You see the word on-chain in pitches, dashboards and rulebooks. It sounds like a quality seal. It is not. On-chain and off-chain answer a plainer question: where is a record kept, and who can check it?
That question matters more than the label. A record on a public blockchain can be looked up by anyone. A record on a company server can be looked up by the company, and by you only when the company chooses to show it. What you care about as a trader can sit on either side, or on both.
This piece stays vendor-neutral until its last section. It explains what each term means, what each side gives you and costs you, how the two are combined, and how to ask which part of a claim you can actually verify.
Key Takeaways
- On-chain and off-chain describe where a record is kept and who can check it, not whether a system is good or bad.
- An on-chain record is public and hard to change afterwards, but it costs a fee, takes time to confirm and shows everything.
- An off-chain record is fast, cheap and can stay private, but you rely on whoever runs it, or on a proof you can check.
- In most designs the two are combined, so the useful question is which part is checkable and who confirms the rest.
- A checkable record shows that something was recorded, not that a claim built on it is complete or true.
What On-Chain and Off-Chain Actually Mean
A blockchain is a shared ledger. Independent computers each keep a copy, agree on which new entries to add, and link every batch of entries to the batch before it. Anything written into that ledger is on-chain: a transfer, a balance, the code of a smart contract, or a value that a contract stores.
Off-chain means everything else. A record kept in a company database, on a server, in a spreadsheet or inside an app is off-chain, even when the company itself works with blockchains. The record exists, but only the people with access to that system can read it directly.
The difference is not about how modern a system is. It is about who can check it. On-chain, you can look yourself. Off-chain, you look at what someone shows you.
What Lives On-Chain, and What That Buys You
Putting a record on-chain gives you a set of properties that come from the design of the ledger itself.
- Anyone can read it. On a public chain you need no account and no permission to look up an entry. You check the ledger, not a summary of it.
- It is hard to change afterwards. Once an entry is confirmed and newer entries are stacked on top, rewriting it would mean rewriting everything after it across the copies the network relies on.
- It carries its own order. Each entry sits in a sequence with a time attached. You can see what happened before and after, not only a final total.
Put together, these properties replace a promise with a lookup. A firm can still say anything it likes about a number. If the number sits on-chain, you do not have to take its word for the record itself.
Notice the limit in that last sentence. You check the record, not the story told about it. A transfer on-chain shows that funds moved from one address to another. It does not tell you who controls either address, why the transfer happened, or whether other transfers are missing from the picture.
What On-Chain Costs You
Those properties are not free. Each one has a price, and that price is the reason a design keeps parts of its work off-chain.
The first cost is the fee. Every write to a public ledger is paid for, and the amount depends on the chain and on demand at that moment. If you have not looked at this before, our explainer on what a gas fee is covers the mechanics. For a system that records a constant stream of small events, paying for every entry adds up fast.
The second cost is time. An entry is not final the moment you send it. The network has to include it and confirm it, and on some chains you wait for further confirmations before you treat it as settled. An off-chain system can answer in the time it takes a server to respond.
The third cost is privacy. A public ledger shows everything to everyone. Balances, transfers and the history of an address are open to anyone who looks, and an address can be linked to a person once it is used somewhere that knows who you are. For a trading firm, putting every order on a public chain would also show its positions to the whole market.
None of this makes on-chain worse. It makes it expensive in specific ways. Verifiability is something you pay for. A sensible design pays for it where checking matters and saves it where checking adds little.
What Lives Off-Chain, and What You Take on Trust
Off-chain is the natural home of fast software. An order book on a server, a risk engine, a user database and a ledger of internal balances can all run off-chain, update instantly, cost little per change and stay private.
The trade-off is trust. When a record lives only on a company's server, you see it through that company's interface. You can read the number. You cannot check it against anything the company does not control. If the database changes tomorrow, there is no outside copy that disagrees.
That does not make off-chain dishonest. It makes it unverifiable from the outside, unless the operator adds something you can check. That something can be a review by a third party, a published report or a cryptographic proof. Each of them moves part of the record from "trust us" towards "check this", and each covers only what it was built to cover.
So the useful question about an off-chain record is not whether it is true. From the outside, you usually cannot know. The question is who vouches for it, how often, and whether you can see their work.
Where the Two Meet: Hybrid Designs
In most designs, the two are combined. Each part of a system goes where its job fits: work that has to be fast or private stays off-chain, and records that have to be checkable go on-chain.
Take order matching. Pairing buy and sell orders has to happen fast, so in many designs the matching runs off-chain while the resulting trades are settled on-chain. Other designs run the whole order book on-chain and accept the cost. At the other end, a custodial exchange keeps both steps in its own books: it matches its customers' orders and updates their balances internally, so a trade between two of its customers leaves no on-chain entry at all. What reaches the chain are transfers into and out of the exchange's wallets. In each case, you can ask where the match happens and where the result is recorded.
Outside data is another case. A smart contract cannot look up a market price by itself. It needs a service, usually called an oracle, to bring that price on-chain. You can see which value the contract used. How that value was chosen happened off-chain, and for that part you rely on the oracle.
Reserves are a third case. A firm can show that it holds assets by pointing to addresses on a public chain. What it owes its customers sits in its own books, off-chain. Our article on what a proof of reserves does and does not show explains why that gap decides what the reserves figure means.
Heavy computation is the last case here. Some checks are too expensive to run on a public chain. They run off-chain, and only a result, or a proof that the result is correct, is written on-chain.
