中文

Daily campaign report — 2026-09-04, against 2026-09-03

The portfolio spent $450.40 on 2026-09-04 against $518.72 the day before, and booked $3,829.23 against $3,501.13. Only Sinai was bought, for a twelfth day running: B1 took $349.24 against $337.39 and Dicey2 $101.16 against $87.90, both on the same creatives they ran the day before. HR2 registered 441 people across all twenty-four hours and none of its spend was measured, so the fall in the headline spend is a fall in coverage rather than in the buy. This is the day's 24 h publication and completes it. Definitions, the report clock and the three revenue measures are in the appendix.

Meta's own line stands in both columns of every Sinai table below and is struck in neither. The strike fires where Meta books more purchases in a day than Google Play processed for the whole app, and on both days its purchase count stayed under the day's transaction count, so its $3,023.60 and $2,327.86 sit beside our $2,939.79 and $2,744.80. The test is one-sided and an unstruck line is not a certified one: on Dicey2's own row Meta books $838.89 against the $99.51 in our records, 8.4 times it, and that excess cannot be split per account or per asset. The Meta bracket beside any asset's bROAS is Meta's attribution and never that asset's revenue. Cause and evidence: meta_conversion_overreport_2026-08-24.md.

1Overview

(i) Both apps, both days. Separate products sharing only the ad accounts; spend is additive, the two revenue columns come from two ledgers.

2026-09-032026-09-04change
spend$518.72$450.40−13.2%
impressions51,83237,338−28.0%
CPM$10.01$12.06+20.5%
registrations (ours)843896+6.3%
cost per registration$0.615$0.503−18.3%
booked$3,501.13 (6.75)$3,829.23 (8.50)+9.4%
Meta booked$3,023.60 (5.83)$2,327.86 (5.17)
within-day$729.28 (1.41)$721.67 (1.60)−1.0%
cohort @ 24h$793.18 (1.53)$815.24 (1.81)+2.8%
booked payers322347+25
payer rate (8h)8.4%10.4%+2.0 pts
ARPU (8h)$0.8614$0.7042−18.2%
ARPPU (8h)$10.23$6.79−33.7%
payer rate (d0)9.1%11.4%+2.2 pts
ARPU (d0)$0.9409$0.9099−3.3%
ARPPU (d0)$10.30$8.00−22.4%

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

This is the 24 h page, so every (d0) and cohort @ 24h cell is closed on both days. The horizon for 09-04 closed at 00:00 UTC on 2026-09-06 and the payment cut was taken at 00:19, so every (d0) cell is at 100% coverage; d3 and d7 close on 2026-09-09 and 2026-09-13, and this run stages no Adjust cohort block for them.

The 8 h page read the day's per-user fall at 18.2% and the closed horizon reads it at 3.3%, ARPU (d0) $0.9409 to $0.9099. Sinai's 09-03 cohort had paid 91.5% of its 24 h money inside 8 h and its 09-04 cohort 77.1%, so the morning comparison was reading how early 09-03 paid.

The measured buy fell 13.2% and the buy itself did not. $93.43 of 09-03's $518.72 is HR2's measured day and 09-04 carries no HR2 spend at all; the two accounts measured on both days went $425.29 to $450.40.

The payer rate rose 2.2 points at d0 and ARPPU fell 22.4%, 102 payers of 896 registrations against 77 of 843 and $10.30 to $8.00 a payer, which is where the smaller ARPU comes from.

B1 took 77.5% of what was measured against 65.0% and Dicey2 22.5% against 16.9%, the difference being HR2's 18.0% leaving the denominator; CPM rose 20.5% at the blend and 13.0% inside B1.

(ii) Every account, both days. Both apps summed, so the rows add to the portfolio line.

account2026-09-03 spend2026-09-04 spendchange2026-09-03 registrations2026-09-04 registrations
Portfolio$518.72$450.40−13.2%843896
B1$337.39$349.24+3.5%412343
Nomad Node$0.00$0.00—24
KBM1$0.00$0.00—01
Dicey2$87.90$101.16+15.1%88107
HR1$0.00$0.00—00
HR2$93.43$0.00−100.0%341441

HR2's $0.00 is an absence of measurement and not a zero: its hourly, day and country exports were all skipped from 09-04, while 441 registrations arrived from its campaign in every hour of the day against 341 in eighteen hours the day before.

