Staking Returns After Costs: Commission, Slashing and Exit Risk

The yield figure on a staking page is gross. What reaches you is that number minus the operator's commission, minus whatever the exit costs when you want your capital back, minus the small probability of a penalty. Three deductions, all published, none of them included in the headline. Compare providers on those and the ranking looks very different from a list sorted by advertised APR.

This guide is the reasoning behind our staking and liquid staking rubrics, written so you can apply it to any provider, including ones we have not covered.

What is the commission actually taken from?

Nearly every provider charges a percentage of rewards, not of your stake. Deposit and get 10% commission, and you keep 90% of what the network pays; your principal is untouched. That distinction matters because it makes the fee proportional rather than fixed, so it never erodes capital during a period of low rewards.

Around that shared model the variations are wide, and each one changes the number:

• What counts as a reward. Consensus rewards, transaction tips and MEV are three separate revenue streams. A provider taking its cut of all three is charging more than one taking the same percentage of consensus rewards alone, and the difference is invisible unless the fee page says which it applies to.
• Whether commissions stack. Some protocols state their fee is taken after the underlying validator commission, which means two layers are being deducted before the figure reaches you.
• Where the fee goes. A 10% cut split 5% to the protocol and 5% to node operators has different governance implications from 10% retained entirely by a treasury, or 8% to protocol with 2% into an insurance fund. The split tells you who is being paid to keep the thing running.
• Whether the percentage is published at all. Some providers state that a commission exists without giving the number, or leave it to be discovered in the interface. That is not a pricing difference, it is a disclosure difference, and we score it as one.

Here is what the published fee structures look like across providers in our ranking. The clustering around 10% is real — and it is precisely why the exit column, not the fee column, usually decides which one suits you.

As published on each provider's own pages, read 24 July 2026. Terms change; the provider's page is the authority and each figure is cited on the review it came from.

Why the exit is a cost, not a formality

Look down that right-hand column. The commissions differ by a factor of five between the cheapest and dearest; the exit terms differ by a factor of eighty — twelve hours at one provider, more than forty days at the tail of another.

An unbonding period is not an inconvenience, it is a position you cannot close. For the length of that window you hold an asset you cannot sell into a falling market, and the cost of that is exactly what an option to exit would be worth over the same period. Nobody prices it, because it does not appear on a fee page, but it is frequently the largest number in the comparison.

Three routes out exist, and providers offer different combinations:

1. The protocol queue. Free, and as slow as the network's exit queue is that day. The published figure is usually a minimum, not an expectation — read the tail, not the headline.
2. An instant exit for a fee. Typically a fraction of a percent to 1%, paid to whoever fronts the liquidity, and generally capped by how much of it is available. That cap is the part to check: an instant exit that works for small amounts and not for yours is not an exit.
3. Selling the receipt token on the open market. Always available, at a price the market sets. In calm conditions that price is near par. In a rush for the door it is not, which is the whole point of the third route existing.

The honest way to compare is to ask what leaving costs on the worst day rather than the median one. A provider with a 0.1% delayed unstake and a deep secondary market is offering something materially different from one with a free queue and no liquid market for its token, even where both advertise the same yield.

What does "non-custodial" mean in staking?

The word covers three arrangements that differ in who could take your assets, and the distinction is worth more attention than it usually gets.

• Delegation without transfer. On networks that support it, you assign your stake to a validator while the tokens stay under your key. The operator can affect your rewards and expose you to penalties; it cannot move your principal. This is the strongest arrangement available, and providers offering it tend to say so explicitly — "tokens stay in your wallet", "never takes custody".
• Deposit into a contract. Liquid staking works this way: you send tokens to a smart contract and receive a receipt token. Nobody holds your assets in the ordinary sense, but they are governed by code that someone may be able to upgrade, and your claim is now a token rather than the underlying.
• Custodial staking. A platform holds the assets and stakes on your behalf. Convenient, and a straightforward counterparty exposure — the risk is the platform, not the protocol.

The test to apply is simple: who could stop me withdrawing? Under delegation, nobody. Under a contract, whoever controls the contract, plus the exit queue. Under custody, the platform. A provider that states plainly which of the three it operates has answered the most important question on the page.

