Notification campaign funnel: people reached, who bought, and the conversion
The funnel of the campaigns that only send a message: how many people the
message reached, and how many of them bought. It is the denominator no other
report has — order.report.sales says how much the message generated, not at
what cost of reach.
Only campaigns with type = notification. That is the attribution rule, not a
convenience: a monetary or point campaign issues a reward card and is attributed
by the redemption of that balance; the message it sends is a cashback notice,
not "the message that brought the customer". Without this cut, on 2026-08-14 one
organization had 24,035 message attributions coming from monetary campaigns
against 41 from notification campaigns — 99.8% of them undue.
⚠️ Two different dates, and the response keeps them apart instead of folding them into a single rate:
totalis the cohort: messages sent inside the window (notificationCard.createdAt), and how many of those have already converted, on any date. Conversion only means something this way — numerator and denominator are the same people. A recent window is immature: most of it is stillopen, and its conversion keeps rising afterwards.total.resultis what happened inside the window (attributionDate): the attributions and orders of the period. This is the block that lines up with the influenced revenue oforder.report.sales, which also filters by the sale date.
The attribution window of each message is
COALESCE(expiresAt, attributionAt, createdAt + 30 days). ⚠️ attributionAt is
not the date a purchase was attributed — it is written at the card's birth and
is deprecated; counting conversion by it would always give 100%. Conversion lives
only in the attribution table.
The three buckets converted / open / expired are exclusive and add up to
sent, so the screen can show where every message stands.
Try-it values populated from environment ACME (Demo)
Authentication
Token
Send your user token in the x-guper-authorization header. Endpoints that
support a Public mode (e.g. acúmulo and resgate previews) accept the same
request without the header — Private mode adds customer data and a
confirmation token to the response.
Headers
Path parameters
Query parameters
Responses
Notes
Why the conversion looks low
Conversion here is people who bought / people reached, and it is normally a fraction of a percent. That is the honest number: a base that buys anyway will produce plenty of purchases inside any 30-day window, and counting those as "influenced" measures coincidence, not influence.
open is the part that can still convert. In a window that ends today it
dominates — the conversion of a recent period is not final, and comparing a fresh
period against a closed one will always flatter the closed one.