Our numbers, on the public record.

Commerce is released by completed UTC month, never as a live or rolling value.

A month-and-currency cell is published only when at least five distinct shops contributed during that exact closed interval. Current-month values remain pending, and suppressed months are never folded into a later total.

Completed-history and trailing-three-month totals are derived only from immutable monthly cells already present in the public payload.

This makes every aggregate change reproducible from previously disclosed intervals instead of exposing a hidden shop’s activity by subtraction.

Survivorship bias applies to reconstructed history.

Registered-account history is bucketed from creation timestamps for accounts that still exist today. Completed-month shops-with-sales values are reconstructed once from surviving paid orders, then frozen in the ledger. Owner-deleted accounts and their cascaded orders make early totals conservative.

The all-history shops-with-sales total is a monotonic lower bound of distinct shop accounts with at least one paid, non-cancelled order in any currency through a published closed-month boundary.

The boundary advances when the observed count is unchanged or has increased by at least five shops; smaller or net-negative changes retain the prior release. Owner-deleted accounts and privacy batching can make it conservative, and it does not imply that every counted storefront is still live or still selling.

The hash ledger proves consistency, not correctness.

It shows that no published row has been altered after the fact, because changing one breaks every link after it. It does not prove the figures were right when captured — we control the chain head. Anyone wanting a stronger guarantee should archive the JSON endpoint independently, which pins our numbers to a date we cannot revise.

We do not convert currencies.

Completed-month currency cells remain separate and are never folded into a blended total at an invented exchange rate.

Small cells are suppressed.

A numeric breakdown is published only when it and every implied complement contain either zero or at least 5 distinct shops. Geography omits an entire level if any stated or unstated partition is smaller than that threshold, so the omitted region cannot be identified.

Changing platform and live-storefront breakdowns are released as one atomic block only on a daily snapshot.

If any visible cell or implied complement changed by one to four shops, the prior block remains published until every changed value has accumulated a safe delta.

Plan buckets are complementarily suppressed as one group: if any free, paid, or lifetime bucket contains one to four live shops, all three bucket values are withheld so the public live-storefront total cannot be used to reconstruct the small bucket..

The monthly chart reports distinct shops with a paid, non-cancelled order across all currencies, plus a USD-currency paid-order count and USD gross.

Each closed-month series is independently qualified by at least five contributing shops. The latest completed-month paid-plan count is the subset of selling shops on a paid or lifetime Cottage CMS plan and is withheld when subtraction would expose fewer than five free-plan shops.

Gross merchandise value is rounded to the nearest $100 per qualified completed-month cell and includes offline payments sellers marked paid.

It measures commerce recorded by these shops, not money that flowed through Cottage CMS, and we take no percentage of it.

Location is self-reported at signup and never inferred from visitor IP addresses.

Sellers who did not state a region are counted in the national total and omitted from the regional breakdown.

A released commerce interval is immutable.

Later refunds, deletions, or status corrections do not revise its public value; any methodology change is introduced in a new version rather than rewriting the ledger.