Why
A yield is a transfer, not a property. Something that pays 12% is receiving 12% from somebody, and the entire question is who. This sounds obvious and is routinely skipped, because a number quoted as a percentage looks like a fact about an asset rather than a claim about a counterparty.
There are exactly three sources, and only one of them is revenue.
Fees are real income: users paid to do something and part of that payment reaches you. This is the only source that survives the protocol's token going to zero, and it is almost always the smallest of the three.
Emissions are dilution wearing a yield's clothes. The protocol prints its own token and hands it to you, so your percentage rises while every holder's share falls. It is not free, it is just paid by someone who is not in the transaction — and it is genuinely useful for bootstrapping, provided everyone reads it as marketing spend rather than as a business. The tell is simple: if the yield is quoted in a token the protocol issues, it is a marketing budget with a denominator.
Leverage is the one that fools careful people, because the arithmetic is invisible from outside. When a protocol pays incentives on both deposits and borrows, the same capital can loop: deposit, borrow against it, redeposit. Total deposits inflate, external capital does not. With outside money X and a collateral ratio c, total deposits are X/(1−c) and total borrows are cX/(1−c) — and the difference, net TVL, is exactly X. The headline number is a leverage multiple, not a size.
So the useful habit is one question asked before any other: who is paying this, and does the payment survive the token? Most advertised yields fail it. The ones that pass are usually much smaller and much more durable, which is the trade being obscured. gross-over-net is where the leverage arithmetic gets done properly, and risk-free-rate-is-the-floor is what any surviving number has to beat before it is interesting at all.
How it works
The three sources, and what each one survives
| Source | Who pays | Survives the token going to zero? | Typical size |
|---|---|---|---|
| Fees | Users, for a service they wanted | Yes — this is revenue | Smallest |
| Emissions | Every existing holder, by dilution | No | Largest, and quoted loudest |
| Leverage | Nobody — it is the same capital counted twice | No — it is not income at all | Invisible from outside |
The one-minute test
- What currency is the yield paid in? If it is the protocol's own token, at least part of it is emissions.
- What did a user pay to generate it? If you cannot name the service, there is no fee revenue underneath.
- Can the same deposit be borrowed against and redeposited? If yes, the TVL is gross and the yield is quoted on a number larger than the money present.
Fees over TVL is the clean version
Strip the token entirely and ask what fraction of deposited capital the protocol earns per year in fees alone. That single ratio is the incentive-corrected demand signal, and it is computable from public data for almost every protocol. A chain or protocol with a large TVL and a fee yield near zero is not being used; it is being rented, and the rent is paid in its own currency.
Where this shows up in the rest of the list
| Card | The same question, elsewhere |
|---|---|
gross-over-net |
The leverage arithmetic, with the identity worked out |
risk-free-rate-is-the-floor |
What the surviving yield has to beat |
monad-last-general-purpose-l1 |
A chain whose fee yield is roughly 0.12% against a $947M TVL |
arb-bots-are-the-peg |
A yield that is real fee income — funding paid between traders, not printed |