A proof-of-reserves report is a photograph of a balance sheet at a block height, and the interval between photographs decides most of what it can tell you. A one-off snapshot proves a venue was solvent on the day someone asked. A monthly attestation proves it was solvent at each month-end, if the liabilities were compiled honestly. A daily Merkle root moves the evidence closer to real time, but only for the part of the balance sheet the venue chose to publish.

That makes cadence the first question to ask, ahead of the ratio. A 105% ratio from eighteen months ago describes a company that may no longer exist in that form. A 100.4% ratio published last week describes a venue that was backed at a known block, with a known list of gaps. The second is more useful.
- Frequency is a claim about control. Publishing daily or per block implies a process that runs continuously, which is a different operational commitment from commissioning an annual report.
- Assets are attested; liabilities are usually compiled. The gap between the two is where most of the residual risk sits.
- The issuer determines the assurance level. A self-published dashboard, an auditor’s agreed-upon procedures report and an on-chain Merkle root are not equivalent evidence.
- No cadence removes snapshot risk. More frequent photographs shorten the window in which assets can be moved, but they do not close it.
What a proof of reserves can and cannot prove
The construction is simple. The venue signs a message from a set of addresses, proving control at a block height, and separately reports the sum of customer liabilities from its internal ledger. If assets exceed liabilities, it publishes a coverage ratio above 100%. What varies is which addresses count, which entities own them, and whether anyone independent checked the liability figure.
A PoR is therefore evidence about a moment, not a guarantee about a period. It can show the claimed wallets held at least the claimed liabilities at one block. It cannot show the liabilities were complete, that the wallets belonged to the operating entity, or that nothing moved the next day.
The asset side is the easy half
Address control is verifiable. A signature over a message, or a transaction from the address, ties the key to the claim. Block explorers show the balance. The weaknesses are attribution and scope: whether the address really belongs to the exchange, whether it is a customer deposit wallet or a corporate treasury wallet, and whether the same asset is counted twice through a wrapped or bridged representation.
Double counting is the common error. A venue holding USDT on Ethereum and an equivalent wrapped balance on another chain can present both as reserves unless the report nets them. The address list has to be read as a portfolio, not a set of independent line items.
Liability proofs are the hard half
Customer balances live in the venue’s database, not on a chain. A Merkle tree can commit to them: each user’s balance becomes a leaf, the leaves are hashed into a root, and the root is published. A user can then check that their own leaf is included, which proves the venue did not quietly omit them from the total.
Inclusion is not completeness. A tree proves that the leaves inside it sum to the published total. It does not prove that every customer account was placed in the tree. Negative balances, margin accounts, internal market-maker accounts and balances held in a separate legal entity are the usual omissions, and none of them is visible from the root alone.
Three cadence tiers
Reports cluster into three schedules, and each corresponds to a different level of operational commitment. The tier tells you more about the venue than the coverage ratio does, because it reveals whether reserve reporting is an event or a process.
Tier 1: the one-off snapshot
This is the report commissioned after a crisis, or published to answer a specific question from depositors. It covers a single date, often chosen after the fact, and it is usually the most detailed document a venue produces. It is also the least useful for ongoing monitoring, because there is no commitment to repeat it.
Read it as a historical statement. It answers whether the venue was solvent on the snapshot date, subject to the scope note. It says nothing about the months before or after, and the absence of a second report is itself information.
Tier 2: monthly or quarterly attestation
The venue publishes on a fixed schedule, usually with an auditor’s report attached. The auditor performs agreed-upon procedures: confirm the address balances, recompute the ratio, check that the liability extract matches the ledger. The output is a statement of findings, not an opinion on solvency, unless the engagement was structured as a higher-assurance review.
Recurring reports usually take this form. It creates a cadence depositors can plan around and a paper trail regulators can inspect. Its weakness is the gap between period end and publication: a report dated 30 November may not appear until January.
Tier 3: real-time or daily Merkle root
A small number of venues publish a Merkle root daily, per block, or on a schedule close to continuous, sometimes anchored on-chain. Users can verify inclusion at any published root, and the sequence of roots makes it harder to present a single favourable snapshot.
The frequency is a real improvement, but it changes the failure mode rather than removing it. A daily root with a self-compiled liability extract is only as good as the extract, and most users never check their leaf, so the proof exists without being used.
| Cadence | Typical issuer | What is covered | Residual risk |
|---|---|---|---|
| One-off snapshot | Venue, sometimes with an auditor engaged for the date | Named addresses and a liability total at one block height | Stale immediately; scope chosen after the fact; no repeat commitment |
| Monthly or quarterly attestation | Venue plus auditor under agreed-upon procedures | Recurring address set and ledger extract, with a findings report | Publication lag; period-end window dressing; liability completeness outside the procedure |
| Real-time or daily Merkle root | Venue, sometimes with an independent verifier or on-chain anchor | Frequent roots and per-user inclusion proofs, usually for selected products | Liability extract self-compiled; partial product coverage; few users verify |
Cadence tells you how often the venue is willing to be checked. Coverage tells you what it is willing to be checked on. A daily report on one product line is not stronger evidence than a monthly report on the whole group.
What more frequent reporting does not fix
A shorter interval narrows the window for moving assets around a snapshot, but it does not eliminate the three structural gaps that matter most: incomplete liabilities, related-party balances, and the difference between the snapshot date and the report date.
