What a transfer count counts

A count and a value on this site are now computed over the same rows. They were not always, and the history is worth stating because the difference used to be visible in the charts.

The count includes a second population since 25 September 2026. On the Brazilian settlement rail a payment is a mint into an address and a burn from it, with no transfer to count, so each mint into an address that never transacts with another is counted as one settlement, inside the figure for its token, currency and chain. It is not a transfer, and the word “transfer count” is kept because it is what the metric was and what every other row of it still is; the settlement layer on its own is published separately.

Nine (chain, token) pairs have a declared decimal exponent that disagrees with the one their ingestion applied. Until August 2026 those rows were kept in counts and dropped from values, on the reasoning that the transfer certainly happened even if its amount could not be trusted. That made a count and a value over the same window describe different populations, and it made Mexican peso value read as near-empty in a metric where the underlying data was sound.

The disagreement was then traced to token metadata blanket-claiming 18 decimals rather than to a wrong applied value, so the exclusion was withdrawn and the applied exponent is used throughout, with MXNB corrected by hand. Counts and values now share one basis, and a count and a value on this site are directly comparable.

The visible consequence: both begin 2021-05-12, where they used to begin two dates seven months apart.

A count still needs no exchange rate and no decimal exponent, which is what makes it the more robust of the two. Where a value figure carries a currency caveat, the count beside it usually does not.

Every count remains a floor. See “Why every count here is a floor”.