Optimistic Rollups vs ZK Rollups: Onchain Traceability

Lower onchain data cost: Post all transaction data (calldata) to L1, leveraging Ethereum's security for ~$0.10-$0.50 per transaction (vs. L1's $5+). This provides full data availability for trustless verification. Matters for protocols requiring maximum transparency and easy data indexing, like DEX aggregators (e.g., Uniswap on Arbitrum) or NFT marketplaces.

Inherent traceability delay: Rely on a 7-day (Arbitrum) challenge period where anyone can submit a fraud proof. This creates a finality vs. traceability gap; state is provisionally final on L2 but not settled on L1. Matters for applications where capital efficiency across bridges is critical, as users/validators must wait for full settlement.

Instant L1 state finality: Post a validity proof (ZK-SNARK/STARK) with each batch, proving state transition correctness. No challenge period. Settlement is immediate (~10-30 min), enabling faster, safer bridge withdrawals. Matters for exchanges, payment networks (zkSync), and protocols where capital efficiency and security are paramount.

Higher proving overhead, but advanced data handling: Generating validity proofs is computationally intensive (specialized provers). However, they can use data compression (e.g., storing only state diffs) and recursive proofs (StarkNet). Some, like Scroll, also post full calldata. Matters for scaling high-throughput dApps where ultimate L1 finality speed justifies the proving cost.

Full transaction data is posted on-chain: All L2 transaction calldata is published to Ethereum L1 (e.g., via calldata or blobs). This provides a permanent, censorship-resistant record from which the entire state can be reconstructed. This matters for real-time analytics dashboards (like Dune Analytics, Nansen) and block explorers that need to index activity without delay.

Every state root is cryptographically verified: Validity proofs (ZK-SNARKs/STARKs) guarantee the correctness of state transitions before they are finalized on L1 (e.g., zkSync Era, Starknet). This provides the highest level of trust minimization for traceability, as the data's integrity is mathematically assured. This matters for financial institutions and bridges where the absence of a fraud window is critical.

Traceability is not final for ~1 week: State updates are considered 'assumed valid' until the challenge period (e.g., 7 days for Arbitrum, Optimism) expires. This creates a traceability lag for finality-dependent systems, as data analysts must account for the possibility of a reorg. This is a poor fit for real-time settlement reporting or systems requiring instant, irreversible proofs.

The proof is the trace: The actual execution trace is hidden within the zero-knowledge proof. While the output is verified, the step-by-step process is not human-inspectable on-chain. This matters for debugging failed transactions or auditing complex contract logic, where developers rely on tools like Tenderly or custom tracers that need full execution visibility.