Coverage & benchmarks

Source

Solana RPC — the complete signature history of every known facilitator fee payer, walked transaction by transaction (43M+ signatures), plus every USDC transfer into a verified seller token account.

Ingest

index-settlement-tx → census_native.solana_facilitator_sig + census_native.raw_solana_bq_candidate → the settlement fact.

Verification

Every settlement row traces to a signature that can be re-fetched from any RPC node; a sample is re-fetched and re-decoded on every gate run.

• Replay ground-truth verify

  Take a random sample of the peer index's own published Solana rows, re-fetch each signature on-chain, and check that the settlement it describes is really there.

  via probe_solana_replay_verify.py over raw_settlement_solana_replay (354,994 rows)

  A 50-row sample proves the decode path, not the completeness of the 354,994-row replay set.

  50 / 50100.0%2026-08-07· measured

• Signature-walk verify

  Re-fetch sampled signatures from the fee-payer walk and confirm the walked record matches what the chain returns.

  via sig-walk verify over census_native.solana_facilitator_sig

  2 of 322 did not verify. They are counted here rather than dropped — the rate is 99.4%, not 100%.

  320 / 32299.4%2026-08-08· reported

Limit: The signature walk reaches back further than the settlement fact does. A facilitator's first measured settlement is a coverage lower bound, not the date it started operating.

Source

Base on-chain settlement events, indexed directly from the chain — no third-party aggregator sits between the chain and the record.

Ingest

index-settlement-tx → census.settlement_tx_fact → the published clean basis.

Verification

Each row keeps its transaction hash, so any figure on this site can be decomposed back to transactions a reader can open in a block explorer.

• Capture-class gate

  Only settlement classes that represent an actual x402 payment are admitted to the published basis; the rest are excluded upstream, in one place, and never re-filtered per surface.

  via int_settlement_daily_all (clean-class list from the canonical constants)

  This is a structural gate, not a sampled measurement — it has no match rate, which is why the rate column reads '—' and not '0%'.

  applied upstreamno rate — structural gate2026-08-08· reported

Limit: Base carries the largest raw volume in the census and also the largest measured share of self-trade. Read it next to the integrity view, never alone.