Costco

The thing to understand first is that Costco's merchandise is close to a break-even operation on purpose. The company caps its markup at 14% on branded goods and 15% on its own Kirkland Signature label, and it holds that line even when a higher price would clear. Gross margin on merchandise ran around 11% in FY2024, against roughly 25% at Walmart and 35% or more at a typical supermarket. Selling and administrative costs eat almost all of that, so the merchandise business, taken alone, earns a rounding error.

The money is in the fee. In the year to 1 September 2024 Costco collected about $4.8 billion in membership fees from roughly 76 million paid members, and that fee income was about half of operating income and about two-thirds of net income. The fee is nearly pure profit — there is no cost of goods attached to a plastic card — and around 90% of members renew every year. The whole strategy follows from that arithmetic: every choice about price, range and pay is a choice about whether a member will renew, not about what a single transaction earns.

The range is the second mechanism. A Costco warehouse carries roughly 4,000 items where a supermarket carries 30,000 and a Walmart supercentre well over 100,000. Fewer items means enormous volume per item, which means the best buying terms in retail, which means the cap on markup still leaves a price nobody else can match. It also means the goods arrive on pallets and sell off the pallet — no shelving, no stocking labour, no decoration. IKEA gets the customer to do the assembly; Costco gets the customer to do the walking and the carrying, and hands the saving back in price.

Costco's moat is scale economics turned into a promise. Volume per item is the largest in retail, which produces the best buying terms in retail, which lets Costco hold a price that a competitor with a normal range cannot reach. That much is copyable in principle — Sam's Club has been trying since 1983 — and yet the gap has not closed in forty years. The reason is the second layer: the markup cap converts the scale advantage into trust. A member does not check the price of a Costco item against Amazon because forty years of experience say it will be fine. That habit is worth more than any single price.

The third layer is the one most analysts skip, and it is where the McDonald's and IKEA cases connect. Costco pays hourly staff well above the retail market — a starting rate near $20 in 2024 and an average around $30, with healthcare for most part-timers. Turnover among staff past their first year runs in the single digits. Jim Sinegal's argument was that this is not charity; it is what lets a warehouse run 4,000 items at high volume with a fraction of the shrinkage and retraining cost of a supermarket. A rival cannot copy the pay without first copying the volume that funds it, and cannot get the volume without the trust. The loop is closed.

Where it is weaker: Costco has no real hold on the customer beyond the annual renewal decision, and the fee is the entire visible expression of its pricing power. A member who stops driving to a warehouse — because delivery got cheap enough — does not defect loudly; she simply does not renew. And the model needs the physical trip: the treasure-hunt aisle, the pallet economics, the petrol station. Every year that same-day delivery gets better is a year the trip has to justify itself again.

Costco's capital allocation is boring in the way the best records are. Nearly all internally generated cash goes into new warehouses on land the company usually buys outright, and the returns on those warehouses are consistent enough that nobody has had to invent a new strategy in forty years. There has been no material acquisition since the 1993 merger with Price Club, no diversification of any consequence, and no attempt to use the balance sheet for anything clever.

The interesting decisions are the ones about what not to keep. Costco has three times returned excess cash with a large special dividend — most recently $15 a share in January 2024 — rather than hoard it or lever up. And it has consistently declined to use the most obvious lever it has, which is to raise the markup. That refusal is a capital-allocation choice as much as a pricing one: it is management deciding, every year, that the renewal rate is worth more than the extra point of margin. It is the same logic that made Kroc's McDonald's a landlord and IKEA a designer of prices — find the part that compounds, and protect it from the part that merely earns.

The one real question is the pay model. Costco spends materially more on hourly wages and benefits than Walmart or Target, and has been criticised by analysts for it since at least 2004. The evidence so far — turnover, shrinkage, sales per square foot — supports the spend. But it is a choice that has never been tested in a serious downturn with a hostile shareholder, and the company's governance gives management no structural protection if one arrives.