NN1 exported at every grain and delivered nothing, which is a measured zero. NN2, KBM1 and HR1 were stopped by Meta on 08-23, 08-24 and 08-24 and are no longer exported at any grain, and HR2 left the pull on 09-04. All of them still book revenue through the payment ledger, which needs no export.

2The two apps, one tab each

Sinai is the only tab with delivery on this day: it took the whole $450.40 while no Vloom campaign in any account served an impression. Each tab carries that app's own accounts, hours, countries, campaigns and assets on its own ledger.

Showing

Sinai — every section below this line is Sinai's: the day, the hours, the countries, the campaigns and the assets, counted on Sinai's own ledger and its own accounts. Use the buttons above to switch.

Sinai's tab. Every section below is Sinai's: the day, the hours, the countries, the campaigns and the assets, on the sinai_lum ledger and the six accounts that buy it.

Sinai's measured spend was $450.40 against $518.72, and it booked $2,744.80 against $2,939.79.

The day, and every account

2026-09-032026-09-04change
spend$518.72$450.40−13.2%
impressions51,83237,338−28.0%
CPM$10.01$12.06+20.5%
registrations (ours)819876+7.0%
cost per registration$0.633$0.514−18.8%
booked$2,939.79 (5.67)$2,744.80 (6.09)−6.6%
Meta booked$3,023.60 (5.83)$2,327.86 (5.17)
within-day$729.28 (1.41)$711.05 (1.58)−2.5%
cohort @ 24h$793.18 (1.53)$804.62 (1.79)+1.4%
booked payers268288+20
payer rate (8h)8.7%10.4%+1.7 pts
payer rate (d0)9.4%11.4%+2.0 pts
ARPU (8h)$0.8866$0.7082−20.1%
ARPU (d0)$0.9685$0.9185−5.2%
ARPPU (8h)$10.23$6.82−33.3%
ARPPU (d0)$10.30$8.05−21.9%

2026-09-04 (the day) — delivery

accountspendshareimpressionsCPMregistrationscost/reg
Portfolio$450.40100%37,338$12.06876$0.514
B1$349.2478%31,863$10.96327$1.068
Nomad Node$0.000%0—0—
KBM1$0.000%0—1$0.000
Dicey2$101.1622%5,475$18.48107$0.945
HR1$0.000%0—0—
HR2$0.000%0—441$0.000

2026-09-04 (the day) — the three measures

accountbookedbROASwithin-daywROAScohort@24hcROASpayers (d0)payer% (d0)ARPU (d0)ARPPU (d0)
Portfolio$2,744.806.09$711.051.58$804.621.7910011.4%$0.9185$8.05
B1$1,534.714.39$315.930.90$357.781.023911.9%$1.0941$9.17
Nomad Node$0.00—$0.00—$0.00—0———
KBM1$100.55—$0.00—$0.00—00.0%$0.0000—
Dicey2$99.510.98$66.790.66$69.980.691211.2%$0.6540$5.83
HR1$185.91—$0.00—$0.00—0———
HR2$824.13—$328.33—$376.85—4911.1%$0.8545$7.69

Each account, 2026-09-03 against 2026-09-04, one table each.

B1

2026-09-032026-09-04change
spend$337.39$349.24+3.5%
impressions34,78031,863−8.4%
CPM$9.70$10.96+13.0%
registrations (ours)390327−16.2%
cost per registration$0.865$1.068+23.5%
booked$1,567.54 (4.65)$1,534.71 (4.39)−2.1%
Meta booked$1,727.84 (5.12)$1,488.97 (4.26)
within-day$199.11 (0.59)$315.93 (0.90)+58.7%
cohort @ 24h$242.91 (0.72)$357.78 (1.02)+47.3%
booked payers135152+17
payer rate (8h)7.2%11.3%+4.1 pts
payer rate (d0)7.7%11.9%+4.2 pts
ARPU (8h)$0.5197$0.8890+71.1%
ARPU (d0)$0.6228$1.0941+75.7%
ARPPU (8h)$7.24$7.86+8.5%
ARPPU (d0)$8.10$9.17+13.3%

Nomad Node — nothing delivered and nobody registered on either day.

