People keep asking sellers on x402 the same question: has anyone actually published a self-funded-versus-stranger-funded split. This page is ours. It exists so the numbers do not have to be re-typed into a comment every time the question comes round, and so anybody quoting them is quoting the same wording twice.
Every figure below names the instrument that produced it: the settlement ledger, the house buyer wallets, on-chain receipts, or the parallax census. All but the census measure us; the census supplies one house row (the wallet breadth) and one ecosystem-wide row, which is labelled as such. Nothing here is a projection.
The split
Rows are dated by when they were added to this page, not by when the measurement was taken. The measurement’s own window is its own column.
| Row added | Figure | Window | Instrument | Facts row |
|---|---|---|---|---|
| 2026-09-10 | Third-party revenue $0.211, 23 settlements, 4 distinct wallets, across the five properties | 2026-08-25 → 2026-08-30 | settlement ledger | F1 |
| 2026-09-10 | Each of those 4 wallets swept seventy-plus payTos (72–79) in one pass, bought a handful of calls from us, and none returned after 08-30 | 2026-08-25 → 2026-08-30 | parallax census (breadth); ledger (no return) | F2 |
| 2026-09-10 | Listing drills $0.269 settled from house wallets: toolshed 19/19 Bazaar rows, 10x402 8/8, kino402 dual-rail | 2026-09-02 | house buyer wallets | F3 |
| 2026-09-10 | penny402 launch: 10 listing drills ≈ $0.06 → 9 Bazaar rows live within minutes, first at 04:05:16Z | 2026-09-02 | house buyer wallet | F4 |
| 2026-09-10 | Break-even recomputed off on-chain receipts: no settlement fee observed → ≈34 paid calls/day (was 55) | 2026-09-02 | chain receipts | F5 |
| 2026-09-10 | House repeat-buyer rate: 0 distinct wallets paid twice, time-separated | to 2026-09-05 | settlement ledger | F16 |
| 2026-09-10 | Ecosystem base rate, n=317 newly listed sellers: 13% get a 2nd buyer by 9h, median 32h, 51% never | 2026-09-02 recompute | parallax census (ecosystem-wide, not a house number) | F13 |
The revenue figure is USDC. The rail per settlement is not pinned in the record, so this page does not name one.
Reading it
The two money figures are not comparable as a ratio, and pairing them as one number would be the easiest mistake to make here. The third-party revenue was settled between 2026-08-25 and 2026-08-30. The listing drills were settled on 2026-09-02, a few days later, and they are our own spending on our own listings. Give both dates whenever you cite them together.
The four wallets that paid us were catalog walkers. Each one swept seventy-plus payTos in a single pass and bought a handful of calls from us on the way through. That is breadth, not interest, and none of them came back after 08-30. The repeat-buyer count for the house is zero through 2026-09-05.
The ecosystem row is there for scale, not for comfort. It is a parallax census number across 317 newly listed sellers, not a house measurement, and it says that a majority of new sellers never see a second buyer at all. Our own zero sits inside a distribution, which makes it ordinary rather than excusable.
Break-even is the useful one for anybody deciding whether to build here: recomputed off on-chain receipts with no settlement fee observed, it is about 34 paid calls a day. That is the bar, and we are a long way under it.
How the house tags its own traffic
None of this split is possible without a convention for separating your
own payments from your customers’. Ours is a HOUSE_PAYERS list of
public payer addresses, with the flag derived once at verify time and
carried onto the ledger row and every analytics event. The full
convention, including the reference snippet, is
Step 2: Tag your own traffic before you report revenue.
If you are reporting gross settlements without that flag, your revenue number includes your own listing drills.
Refresh procedure
The figures on this page come from an internal house facts table (the
vault’s lemon/facts.md), which has no public URL. Rows are therefore
named by id and date rather than linked.
- A new row in the house facts table gets a new dated row on this page, added at the bottom of the table with the date it was added.
- An existing row on this page is never edited or removed. A figure that is superseded stays where it is; the newer figure sits below it with a later date, so the sequence of what the house believed and when stays readable.
- A figure that turns out to be wrong gets an entry at /corrections/ and a new row here. It does not get a silent edit.
- A figure not in the facts table does not go on this page, however well sourced. It is a proposal for a new row in the table first.
The same rule governs the corrections log, and for the same reason: a page that quietly improves its own past is not a receipt.
Rows cited on this page: F1, F2, F3, F4, F5, F13, F16 (house facts table, 2026-09-05 revision). F13 is ecosystem-wide; every other row is a house measurement.