Ask a crypto forum how to escrow a deal and you will get three confident, contradictory answers: use a custodial service, use a smart contract, use 2-of-3 multisig. Each model has real strengths and a failure mode its fans rarely mention. Here is the honest comparison.
Custodial escrow: a company holds the coins
The classical model, and the one we run: a service takes custody of the buyer's deposit, verifies it, and releases it when the buyer confirms delivery or a dispute is decided. Strengths: simple for non-technical users, human arbitration built in, works identically across every coin, and mistakes can be caught by process (our operators verify every movement against written terms, on an append-only ledger). Failure mode: you must trust the operator, which is why transparency is everything. A custodial escrow without published fees, a written dispute policy and auditable records is just a stranger holding your money. Judge the operator, not the model: our controls are documented here.
Smart-contract escrow: code holds the coins
Escrow logic deployed on a blockchain, usually Ethereum: the buyer funds the contract, and code enforces the release rules. Strengths: the operator cannot quietly spend escrowed funds; anyone can read the contract. Failure modes: code cannot judge whether a domain transfer completed or delivered work matches a spec, so real disputes still route to a human arbitrator chosen by the platform, and you are back to trusting an operator, now with extra steps. Contract bugs are unrecoverable by design. Gas costs and wallet complexity exclude exactly the nervous first-time buyer who most needs escrow. It also ties you to one chain: a TRC20 USDT deal cannot use an Ethereum contract.
Multisig escrow: keys shared three ways
A 2-of-3 multisignature address where buyer, seller and an arbitrator each hold a key; any two can move the funds. Services like Bisq and Hodl Hodl built on this. Strengths: genuinely non-custodial, the arbitrator alone can never take the money, and a cooperative deal needs no third party at all. Failure modes: the tooling is unforgiving for non-technical users, a lost key or a botched signing bricks the deal, it is Bitcoin-centric in practice, and disputed cases still depend entirely on the quality of the arbitrator. Multisig is excellent between two technically competent traders; it is a poor fit for a marketplace buyer meeting crypto for the first time.
The honest bottom line
All three models converge on the same weak point: when a deal is disputed, a human decides. The model matters less than the quality and accountability of that human layer. Choose by user, not by ideology: technical Bitcoin traders can use multisig; DeFi-native counterparties may prefer a well-audited contract; for everyone else, a transparent custodial service with published rules, an immutable ledger and reachable staff is the protection that actually gets used. That is the service we built: see exactly how it works, or start with the basics at what is crypto escrow.