a number from a different table on a different rhythm, though the question underneath is identical: does the customer stay when it gets more expensive. The heading did not change. The place you look for the answer did.

At a pharma company you get patent expiry dates, trial candidates and reimbursement. So the template is not a form to fill in. It is a sequence that pushes whatever is load-bearing in a given company to the surface, and it is something different every time. If you want to see the same "who pays and what survives" question traced inside the income statement itself, follow one dollar through a company.

The page has one more virtue that is easy to forget: the date at the top. "As of September 2026" tells the future version of you that the conclusion was built on two annual filings, so when the third one lands the question is "what changed", not "who is this company". That third filing announces itself, because Monitoring checks the recorded conditions then, and it costs no Credits on any plan.

Three questions at the bottom of the page#

The tenth heading gets three questions, each with a threshold, and all three fall out of the nine above.

1. Renewals in the US and Canada in the next annual filing: above 92.5% or below 92%?
2. Merchandise gross margin: does it stay near 11 cents? If it climbs to 12 or 13, the company has started earning on the shelf, which is a change of model that nobody will announce.
3. The Executive share of sales: is 73.6% rising or standing still?

What you are doing by hand on one page, Taufolio does as a Full report: the same questions spread across 25 checks, each answer pinned to a page in a filing. The monthly credit allowance on the Free plan covers one such company a month without paying. Out of it comes an Investment thesis, the reason for holding broken into assumptions with measurable conditions, exactly like the three questions above. What such a report looks like inside is on the product page.

One closing rule: if your page on any company contains no "unclear", it is not a research note. It is a slide.