KBM1

2026-09-032026-09-04change
spend$0.00$0.00—
impressions00—
CPM$0.00$0.00—
registrations (ours)01new
cost per registration$0.000$0.000—
booked$275.60$100.55−63.5%
Meta booked——
within-day$0.00$0.00—
cohort @ 24h$0.00$0.00—
booked payers1711−6
payer rate (8h)0.0%0.0%+0.0 pts
payer rate (d0)0.0%0.0%+0.0 pts
ARPU (8h)$0.0000$0.0000—
ARPU (d0)$0.0000$0.0000—
ARPPU (8h)$0.00$0.00—
ARPPU (d0)$0.00$0.00—

Dicey2

2026-09-032026-09-04change
spend$87.90$101.16+15.1%
impressions6,6605,475−17.8%
CPM$13.20$18.48+40.0%
registrations (ours)88107+21.6%
cost per registration$0.999$0.945−5.4%
booked$175.20 (1.99)$99.51 (0.98)−43.2%
Meta booked$437.28 (4.97)$838.89 (8.29)
within-day$146.76 (1.67)$66.79 (0.66)−54.5%
cohort @ 24h$159.72 (1.82)$69.98 (0.69)−56.2%
booked payers1318+5
payer rate (8h)11.4%11.2%−0.1 pts
payer rate (d0)12.5%11.2%−1.3 pts
ARPU (8h)$1.6677$0.5652−66.1%
ARPU (d0)$1.8150$0.6540−64.0%
ARPPU (8h)$14.68$5.04−65.7%
ARPPU (d0)$14.52$5.83−59.8%

HR1 — nothing delivered and nobody registered on either day.

HR2

2026-09-032026-09-04change
spend$93.43$0.00−100.0%
impressions10,3920−100.0%
CPM$8.99$0.00−100.0%
registrations (ours)341441+29.3%
cost per registration$0.274$0.000−100.0%
booked$856.30 (9.17)$824.13−3.8%
Meta booked$858.48 (9.19)—
within-day$383.41 (4.10)$328.33−14.4%
cohort @ 24h$390.55 (4.18)$376.85−3.5%
booked payers9699+3
payer rate (8h)9.7%9.5%−0.2 pts
payer rate (d0)10.6%11.1%+0.6 pts
ARPU (8h)$1.1046$0.6104−44.7%
ARPU (d0)$1.1453$0.8545−25.4%
ARPPU (8h)$11.41$6.41−43.8%
ARPPU (d0)$10.85$7.69−29.1%

B1 bought fewer and dearer registrations and each was worth more at d0: 327 against 390 at $1.068 against $0.865, payer rate 7.7% to 11.9% and ARPU (d0) $0.6228 to $1.0941 on 30 and 39 payers, its cohort return 0.72 to 1.02.

HR2 is 46.8% of the day's 24 h cohort revenue and none of its cost is in any denominator on this page, $376.85 on 441 registrations at an ARPU (d0) of $0.8545 against $1.1453, on 49 and 36 payers.

Held to the two accounts measured on both days, the 24 h cohort return is 0.95 on each: B1 and Dicey2 together earned $402.63 on $425.29 and $427.76 on $450.40, on 41 and 51 payers at d0.

Dicey2's cohort return went 1.82 to 0.69 on 12 payers at d0 and is not readable at this sample size, while its CPM rose 40.0%.

KBM1 and HR1 booked $100.55 and $185.91 with no delivery and one registration between them, from cohorts bought before Meta stopped them, so neither has a per-user row.

Hourly, 2026-09-04

(i) The portfolio, hour by hour, 2026-09-04. Every account placeable on the report clock, summed; delivery on the delivery hour, our columns on the hour the appendix names.

