How wide is the spread?#
Through 2026, published figures for x402 activity included cumulative transaction counts from roughly 100 million to 200 million, monthly volume figures from a few thousand dollars to several billion, and 30-day totals quoted at both $24M and $1.6M.
These are not contradictions in the data. They are six methodological choices producing six different quantities, all described with the same words.
The six differences#
1. Evidence standard. Counting every gasless authorised transfer gives a much larger number than counting the ones a named facilitator relayed. The marker is shared with systems that have nothing to do with x402. What proves a transfer was an x402 payment covers why this difference is large.
2. Chain coverage. A Base-only tracker and a five-chain tracker are not measuring the same thing. Chain shares also moved a lot during 2026: Solana was reported at around 70% of transaction activity by mid-August while Base retained the cumulative lead, and weekly shares shifted repeatedly.
3. Self-dealing. Removing payer-equals-payee rows, and flows concentrated in one or two counterparties, cuts the figure substantially. The a16z and Allium analysis took about $3M to about $1.6M this way.
4. Transactions against dollars. These move independently and sometimes in opposite directions. Artemis's December 2025 estimate put the artificial share at 48% of transactions and 81% of dollar volume, which tells you a few large transfers dominated the dollar figure. A protocol with a few-cent median payment will always look better by count than by value.
5. Batched and deferred payments. A deferred scheme settles once for many calls. Counted on-chain that is one payment; counted at the application layer it is a thousand. Neither is wrong and they differ by three orders of magnitude.
6. Window. Cumulative since launch, trailing 30 days, and a single week produce very different numbers, and sources often omit which they used.
| Choice | Direction it moves the figure |
|---|---|
| Markers rather than facilitator-relayed | Much larger |
| More chains | Larger |
| Self-dealing left in | Larger, especially in dollars |
| Dollars rather than count | Noisier, dominated by few rows |
| Batched payments counted once | Much smaller |
| Cumulative rather than trailing | Larger |
How to read a figure you are handed#
Four questions. If a source does not answer all four, the figure is not usable for comparison.
- Which chains? Pooled multi-chain totals cannot be compared with single-chain ones, and the shares move.
- What window? Cumulative, 30 days, one week.
- What evidence standard? Every transfer with the marker, or only facilitator-relayed.
- What was excluded? Self-payments, concentrated flows, low-counterparty sellers, or nothing.
Note
This is why like-for-like comparison is the only comparison worth making. Roundhouse's Base figures for a 90-day window in 2026 landed within about 1% of Token Terminal's for the same chain and window. Against a blended multi-chain figure for the same period, the gap was enormous. The index did not change; the comparison did.
Why does nobody agree on a standard?#
Because the choices are genuine judgement calls rather than errors, and the right answer depends on the question.
If you are sizing the protocol's reach, counting every call including batched ones is right. If you are sizing the money moved, settled dollars net of self-dealing is right. If you are deciding whether to trust a seller, distinct counterparties matters more than either.
The failure is not that different methods exist. It is that the resulting figures are published with the same label and no method attached, so readers compare quantities that were never comparable.
What should a figure carry?#
Three things, always, and the absence of any of them is a signal.
A basis. What population was counted and what evidence standard it met.
A window. The exact period, and whether it is closed or runs to now.
A coverage statement. Which chains, and what is known to be missing. An index that cannot see a chain should say so rather than implying the record is complete.
A figure with those three can be checked and compared. One without them cannot be, no matter how precise it looks. Precision and accuracy are different things, and a number quoted to seven significant figures with no basis is less trustworthy than a range with one.
Frequently asked questions#
Which figure should I quote?#
The one whose basis matches your claim, stated with that basis. If you are arguing agents are transacting, quote facilitator-relayed settlements between distinct parties, per chain. If you are arguing the protocol is widely deployed, a broader count is defensible as long as you say what it includes.
Is a bigger number always the inflated one?#
Usually but not always. Batched payments make the on-chain count smaller than the real call count, so a settlement-based figure understates activity for sellers using deferred schemes.
Why do chain shares move so much?#
Because the base is small and concentrated. A handful of active sellers changing chain, or one campaign, moves a share by tens of points. Shares in a market this size are not stable properties.
Can I compare x402 to card volume?#
Not usefully. The units differ by orders of magnitude and the populations do not overlap. A comparison that puts a sub-cent API call next to a retail order is adding things that are not the same kind of thing.
Does any of this get easier?#
The measurement gets easier as facilitator registries get more complete, because the evidence standard tightens. The judgement calls do not go away, which is why publishing the method alongside the number is the durable fix rather than agreeing on one number.
Where to check this yourself#
/stats states the basis, the window and the chain for every figure it publishes, and breaks the record down by evidence grade so the verified band is visible separately from the total. Coverage and confidence documents what each grade means, and /transactions is the raw record if you would rather compute it yourself.