The business model question is not where the money comes from. It is who sets the rate
American Express set its own fee in 1958 and was still collecting it in 2024. Booking pays Google, and its marketing bill rose to $8.2 billion while it was trying to escape. Fifty-seven money machines on record.
Two companies with the same revenue are not in the same business
Fifty-seven business-model decisions are on record and 47 were never reversed, the second-highest persistence here. A money machine is hard to undo once it runs, which is the argument for getting the shape right rather than the size.
The shape that matters is not the revenue line. It is which party decides how much comes in. Read the record for that and the set splits cleanly, and the split has nothing to do with margin, sector or era.
The ones that own the rate
American Express priced above the market in 1958 as a deliberate signal, and the premium fee model was still producing growing card-fee revenue through fiscal 2024. Sixty-six years, one pricing decision, and nobody outside the company has a say in it.
Bloomberg bought Merrill Lynch's stake back in 2008, retook the equity in full, and kept raising the terminal price into a locked-in network. Capital One built test-driven near-prime lending from 1988 and the information-based strategy scaled across cards, auto and eventually a payment network. Ferrari capped volume and monetized the overflow; the cap held and the margins stayed fat.
In each case the company holds the instrument that sets the rate — the fee, the price, the underwriting model, the scarcity. The rate is not negotiated annually with somebody bigger.
The ones that rent it
Booking Holdings decided in 2024 to convert customers it had paid Google to acquire into direct and loyalty relationships, on the sound reasoning that a customer paid for once can be re-acquired for free later. Direct-channel room nights reached the mid-50% range and grew faster than paid. Marketing spend climbed to $8.186 billion in 2025, rising rather than falling. The record's verdict is that the strategy stood only in part.
That is what escaping a rate-setter looks like when it is going well. The strategy is right, it is working, and the bill still goes up, because the part of the business that has not converted yet is priced by somebody else.
BuzzFeed raised SPAC capital in 2021 to scale a model whose traffic was priced by platforms it did not control; the capital that was supposed to save the model never arrived and revenue kept falling. 23andMe accepted a non-exclusive license in 2023 and the downgrade was followed by heavy losses, the closure of Therapeutics and Chapter 11.
And Ford split off Model e in 2022 to pre-build electric capacity against a two-million-unit target, because a legacy automaker that sat out the transition risked becoming the next Kodak. By 2025 that was more than $16 billion of cumulative EBIT losses and a $19.5 billion pre-tax restructuring charge. The rate there was set by adoption, and adoption is not a counterparty you can negotiate with.
How to run this decision
Drawn from what separated the records that stuck from the ones that were undone. A step nobody in the record can be seen taking is not here.
- 01
Write down who sets the rate, by name
For each revenue line: who decides what you charge, and who decides what you pay to acquire the customer. If the answer to either is a platform, a channel, a regulator or a single large buyer, that is the business you are actually in, whatever the deck says.
- 02
Separate the customer you paid for from the customer you own
Booking's distinction between paid-channel and direct room nights is the one that matters, and most companies do not report it internally. A customer acquired once and retained is an asset. A customer re-acquired every purchase is a cost line with a supplier attached.
- 03
Price the escape before you commit to it
Converting rented demand into owned demand is right and slow, and the bill rises during the transition because the unconverted part still has to be bought. Model the years where you are paying for both, and say out loud how many of them the business can absorb.
- 04
Check whether your rate-setter can change the rate unilaterally
A fee you set is different from a rate you negotiate, which is different again from a take rate somebody announces to you. The third kind is not a business model, it is a tenancy, and it should be planned with the same time horizon as one.
- 05
Find out what your model costs when the volume assumption is wrong
Ford's machine was designed for two million electric vehicles a year. The design was not wrong about direction; it was wrong about rate, and the fixed cost of that design is what produced the $19.5 billion charge. Ask what the model earns at half the volume, and whether it still functions at all.
- 06
Name the instrument you would have to give up to grow faster
23andMe gave up exclusivity. Coca-Cola gave away bottling rights. Every growth path has one of these, and it is usually invisible in the paper because it is a term rather than a number.
What it looks like when this is going wrong
- The acquisition cost is set by a party you do not negotiate with.Booking Holdings, 2024. BuzzFeed, 2021.
- Nobody separates customers you paid for from customers you own.Booking Holdings, 2024.
- The escape plan is working and the cost line is still rising.Booking Holdings, 2024.
- Capacity is being designed against an adoption curve rather than a contract.Ford, 2022.
- Exclusivity or a rate-setting term is being traded for speed.23andMe, 2023.