hrspendimpressionsCPMregistrationscost/regbookedbROASwithin-daywROAScohort@24hcROASbooked payers
00$12.94858$15.0818$0.719$96.697.47$2.980.23$2.980.239
01$12.581,294$9.7242$0.300$81.296.46$69.695.54$76.006.0413
02$18.301,813$10.0930$0.610$73.634.02$26.951.47$26.951.4712
03$20.662,195$9.4135$0.590$76.143.69$48.032.32$48.032.3218
04$21.812,289$9.5343$0.507$143.886.60$32.281.48$32.281.4818
05$21.941,986$11.0552$0.422$97.544.45$108.084.93$108.084.9316
06$19.731,793$11.0036$0.548$124.406.31$25.881.31$29.441.4917
07$15.101,509$10.0135$0.431$169.8211.25$23.811.58$23.811.5811
08$17.561,717$10.2331$0.566$96.695.51$0.000.00$0.000.0013
09$21.251,856$11.4527$0.787$106.865.03$29.011.37$29.011.3712
10$19.521,704$11.4635$0.558$53.682.75$9.910.51$9.910.5115
11$18.531,663$11.1428$0.662$129.566.99$6.870.37$6.870.3716
12$17.531,524$11.5039$0.449$82.944.73$38.032.17$61.403.5017
13$19.841,576$12.5929$0.684$37.881.91$39.081.97$42.072.128
14$19.771,576$12.5429$0.682$273.7613.85$6.150.31$6.150.3121
15$17.161,520$11.2942$0.409$59.303.46$38.452.24$38.452.2414
16$15.631,269$12.3237$0.422$130.188.33$35.812.29$50.993.2620
17$20.711,521$13.6239$0.531$195.669.45$21.641.04$37.421.8126
18$16.941,520$11.1444$0.385$127.277.51$54.423.21$61.153.6124
19$19.641,472$13.3455$0.357$131.056.67$21.251.08$24.651.2629
20$21.341,333$16.0139$0.547$160.667.53$33.321.56$33.321.5628
21$20.031,256$15.9541$0.489$132.166.60$17.240.86$24.351.2226
22$21.071,069$19.7139$0.540$56.262.67$13.080.62$19.010.9011
23$20.821,025$20.3131$0.672$107.505.16$9.070.44$12.270.5916

Spend ran $12.58 to $21.94 an hour with no gap, against $11.56 to $31.95 the day before; the flatter shape is B1's, which ran $9.13 to $17.58 and no longer climbs to one afternoon peak.

CPM climbed from $9.41 at 03:00 to $20.31 at 23:00, and every hour from 20:00 ran above $15.

Registrations ran 18 to 55 an hour, and 441 of the day's 876 are HR2's, 379 of them from one ad.

The 24 h cohort column is now filled and its payers run 0 to 8 an hour, so the hours carry shape and the ALL row carries the level.

2026-09-05T17:24:32.102129 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 3 6 9 12 15 18 21 hour (UTC) 0 5 10 15 20 USD Spend by hour 0 3 6 9 12 15 18 21 hour (UTC) 800 1000 1200 1400 1600 1800 2000 2200 impressions Impressions and installs by hour impressions installs 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 installs (Meta Leads) The day's motion — whole account, 2026-09-04 (UTC)
The day's motion: spend by hour, and impressions against installs
2026-09-05T17:24:32.244962 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 4 8 12 16 20 hour (UTC) 5 6 7 8 9 10 11 12 CTR % CTR 0 4 8 12 16 20 hour (UTC) 10 12 14 16 18 20 USD per 1,000 impressions CPM 0 4 8 12 16 20 hour (UTC) 10 12 14 16 18 20 22 24 installs / clicks % Click → install 0 4 8 12 16 20 hour (UTC) 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 USD Cost per install Delivery by hour — whole account, 2026-09-04 (UTC)
CTR, CPM, click-to-install and CPI by hour, whole account

(ii) The same 24 hours per account, 2026-09-04.

B1 — $349.24 on this day

hrspendimpressionsCPMregistrationscost/regbookedbROASwithin-daywROAScohort@24hcROASbooked payers
00$9.13739$12.3510$0.913$83.349.13$2.980.33$2.980.335
01$10.081,088$9.268$1.260$38.413.81$7.150.71$7.150.715
02$14.351,558$9.216$2.392$29.082.03$0.000.00$0.000.005
03$17.581,974$8.9119$0.925$50.202.86$19.981.14$19.981.1411
04$15.612,032$7.6817$0.918$58.823.77$0.000.00$0.000.009
05$17.371,696$10.2417$1.022$25.421.46$52.763.04$52.763.044
06$14.931,549$9.6417$0.878$76.535.13$10.310.69$10.310.698
07$11.811,295$9.1217$0.695$137.6111.65$12.991.10$12.991.108
08$14.701,464$10.0418$0.817$38.162.60$0.000.00$0.000.007
09$15.281,575$9.706$2.547$70.174.59$25.861.69$25.861.695
10$14.641,463$10.0112$1.220$17.191.17$6.920.47$6.920.475
11$14.221,427$9.9610$1.422$84.445.94$0.000.00$0.000.008
12$13.971,320$10.5817$0.822$57.364.11$19.591.40$36.162.5910
13$15.721,359$11.579$1.747$19.711.25$39.082.49$42.072.686
14$16.481,363$12.0912$1.373$120.837.33$2.990.18$2.990.1813
15$15.071,457$10.3419$0.793$39.542.62$18.511.23$18.511.238
16$14.511,225$11.8412$1.209$87.666.04$18.741.29$33.922.3411
17$15.191,158$13.1210$1.519$80.025.27$3.560.23$3.560.2311
18$11.421,156$9.8813$0.878$78.306.86$22.301.95$22.301.9513
19$14.531,159$12.5422$0.660$54.363.74$7.120.49$7.120.4910
20$15.541,064$14.6120$0.777$79.395.11$15.460.99$15.460.9915
21$15.971,062$15.0416$0.998$87.715.49$10.670.67$17.791.1114
22$14.77844$17.5015$0.985$29.922.03$13.080.89$13.080.898
23$16.37836$19.585$3.274$90.535.53$5.880.36$5.880.3611

