What actually happens when you send crypto

Almost every expensive mistake made by someone new to crypto comes from carrying over an assumption from banking. A bank transfer is a request to an institution that can be queried, delayed, reversed or refunded. A crypto transfer is not a request at all. Following one from start to finish explains the difference better than any list of warnings.

Nothing is actually sent

There is no file, token or object that travels. What exists is a shared ledger, replicated across thousands of independent machines, recording which addresses control what. A transfer is an instruction to update that ledger, and the whole system exists to answer one question: is this instruction genuinely authorised by whoever controls the funds?

That authorisation comes from a private key. Your wallet holds the key; the address others see is derived from it. When you send, the wallet produces a cryptographic signature proving the instruction came from the key holder — without ever revealing the key. Anyone can verify the signature. Nobody can forge it, and nobody can reverse-engineer the key from it.

This is why the phrase “not your keys, not your coins” is a statement of mechanics rather than ideology. Whoever holds the key can produce valid signatures, and therefore controls the balance. Everything else is bookkeeping.

Broadcasting, and the auction you just entered

The signed transaction is broadcast to the network and sits in a waiting area of unconfirmed transactions. Validators or miners choose which to include in the next block, and they are not obliged to take yours.

They choose by fee. Every transaction carries a fee offer, block space is finite, and the highest offers get included first. That is the whole mechanism, and it explains behaviour that otherwise looks arbitrary: fees are not a published price, they are a live auction, so the same transfer can cost trivially little at one hour and a great deal at another. When the network is quiet, a low fee confirms quickly. When it is busy, an identical transaction can sit for hours.

Wallets estimate a fee for you, and their estimates vary in quality. Most let you set it manually, which is worth learning if you transact often. Set it too low and the transaction may sit indefinitely; some wallets let you replace it with a higher-fee version, and some do not.

Two details surprise people. On networks that run programs, a transaction that fails still consumes its fee — you paid for the attempt, not the outcome. And the fee is always paid in the network’s own asset, so you must hold some of it to move anything else. Running out of the native asset while holding plenty of a token is a common way to become temporarily stuck.

Confirmation is a probability, not a switch

Once included in a block, the transaction has one confirmation. Each subsequent block adds another. Exchanges typically wait for several before crediting a deposit, and the reason is not bureaucratic caution.

On a proof-of-work network, blocks are found probabilistically, and two valid blocks can occasionally be found at nearly the same moment. The chain briefly forks, then one branch wins and transactions in the losing branch return to the waiting pool. A transaction one block deep could still, in principle, be reorganised out. Six blocks deep, that becomes vanishingly unlikely. Confirmations are a measure of confidence, not a state change, which is why different platforms require different numbers for different amounts.

Nobody checks the destination

This is the part that has no banking equivalent, and it is where the losses happen.

The network verifies that the signature is valid and the balance is sufficient. It does not verify that the destination address is the one you intended. There is no name attached, no confirmation screen at the other end, and no institution to call. Send to a wrong address and the funds are gone — either sitting at an address nobody controls, or in the hands of whoever does.

Two failure modes dominate. Malware that silently swaps a copied address for an attacker’s is long-standing and effective, because nobody reads a long string carefully. And sending to an address on the wrong network — the same-looking address on a different chain — is increasingly common as assets exist in multiple places.

The defences are unglamorous and they work. Verify the address on a hardware device’s own screen rather than trusting the computer’s. Send a small test amount first for anything large or unfamiliar. Confirm the network as carefully as the address. And treat checking only the first and last few characters as insufficient, because attackers now generate addresses that match those.

There is one more wrinkle worth knowing: some networks do not track balances at all. Bitcoin and several others track discrete unspent outputs, so a wallet showing a single balance may actually hold dozens of separate pieces. Spending combines them and produces change back to an address you control. This is invisible in most wallets, but it explains why a transfer of a small amount can occasionally carry a surprisingly large fee — the transaction had to gather many small pieces, and fees scale with transaction size in bytes rather than with value moved.

Why an exchange transfer feels different

Moving funds between accounts on the same exchange is usually instant and free, and that is because nothing touched the blockchain. The exchange simply edited its own internal ledger. No transaction was broadcast, no fee was paid, and no confirmation was needed.

That convenience is exactly the trade described in custodial versus non-custodial holding. Internal transfers are fast because you are trusting a company’s bookkeeping rather than a network’s consensus. Both are legitimate choices; they are simply different systems with different failure modes.

What this explains

Most of what confuses newcomers follows directly from the mechanics. Fees fluctuate because they are bids. Transfers are irreversible because there is no institution to reverse them. Confirmations take time because certainty accumulates. Wrong addresses are unrecoverable because the network only checks authorisation, never intent. And a seed phrase is catastrophic to lose because it is the key, and the key is the only thing that controls the funds.

None of this makes crypto unusable. It makes it a different system with different assumptions, and the assumptions are worth learning before the first large transfer rather than after it. The glossary defines the vocabulary, and our security coverage goes through the specific defences in detail.