How to Choose the Right Business Model for Your Business

Costco prices its registers to barely break even. HP's own CEO has gone on camera and called a printer sale a bad investment for HP itself, if the ink doesn't follow. Comcast spent thirty years wiring homes to sell television, in a business that isn't the one paying the bills anymore. Visa, the company you assume charges you every time you tap a card, doesn't lend you a cent and never touches the risk that you won't pay it back. Four companies, four industries, and not one of them makes its money where you'd point first.

That isn't four unrelated stories. It's one pattern wearing four costumes. Every business has a place where cash is designed to enter - and it is very often not the product on the label. Two companies can sell the exact same object and still run two different machines underneath it, depending only on where the price tag is actually doing its work. The method is simple to state and hard to run honestly on your own company: find the transaction that actually clears a profit, trace what protects it, and ask whether your own cost structure could survive charging there. Four repeatable patterns cover most of what you'll find in the wild. Below is one real company per pattern - not the whole story of any of them, just the single mechanism each one proves, and the one place each pattern is most likely to snap.

Pattern one: the membership margin: price the product to move it, and let the annual fee carry the entire profit

Costco's warehouses run on a markup capped so low it would be reckless for an ordinary retailer to try it - merchandise is priced to be obviously, provably cheap, not to earn a margin. The profit shows up somewhere else: the annual card every member pays simply to be allowed to shop there. That fee is close to pure margin, and it only works if members keep deciding, year after year, that it's worth paying again - which is exactly why Costco's own fiscal 2023 filings put the renewal rate at 92.7% in the U.S. and Canada and 90.4% worldwide.¹ Once the fee is carrying the business instead of the cart, the incentive structure flips: cheaper merchandise doesn't cut into profit, it protects the number the profit actually depends on, and the cheaper the merchandise gets, the harder the card is to justify not renewing. The break risk lives in the same place as the discipline. The model survives only as long as leadership keeps refusing the easy move - a fatter margin point here, a trimmed benefit there - that every quarter gives them a reason to make. The moment that discipline slips, the renewal rate is what erodes, and the renewal rate is the entire business - not a metric next to the real one.

Pattern two: razor-and-blades: sell the durable thing at a loss, and make the real margin on what it consumes

HP's printers are the clearest version of this pattern still running in public, because HP's own CEO has said what most companies leave unspoken. Speaking to CNBC in January 2024, Enrique Lores described a printer purchase itself as a bet on the customer's future ink spending - and, in his own words, a bad investment for HP if the customer doesn't print enough or doesn't use HP's supplies.² The printer is the razor, priced to win the install base. The ink is the blade, bought again and again, at a margin the hardware alone could never produce: HP's Printing segment ran at roughly a 17-20% operating margin across fiscal 2024, 19.6% in the fourth quarter.³

None of that margin survives if a customer can buy the same ink from somebody else, which is why the fiercest fighting in this pattern happens on the consumable, not the device. Lores was just as direct about the enforcement side: HP pushes firmware that can disable a cartridge it judges to infringe its IP, in his words, to 'stop the printer from working.'⁴ That firmware isn't a side quarrel with customers - it's the model's real cost structure showing its hand. A compatible, cheaper refill doesn't dent HP's revenue at the margin; it deletes the only half of the transaction that was ever profitable. Pick that lock - a good-enough generic cartridge, or a regulator who decides the firmware itself is anticompetitive - and razor-and-blades quietly turns back into what it looks like from the outside: a company selling hardware at a loss, for no reason at all.

Pattern three: the cross-subsidy: let one line fund another, and watch what happens when their roles quietly reverse

Comcast spent decades building a coaxial network to sell television, back when the network was a means and the channel lineup was the point. Somewhere in the last decade, the two swapped places. The wire Comcast built to deliver channels turned out to be an even better way to deliver the internet, and the internet turned out to be the thing people would pay for no matter what. By Q1 2026, Connectivity & Platforms - the segment broadband sits in - generated roughly 99.8% of Comcast's adjusted EBITDA, on close to 28.7 million domestic broadband customers.⁵ Video didn't disappear. It became the passenger. Comcast keeps selling it, at thin or negative margin, because a bundled customer churns less than a broadband-only one, and the broadband relationship is the one now worth defending. Notice that nobody planned this in advance - the cross-subsidy is the fingerprint of a center of gravity that already moved, discovered after the fact rather than designed on a whiteboard, which is exactly why it's worth learning to spot in your own numbers before a market forces you to notice it.

A subsidy like this only makes sense as long as the donor line stays strong, and both halves of this pattern are under pressure at the same time. On the funded side, cord-cutting is not a slowdown: major U.S. pay-TV providers lost roughly 5 million subscribers in 2023 alone, part of a slide from about 91.5 million subscribers in 2018 to about 71.3 million in 2023⁶, while the programming costs paid to networks and leagues keep climbing - the weak line gets squeezed from both directions at once. On the donor side, the very thing that makes broadband so profitable - a network that's already built, so one more customer costs almost nothing to serve - is what fiber and fixed-wireless rivals are now underpricing. A cross-subsidy survives exactly as long as the strong line stays strong. The day it doesn't, the weak line it was propping up gets cut loose fast.

Pattern four: the toll booth: collect a fee on volume passing through, and let someone else carry the risk

Visa is the pattern almost everyone mislabels, because it looks exactly like a bank and isn't one. Its own 10-K says so in language too plain to misread.⁷ Four parties move around your money on every purchase - you, your card-issuing bank, the merchant, and the merchant's bank - and Visa is none of them. It doesn't issue your card, doesn't set your credit line, and doesn't lose anything when you fail to pay your bill; the issuing bank carries all three. What Visa actually sells is the road: a network that lets an issuing bank and a merchant's bank clear a transaction in seconds, for a thin fee on the traffic passing through - not the loan, and not the risk.

That structure is what makes the toll-booth pattern worth studying on its own. Visa's network was built once, so an additional transaction costs it almost nothing to process, and a fee that's thin per swipe compounds, at that scale, into real money: $35.9 billion in net revenue and a 65.7% operating margin in fiscal 2024.⁸ That margin is the direct answer to whether your cost structure supports charging here - it does, because the cost of processing one more transaction barely moves. The honest risk is that a toll only holds as long as the road is the only way through, and it no longer is everywhere: government-built instant-payment systems like Brazil's Pix and India's UPI already settle bank-to-bank, with no network fee in the middle, across some of the largest consumer economies on earth. Same pattern, toll removed.

The product on the label is rarely the product that pays the bills. It's the thing that gets you close enough to charge for the one that does.

Which one do you actually have?: the four patterns, side by side, and the one question underneath all of them

Copying a competitor's product is easy, and mostly useless. Copying their entry point only works if your own costs can survive standing where theirs do - a near-break-even storefront, a subsidized device, a line you're willing to fund at a loss, a fee so thin it only works at enormous volume. None of these four companies chose their pattern because it looked clever on a slide; each one held its nerve on the losing half of the transaction long enough for the winning half to compound. Before you borrow anyone's business model, find their real entry point for cash, not their advertised product. Then ask the harder question, the one that actually decides whether the model is yours to run: where does the money actually get in, and can your cost structure support charging there?