Methodology

Method version organic-1.0 · Effective 2026-11-09 · Published before the first number. Every figure Orion Seer publishes carries the method version that produced it. A change to anything on this page is a new method version; the old version's series is left standing, never silently restated.

1. What the series counts

Scope of v1. Settlements are x402 payments settled in USDC on three chains — Base, Solana, and Polygon — observed as facilitator USDC transfers on the public ledger. Other assets and other chains are out of scope for organic-1.0 and are stated as out of scope wherever the series is shown, never silently omitted.

A settlement is one on-chain transfer in the store, keyed by chain, transaction hash, and log index. A multi-transfer transaction contributes one settlement per transfer log.

Two series, never blended. Every published figure exists in exactly two forms: raw (every settlement in scope that we observed — everything the facilitators settled) and organic (raw minus what the wash filter in section 3 excludes — what survives the wash filter). The two never share a row and never sum into a blended figure. Where a panel shows one, it names which.

Counting rules. Days are UTC calendar days, by the settlement's chain timestamp. Volume is the sum of settlement amounts; USDC atomic units convert at 10⁶ per dollar, converted on display, never rounded in storage. Distinct buyers per day are distinct payer wallets in the stored rows; distinct sellers are distinct payee addresses — wallet counts from the chain record, stated as wallet counts. A settlement whose facilitator cannot be determined rolls into a stated unattributed row, never spread across known facilitators. Zero is a measurement; null is unknown. A slice we did not observe renders as a gap, never as 0.

2. Where the data comes from

The source of record is the public ledger itself: facilitator USDC transfers recomputed from chain data, so no intermediary's terms can revoke the record. Published series from others (x402stats, CC BY 4.0; agenteconomy.to's open data.json) are cross-checks only: discrepancies against our series are logged and explained in the open. They are never republished as our series and never merged into it.

3. The wash filter, organic-1.0

The filter is a pure function over stored settlements. Each excluded settlement is excluded by exactly one rule, evaluated in this order, and the per-rule exclusion counts are published with the series. A settlement no rule touches is organic.

W0, zero-value transfers. Amount = 0. Not a payment; excluded from the organic series (it remains in raw counts of transfers observed).
W1, self-settlement. Payer address = payee address. A wallet paying itself is not demand.
W2, repeat-pair cycling. The same (payer, payee) pair settles the same amount 10 or more times within one UTC day. The first 9 stand; the 10th and later occurrences are excluded as cycling.
W3, mint-frenzy bursts. For one payee on one UTC day: if the payee receives 500 or more settlements that day and 90% or more of them are a single identical amount, the payee-day is a mint-frenzy pattern (the late-2025 PING pattern). All settlements to that payee that day at the modal amount are excluded; settlements at other amounts stand on their own rules.

What the filter deliberately does not do in v1. It does not cluster wallets by inferred common ownership, does not score counterparties, and does not exclude small payments for being small (the ecosystem's median genuine payment is under a cent; size is not evidence). Ownership inference, if it ever ships, ships as a labeled house indicator under its own method version, never folded silently into "organic."

4. Restatement and corrections

A methodology change creates a new method version. Derived rows keep the version that produced them; the public series shows the seam. A wrong evidence row is answered by a correction row, never an edit: evidence tables are append-only, and the audit trail is the product's claim to be checkable. Intraday figures are provisional, labeled provisional with their as-of time wherever they show.

5. Caveats printed with the series

Self-dealing across wallets overcounts distinct buyers. A sophisticated operator paying itself from many wallets looks like many buyers. We say this on our own series before anyone says it about us.
Wallet counts are not people counts. One person, many wallets; one custodial wallet, many people. The series counts wallets because wallets are what the chain proves.
Coverage is observational. We publish what we observed on the chains in scope. Settlements on other chains or in other assets exist and are not in these numbers.
Fixture labeling (PoC). During the proof-of-concept pass the store holds a synthetic 90-day fixture window at reduced scale, generated to census-shaped patterns and labeled FIXTURE in the store's ingest record and on every console panel. Fixture figures demonstrate the method; they are not measurements of the live economy and are never presented as such.

6. The caveat that governs everything

Orion Seer sells on the x402 rail only (a) data built on open, publicly verifiable evidence, or (b) its own proprietary indicators, labeled as such. Nothing in between is ever shipped as settled fact. This methodology is a prong-(a) instrument: every number it produces can be recomputed from public chain data and this page.