These examples carry the core point of this article. The label hides the real question. You are never really asking whether a system is on-chain. You are asking which part of it you can check yourself, and who confirms the rest.
| Part of the system | Where it usually sits | What you can check | Who confirms the rest |
|---|---|---|---|
| Order matching | off-chain or on-chain, by design | the settled trades, if they are recorded on-chain | the operator of the matching engine |
| Outside prices | brought on-chain by an oracle | the value the contract used | the oracle and its data sources |
| Reserves | assets on-chain, debts off-chain | the balances at the published addresses | whoever examines the books |
| Heavy computation | off-chain, with a proof on-chain | that the proof is valid | whoever supplies the inputs |
A Third Case: Off-Chain Work With an On-Chain Proof
The last row of the table deserves a closer look, because it changes what off-chain can mean. Instead of asking you to trust a calculation, a system can run it off-chain and publish a cryptographic proof that the calculation was carried out as specified. Anyone can then check that proof on-chain without redoing the work.
Different techniques can produce such a proof, from zero-knowledge proofs to signed results from protected hardware. They differ in how they work, but they share the same edge.
A valid proof shows that the calculation ran correctly on the inputs it was given. It does not show that the inputs were complete, current or honest. If a record is left out before the calculation starts, the proof still checks out.
So when you read that a result is "proven on-chain", split it in two. The computation is checkable. The data that went into it is confirmed by whoever supplied it.
How to Ask Which Part Is Checkable
You do not need to read code to apply any of this. You need a short set of questions, and the patience to ask them in writing before you rely on a claim.
- Is there a transaction hash or an address you can look up yourself, without going through the firm's own interface?
- Which part of the claim stays off-chain, such as a calculation, a balance in an internal ledger or a rule applied by staff?
- Who confirms that off-chain part, how often, and can you see their report?
- What does the on-chain part prove, and what does it leave out?
- If the record disappeared from the firm's website tomorrow, would an outside copy still exist?
The first question filters fast. If the answer is a screenshot rather than a hash or an address, you are looking at an off-chain record, however it is labelled. Once you have a hash or an address, our guide on how to read a block explorer shows you what to look for on the page.
The fourth question guards against a quiet overreach: treating a checkable record as if it proved a complete one. A reserves figure without the matching debts is a clear example, and the proof-of-reserves article linked above walks through it.
What This Looks Like at CarrotFunding
CarrotFunding is a hybrid too, and its rulebook describes both sides. That makes it a useful test case for the questions above.
Start with what the rulebook places on-chain. It states that Each challenge is minted as an NFT that stores its performance on-chain. For its verification layer, it describes the pattern from the section on off-chain work with an on-chain proof: ROFL calculations run off-chain and produce cryptographic proofs anyone can check on-chain.
Then the trading itself. The rulebook states: Your funded account is simulated, like every Carrot account. When you trade, CarrotFunding decides at its sole discretion whether to place a matching trade on Hyperliquid with its own vault capital (A-book), keep the risk internally (B-book), or do partly both. The rulebook labels the first route A-book · Carrot trades on-chain and the second B-book · no on-chain trade, and adds: Your P&L is identical whichever route a trade takes.
Read with the questions above, the split is visible. The challenge NFT, the proofs of the evaluation checks and the matching trades of the A-book are on-chain. The routing decision is made by the firm at its sole discretion, and B-book risk is kept internally. The transparency and on-chain proofs documentation lists the wallets and smart contracts behind the platform.
Conclusion
On-chain and off-chain are not grades. They answer a narrower question about where a record lives and who can read it. The decision in front of you is not which label a firm uses. It is whether the parts you depend on are the parts you can check. A firm that sounds fully on-chain in its marketing but keeps its rules in a private database leaves you trusting the part that matters. A firm that is openly hybrid, and tells you which side each part sits on, gives you something to verify. Ask for that split in writing before you pay, and treat any part nobody can show you as a promise, not a record.
FAQ
What is an off-chain transaction?
It is a transfer of value recorded somewhere other than the blockchain, so the chain never records the individual step. A trade between two customers of a custodial exchange, as described above, is one kind. A payment channel is another: two parties lock funds on-chain once, swap signed balance updates off-chain as often as they like, and write only the final balances back on-chain when they close the channel.
Does on-chain mean the data is correct?
No. On-chain means the data was recorded and can be checked against the ledger. If a value was wrong when it was written, the ledger keeps the wrong value as faithfully as a right one, and makes it hard to correct. A checkable record tells you what was written and when, not whether it was true.
Does on-chain mean decentralized?
Not necessarily. On-chain describes where a record sits. Decentralization describes who controls the system around it. A single company can deploy a smart contract, hold the keys that upgrade it and decide what gets written to it. The record is still public, but the power to change what happens next may sit with that company alone. Check both questions separately.
Can an off-chain record be moved on-chain later?
Part of it can. An operator can publish a fingerprint of a record, called a hash, on a public chain. Later, anyone holding the record can check that it still matches that fingerprint. That shows the record has not changed since it was anchored. It does not show that it was complete or correct at that moment, and the record itself can stay private.
If a firm says something is on-chain, how do I check it?
Ask for the exact transaction hash or address, then open it yourself in a block explorer rather than trusting a screenshot or an embedded widget. Compare what you see with the claim: amount, direction and time. If the firm cannot name a hash or an address, the claim is about an off-chain record, whatever the wording says.
Is "verified on-chain" the same as "audited"?
No. An audit is a review by a third party, carried out off-chain, that covers whatever scope the auditor agreed to. On-chain verification is a mechanical check of a specific record or proof against the ledger. The first relies on the reviewer's judgement, the second on the rules of the ledger. Neither covers what sits outside its scope, so ask what each one actually looked at.