What kind of receipt token are you holding?

Liquid staking hands you a token representing the staked position, and the two designs in circulation behave differently enough to matter for accounting and for integrations.

Value-accruing tokens keep your balance fixed and increase in redemption value: hold one unit today and one unit next year, but each is worth more of the underlying. Every provider in the table above uses this model, and it is the friendlier of the two for tax records — no stream of tiny balance changes to reconcile — as well as for DeFi contracts that assume balances do not change on their own.

Rebasing tokens hold the redemption value near par and increase your balance instead. Simpler to read in a wallet, harder to account for, and a known source of integration bugs in protocols that were not written to expect a balance changing without a transfer.

Two practical consequences. First, the market price of a receipt token is not the peg — for a value-accruing token it drifts upwards by design, and comparing it to one is meaningless. Second, the token's usefulness elsewhere is part of what you are buying: a receipt accepted as collateral across major lending markets is a materially different asset from one that only its issuer supports, even where the staking terms are identical.

What does slashing actually cost, and who pays it?

Slashing is a protocol-level penalty for validator misbehaviour — signing two conflicting blocks, or being offline in the specific ways a network punishes. It is rare, it is real, and when delegating it lands on the delegator's stake, not on the operator's balance sheet.

Providers respond with insurance funds, coverage tiers, or a share of the commission set aside. Read those claims for three things: what the fund actually covers (operator fault only, or any loss), what its limits are, and whether the amount held is published. A provider naming a percentage of its fee that flows into an insurance fund has told you something checkable. One offering "coverage" with no stated limits has told you very little, and we score the two differently.

It is worth keeping this in proportion. Across the major networks, slashing losses are a rounding error next to the cost of picking a provider with a bad exit, or of leaving assets on a platform that fails. It belongs in the assessment, not at the top of it.

Who runs the validators, and can anyone?

The operator set is where a staking product's risk profile is genuinely decided, and it splits along one line: can anyone run a node, or is entry curated?

Permissioned sets are named — a handful of professional operators meeting performance and compliance standards. That produces reliability and accountability, and it concentrates the network's stake in a small group whose failures may correlate: shared clients, shared clouds, shared jurisdictions. Permissionless sets let anyone participate against a bond, which distributes the stake and accepts a wider spread of operator quality. Several providers run both, a curated pool alongside an open one, and split rewards differently between them.

Neither is the right answer in the abstract. What is not defensible is a provider that will not say which it operates: if the documentation does not describe the operator set at all, you cannot form a view on the single factor that most determines whether your stake is at risk from a correlated failure.

Working out the number that matters

To compare two providers honestly:

1. Start from the network's reward rate, not the provider's marketing figure.
2. Subtract the commission, checking which revenue streams it applies to and whether an underlying validator commission is deducted first.
3. Add the cost of the exit you would realistically use, spread over your expected holding period. A 1% instant exit on a six-month position is 2% annualised — usually larger than the difference between any two commissions in the table above.
4. Consider the tail: what happens if you need out on the worst day of the year, when the queue is long and the secondary market is thin.

For a holding period measured in years the commission dominates. For anything shorter, the exit dominates — and most people are shorter-term than they intend to be.

How our ranking scores this

The staking rubric scores five criteria at equal weight: fee transparency, custody and control, decentralisation, risk disclosure, and exit and liquidity. Every one is scored from what the provider publishes on its own pages, with the date recorded on the review. A provider quoting an APY without stating the cut behind it scores poorly regardless of how attractive the number is, because an unstated fee is not a low fee — it is an unknown one. The full rubric is in our ratings methodology.

What we cannot verify

Reward rates move with network conditions and validator performance, so no yield figure is stable enough to publish as a criterion; we score the terms around the yield instead. Where documentation is unreachable — one provider's docs sat behind a challenge page that would not resolve — we record the criterion as unverified rather than filling it from a secondary source. And exit times are what the provider states they are: we have not queued for a withdrawal ourselves, and a stated minimum is a claim about the best case, not a measurement of the worst one.

Sources

• Commission, exit, operator and insurance terms as published by each provider and cited individually on the reviews linked above, read 24 July 2026
• Our scoring rubric for these categories — Ratings Methodology