Skip to content
Reading depth

Each view includes the earlier layers; the complete engineer or auditor page is shown by default.

Settlement records a state change, while finality describes the confidence that the accepted history will not later be replaced under the network's consensus rules.

Warning cross category cue

Settlement says a state change has been recorded; finality says when an application can rely on that history under a particular consensus model.

Financial actions build on earlier actions. An exchange may credit a deposit, a bridge may release value, and a lender may accept collateral only after deciding that the source event is stable enough. Treating submission, inclusion, execution success, confirmation, and finality as one moment can create assets or claims on one system that no longer have backing on another.

Final securities settlement is a helpful analogy because both systems need a point after which participants can rely on transfer. The analogy is limited: blockchain finality follows protocol-specific consensus, validator, timing, and recovery assumptions, whereas traditional finality also relies on operators, contracts, regulation, and courts. “Irreversible” is never permission to omit the assumptions that make reversal sufficiently unlikely or disallowed.

Transaction parties create the requested state change. Consensus participants order and finalize history. Applications choose the confirmation rule they will accept. Exchanges, bridges, custodians, and protocols may release destination assets based on a source receipt or state commitment. That downstream position becomes a claim on the correctness and persistence of the source event.

  1. A valid transaction is included and executes in a candidate history.
  2. Later blocks, votes, checkpoints, or proofs increase settlement confidence according to the chain’s consensus design.
  3. An application applies its own policy to the protocol signal and value at risk.
  4. Once the declared condition holds, the application recognizes the result or releases a dependent asset.
  5. Exceptional reorganization, halt, or recovery handling follows a separately specified reconciliation path.

Capital flow may wait until confirmation before assets are credited or released. Claim flow begins when another system relies on the source event. Information flow carries blocks, votes, proofs, and status. Control flow includes protocol governance and the application’s confirmation policy. Return flow is absent from finality itself; finality only changes when another economic return can be treated as settled. Risk flow runs from a replaced or stalled history into every dependent balance sheet.

Finality does not pay return. It changes settlement risk and capital timing. When one system releases value too early and the source history is replaced, the bridge, exchange, liquidity provider, reserve, insurer, or users holding the unbacked destination claim must absorb the mismatch.

A transaction can appear on screen before it is safe for another financial system to treat it as irreversible. Look for separate status and confirmation language instead of treating “complete” as a universal state.

Set policies using the chain’s real finality model, asset value, attack cost, liquidity, reorganization history, sequencer behavior, and downstream action. Document how deposits, withdrawals, liquidations, and cross-chain messages behave during a halt or reorganization.

Ethereum proof of stake uses attestations and checkpoint finality with protocol-specific supermajority conditions and penalties. That is an example, not a generic confirmation formula. Probabilistic chains, BFT-style chains, rollups, and centralized sequencers expose different signals and recovery paths; applications must integrate the one they actually depend on.

Model observed, included, successful, confirmed, and finalized states separately. Test reorganization and halt paths, ensure irreversible downstream release waits for the declared condition, and reconcile any derived claim when source history changes. Record whether the policy comes from protocol finality or an application-specific confirmation heuristic.

  • “Included means settled.” Inclusion is only one lifecycle step.
  • “More confirmations always mean finality.” The meaning of confirmations depends on the consensus model.
  • “Finality eliminates all reversal.” Protocol recovery, software faults, governance, and off-chain enforcement remain distinct boundaries.

Read blockchain and transactions and gas. Continue to bridges to see why two finality systems must be reconciled explicitly.

  • Ethereum.org, Proof-of-stake consensus — Ethereum-specific attestation and finality behavior (accessed 2026-08-09).
  • Ethereum.org, Transactions — Ethereum transaction inclusion and execution context (accessed 2026-08-09).
  • The supplied DeFi seed, sections 10, 12, and 14 — dependency, loss, and bridge-learning questions.

Machine-readable model

Key equations

Canonical expressions come from the structured concept record. KaTeX renders the notation, while the plain-text expression and variable table keep its meaning and units inspectable without JavaScript. Read the narrative above for the model's domain, assumptions, and rounding rules.

This concept does not require one canonical equation. Its mechanism and state transitions remain the authoritative explanation; do not invent a formula merely to make the topic look quantitative.

Assurance contract

Security properties

These structured statements define desired behavior. Their stable IDs can bind tests, invariants, specifications, audit findings, or proof results without turning descriptive review advice into an assurance claim.

Desired · not evaluated: No test, audit, or proof result is implied until scoped evidence is linked to this property.

  1. Applications distinguish observed, included, successfully executed, confirmed, and finalized states

  2. The confirmation rule is parameterized by the actual consensus model and value at risk

  3. A reorganization cannot silently leave downstream accounting inconsistent with the canonical source state

  4. Cross-system consumers delay irreversible release until their declared source-finality condition holds

Knowledge check

Quiz

Answer in your own words, then open the model answer.

Why might two applications wait different lengths of time?

Model answer

They may face different values at risk, downstream reversibility, consensus assumptions, and operational policies even when observing the same chain.