Acquisition Cohort Report
Pull installs, trials, ad spend, CPI, and install-cohorted revenue per campaign, ad set, ad, or country through the Superwall MCP or API.
The acquisition cohort report answers a single question: for the users acquired through a given campaign, ad set, ad, or country, what did they cost and what have they earned back so far?
It is available through the Superwall MCP — ask your agent for something like "show me installs and D30 revenue by campaign for the last 30 days" — and over the API at POST /v2/charts/acquisition-cohort-report.
What it returns
Each row is one value of the dimension you grouped by, plus a row for the organic/unattributed bucket. For every row you get:
| Column | Meaning |
|---|---|
installs | Installs in the window. For campaign, ad set, and ad, these come from MMP install matching; for country, from the SDK's install geo. |
trials | Trial starts, cohorted by install date. |
install_to_trial_rate | trials ÷ installs. |
spend | Ad-network spend. Campaign dimension only. |
cpi | Cost per install: spend ÷ installs. Campaign dimension only. |
revenue | Net proceeds per requested window, keyed by window — for example d4, d14, d30. |
Revenue is cohorted by install date, not by purchase date: the d30 figure for a campaign is what the users it acquired had earned you within 30 days of installing, however long ago that was.
Choosing your windows
By default the report returns d4, d14, and d30. You can request any combination of d1, d3, d4, d7, d14, d30, and d90.
Shorter windows tell you sooner whether a campaign is working; longer windows tell you more accurately. A campaign that looks weak at d4 may look very different at d30.
Reading zero versus null
The report distinguishes "we measured nothing" from "we could not measure":
- 0 means the source was available and genuinely saw nothing. A campaign with
installs: 0is one the MMP has no matched installs for, even though it has revenue attributed to it. - null means the source itself was unavailable, or a rate or cost had nothing to divide by. On the campaign, ad set, and ad dimensions,
installsis null when the report is not scoped to production only, because MMP installs are production-only. The country dimension counts installs from SDK data instead, so it keeps its install column on any environment filter.spendandcpiare null on every dimension except campaign, and null on all rows when the window contains no spend at all — so a report never claims your installs were free just because no ad account is connected.
Totals are blended
totals.cpi and totals.install_to_trial_rate divide by all installs in the window, organic included. They are blended figures and deliberately are not the sum of the per-row values: a paid campaign's own CPI is typically higher than the blended CPI, because the blend is spreading the same spend across organic installs too.
Things worth knowing
Ad spend has campaign granularity only. It carries no ad set id, ad id, or country, which is why spend and cpi appear on the campaign dimension and nowhere else, and why they need a connected ad account. See Meta Ads.
On ad dimensions, installs are windowed by MMP match time while trials and revenue are cohorted by SDK install date. At the edges of a window those two can disagree slightly, which occasionally pushes install_to_trial_rate above 1.
The report defaults to production only. Widening the environment filter to include sandbox makes the revenue side describe a different population than MMP installs can, so on the campaign, ad set, and ad dimensions the install column drops to null rather than comparing two different things. The country dimension is unaffected — its installs come from SDK data and follow the same environment filter as revenue.
How is this guide?