Completeness of liabilities
The liability side is a database query, and the query is written by the party being audited. Accounts in a separate subsidiary, balances held in a lending product, and negative balances that offset customer deposits can all sit outside the extract. An auditor testing the extract will not find what was never included, unless the engagement explicitly tests completeness against the full account population.
The scope note therefore matters more than the ratio. A report covering spot accounts at one entity and excluding the earn product is accurate for that part and irrelevant to a depositor whose balance sits in the excluded product.
Related entities and affiliated funds
Reserves held by an affiliate, a market maker connected to the group, or a treasury company in another jurisdiction can be counted as assets while the corresponding liabilities sit elsewhere. The wallets may genuinely be controlled by the group, but control is not the same as legal availability to depositors in an insolvency.
The question to ask is which legal entity holds each wallet and which entity owes the customer balance. If the two differ, the report is a group-level statement and the depositor’s claim may be against a subsidiary with a different balance sheet. Which entity actually holds your account is a separate question from whether the group has enough assets somewhere.
Movement after the snapshot
Any point-in-time proof can be gamed by timing. Assets can be borrowed or moved onto the attested addresses before the block height and returned afterwards. Liabilities can be reduced by delaying withdrawals or by settling internal balances. A monthly cadence makes the game harder and more expensive, but it does not make it impossible.
Real-time roots reduce the window to the publication interval, which is a genuine improvement. They do not address an incomplete liability extract at every root, or addresses belonging to a different entity from the one owing the deposits.
Who signs the report matters as much as how often
Three issuer types appear in practice, and the differences are not cosmetic. A self-published dashboard is a management assertion with a signature. An auditor’s report is a statement about procedures performed. An on-chain verification is a cryptographic proof about a specific data set.
Self-published
The venue controls the addresses, the ledger extract and the presentation. This can still be useful, particularly when the report is frequent and the address list is complete, because the cost of falsifying a continuous series is higher than the cost of falsifying one snapshot. But nothing independent confirms the liability total.
Auditor attestation
Assurance engagements differ in what the auditor is allowed to say. Agreed-upon procedures produce findings without an opinion: the auditor confirms that the assets exist and the arithmetic is right, and explicitly disclaims a view on solvency or on the completeness of liabilities. A reasonable-assurance engagement goes further, at higher cost and lower frequency.
Reading the opinion paragraph is part of reading the report. A document that says the auditor “confirmed reserves” in a press release may contain a procedures report that says nothing about whether liabilities are complete.
Independent on-chain verification
A verifier can recompute the Merkle root from a published liability set, check the address signatures, and anchor the result on-chain. This removes the venue from part of the verification path, but it still trusts the input data. If the liability set is incomplete, a correct root commits to an incomplete set.
Five checks when you read a PoR report
The checks are ordered by how much they change the conclusion. A report that fails the first two does not need the rest.
- Find the scope note. List the entities, chains and products covered, then write down what is excluded. The exclusions are the part that matters to your balance.
- Note the date and block height, then compare with today. Calculate the age. A monthly report published six weeks after its date is describing a different company from the one you are depositing into.
- Check who compiled the liabilities. A venue-prepared extract with no auditor testing is a management claim. An agreed-upon procedures report is stronger but still limited to the procedures described.
- Look for a related-party note. Identify which wallets belong to affiliates or group treasury entities, and whether those assets are legally available to the entity that owes you.
- Verify your own inclusion if a Merkle proof is offered. A proof you have not checked is a claim. The exercise also shows you which products the tree actually covers.
Each step narrows the claim: scope says what is discussed, date says whether it is current, issuer says how far to trust the numbers, related parties say whether the assets are reachable, and inclusion says whether you are in the data set at all.
Cadence only helps if you know where the report lives. Binance keeps its reserves page and the auditor’s scope in one place, so the claim can be checked rather than taken on trust — the current offer card is below.
Welcome bonus
Up to $600
-
Code is applied at sign-up, not after -
Terms are stated before you deposit
Claim offer
Ends Nov 17, 2026
Matching cadence to exposure
A depositor with a small balance and a short horizon can treat a monthly attestation as adequate, because the cost of being wrong is bounded and the exit is quick. A depositor with a large balance or a long horizon needs a shorter interval and a broader scope, and should treat a one-off report as no ongoing assurance at all.
The same logic applies to the product. Spot balances are the easiest to attest and the most commonly covered. Earn products, margin accounts and lending books are where the liability side gets complicated, and a spot-only report is not a report on the group. The three numbers inside a single report — ratio, timestamp and scope — are the starting point; cadence is the fourth, telling you whether the other three are maintained.
There is a related concentration question that frequency cannot answer: whether the reserves themselves are made up of assets that would hold value under stress. A stablecoin float can be three balance sheets, and a reserve report that counts a large position in a single token inherits that token’s risk. Coverage of 100% in a depreciating asset is not coverage of the liability.
The practical conclusion is unglamorous. Ask for the scope, the date, the issuer and the schedule, in that order. A venue publishing monthly with a narrow scope is more checkable than one that published once with a broad claim, because the series can be compared against itself and against the chain.
0 comments
No comments yet — ask a question and we will answer it here.