Daily campaign report — 2026-08-04, against 2026-08-03

1Overview

Cohort value — ARPU and ARPPU at d0, d3 and d7

Keyed on the INSTALL day and measured from each user's own install stamp: dN is their first (N+1) x 24 h. ARPU is Adjust's own series, the one the partner's dashboard reports; ARPPU is our payment ledger, because Adjust exposes no cumulative distinct-payer count. A horizon the window has not lived through yet is left empty rather than written as zero.

Ignore the impression columns. One broken build's single hour is 85.4% of 08-03's impressions; §2's hour-by-hour read finds nothing.

Read within-day. Cohort at 24h is 35% covered against 08-03's 100% and is not compared here (Caveats). Within-day is sealed both days: 0.89 → 1.32.

The day gained on basket size. 39 within-day payers both days, payer rate 4.8% to 4.7%, ARPPU $10.20 to $15.09 — the whole 48% gain.

A third of booked is inherited. Earlier installs paid 36.1% of 08-04's booked revenue, the gap between booked 2.07 and within-day 1.32; §4 splits it.

2Hourly, 2026-08-04

Delivery columns are Meta's on both sides of each ratio. Payer counts and revenue are ours. Booked sits on the payment hour; within-day and cohort sit on the install hour.

Spend is close to flat across the clock, between $10.59 and $28.73, with no hour holding more than 6.4% of the day. CPI runs $0.348 to $1.036 with no clean diurnal shape on this clock: the four cheapest hours are 20, 21, 03 and 05. The revenue columns are not readable at this grain. Within-day payers run 0 to 3 per hour, so a single person moves an hour outright: hour 10 shows within-day ROAS 7.07 on two payers and hour 07 shows 5.11 on one.

The cohort column decays as the day goes on. Its coverage is 100% through hour 09, 96% at hour 10 and 0% from hour 11 onward, so every cohort figure after hour 10 counts only the payments already in hand and will rise.

3Country

4Aggregate, both days

Pooled over the 19 hours both days delivered in, on the UTC clock. Meta's instrument on both sides of every ratio.

• click → install 13.41% → 23.95% (+78.6%, p = 0.000) — rose.
• CTR 2.65% → 9.19% (+246.9%, p = 0.000) — rose.

Pooling is a mix comparison: each hour weighted by what it delivered. §2's hour count separates neither ratio; both p-values are hour 14 of 08-03.

Held to a single creative that ran on both days, on the ad account's own clock, only two tests clear p < 0.05:

• bodysuit click → install fell 36.33% to 25.55%, a 29.7% decline, p = 0.001. This is the one genuine negative in the comparison. Its CPI is flat across the two days, $0.511 to $0.506, so what moved is the share of clicks that became installs and not the price of an install. Coverage at H=6 is 72% on 08-04. bodysuit also went from $100 to $150 per day inside the compared window, so the decline and a budget change share it and cannot be separated here.
• spill click → install rose 2.61% to 26.42%, p = 0.000. This is the rejected build being replaced by its copy, already recorded as DECISION_LOG.md #50. It is not new information and it is the source of the pooled improvement above.

bikini (−10.2%, p = 0.224), pool5 (−11.9%, p = 0.240) and mugshot (−31.4%, p = 0.052) do not separate. No install → pay test on any single creative clears 0.05 on either day. These per-creative tests key on cells the estimator builds on Meta's clock, so their day boundaries are the account's and their totals do not reconcile against the tables above.

Meta's bracketed figures on both rows are quoted on its own conversion clock and are shown so the day is traceable to what Meta will report.

Booked revenue on each day, split by the day its payer installed:

The top row of each column is the within-day figure. That identity is why the gap between booked and within-day is exactly the inherited cohort and nothing else.

5Per ad set, hourly

One ad per ad set delivered on this day, so the ad-set row is the ad row, labelled by asset. §8 is the exception.

A single-clock delivery cut in six-hour UTC blocks. All columns are Meta's instrument. No revenue column appears here because per-asset per-hour payer counts are 0 to 2 and carry nothing.

spill2 holds the account's highest CTR in the 18–23 block at 15.37%, and the cheapest installs in both blocks where it ran with any volume, $0.341 and $0.370. mugshot runs 12 to 21 CPM against a pack of 9 to 15 and converts 11.6% to 20.0% of clicks against 19% to 34%, which keeps its CPI above every other arm in every block it ran. bodysuit, pool5 and mugshot each buy their cheapest installs in the 18–23 block, the ad account's late morning and afternoon; bikini is cheapest in 00–05 and spill2 in 12–17, so the diurnal pattern is not shared across the pack.

6Per ad set × country

⚠ Spend and installs are a META-day cut, as in §3; revenue is the UTC day. Cells above $5 of spend, ordered by spend. Revenue is within-day.

pool5 leads every revenue measure on a two-person cell. $93.94 of its $228.50 within-day revenue (41.1%) is one United States cell, $88.95 and $4.99: 94.7% of the cell and 38.9% of the asset's day is one person. Its India cell was $15.17 at wROAS 0.76.

That single payer is also 15.1% of the whole account's within-day revenue. Remove them and pool5 falls from within-day 2.21 to 1.35, below bodysuit's untouched 1.57. The ranking between the day's top two arms therefore turns on one person, and that is true whichever country they are in.

7Per ad set, aggregate, all three measures

The three tables below are the standing per-asset block, emitted by report_tables.py and pasted unedited. They carry the same columns in the same order in every report in this series, so any two days can be read against each other. Rows are every ad object with spend or installs on the day, including ones that took no spend and still attracted late-attributed installs.

A. Per asset — delivery

Meta's instrument on both sides of every ratio. Our registrations never enter this table.

B. Per asset — the three revenue measures

Ours on every column. Meta can express only booked, so it rides bracketed beside that column and never as a row. Payers are distinct people, never transactions. Cohort horizon H=24h, 35% covered on this day. Report clock UTC.

The 16-payer bar (#37): any row above whose payer count is under 16 is a direction and never a magnitude.

C. Per asset — country mix

⚠ This table is cut on the META day, while every other table on this page is on the UTC clock. Meta serves no hour × country grid, so per-country delivery exists only on the ad account's own UTC-7 day and cannot be restitched. It is a seven-hour-offset window against tables A and B above; read it for mix, never for level against them.

The last column is the share of that asset's own within-day revenue that came from the United States: where it far exceeds the US spend share, the return is a property of the geography the asset bought.

No arm reaches the 16-payer bar. Table B's best are pool5 at 12 and bodysuit at 11, so every revenue row is a direction.

Read the three tables together and the day splits cleanly. Table A says spill2 bought the cheapest installs on the account at $0.369 and the highest CTR at 13.34%, and that mugshot converted 14.99% of clicks against a 24–28% pack at roughly twice the price. Table B says pool5 leads the five worldwide arms on all three revenue measures and on both per-person measures, at a 6.2% payer rate and $19.04 ARPPU. Table C says 41.1% of pool5's revenue came from a country holding 13.5% of its spend, a gap of 28 points; bodysuit is the only other arm on the positive side of that gap, at eight points.

A further caution on table B's level: $57.72 of the day's $588.70 within-day revenue, 9.8%, came from two arms that spent nothing at all, on installs attributed late to builds that were switched off. The five worldwide arms that actually bought media returned $505.58 on $434.33 of spend, a within-day ROAS of 1.16 against the day's headline 1.32.