Nomad Node — $0.00 on this day

hrspendimpressionsCPMregistrationscost/regbookedbROASwithin-daywROAScohort@24hcROASbooked payers
00$0.000—0—$0.00—$0.00—$0.00—0
01$0.000—0—$0.00—$0.00—$0.00—0
02$0.000—0—$0.00—$0.00—$0.00—0
03$0.000—0—$0.00—$0.00—$0.00—0
04$0.000—0—$0.00—$0.00—$0.00—0
05$0.000—0—$0.00—$0.00—$0.00—0
06$0.000—0—$0.00—$0.00—$0.00—0
07$0.000—0—$0.00—$0.00—$0.00—0
08$0.000—0—$0.00—$0.00—$0.00—0
09$0.000—0—$0.00—$0.00—$0.00—0
10$0.000—0—$0.00—$0.00—$0.00—0
11$0.000—0—$0.00—$0.00—$0.00—0
12$0.000—0—$0.00—$0.00—$0.00—0
13$0.000—0—$0.00—$0.00—$0.00—0
14$0.000—0—$0.00—$0.00—$0.00—0
15$0.000—0—$0.00—$0.00—$0.00—0
16$0.000—0—$0.00—$0.00—$0.00—0
17$0.000—0—$0.00—$0.00—$0.00—0
18$0.000—0—$0.00—$0.00—$0.00—0
19$0.000—0—$0.00—$0.00—$0.00—0
20$0.000—0—$0.00—$0.00—$0.00—0
21$0.000—0—$0.00—$0.00—$0.00—0
22$0.000—0—$0.00—$0.00—$0.00—0
23$0.000—0—$0.00—$0.00—$0.00—0

KBM1 — $0.00 on this day

hrspendimpressionsCPMregistrationscost/regbookedbROASwithin-daywROAScohort@24hcROASbooked payers
00$0.000—0—$0.00—$0.00—$0.00—0
01$0.000—0—$0.00—$0.00—$0.00—0
02$0.000—0—$0.00—$0.00—$0.00—0
03$0.000—0—$3.16—$0.00—$0.00—1
04$0.000—0—$0.00—$0.00—$0.00—0
05$0.000—0—$6.58—$0.00—$0.00—1
06$0.000—0—$3.39—$0.00—$0.00—1
07$0.000—0—$0.00—$0.00—$0.00—0
08$0.000—0—$19.66—$0.00—$0.00—1
09$0.000—0—$16.10—$0.00—$0.00—1
10$0.000—0—$13.96—$0.00—$0.00—4
11$0.000—0—$0.00—$0.00—$0.00—0
12$0.000—0—$3.28—$0.00—$0.00—1
13$0.000—0—$0.00—$0.00—$0.00—0
14$0.000—0—$3.35—$0.00—$0.00—1
15$0.000—0—$0.00—$0.00—$0.00—0
16$0.000—0—$3.16—$0.00—$0.00—1
17$0.000—0—$14.04—$0.00—$0.00—1
18$0.000—1$0.000$0.00—$0.00—$0.00—0
19$0.000—0—$3.35—$0.00—$0.00—1
20$0.000—0—$0.00—$0.00—$0.00—0
21$0.000—0—$3.48—$0.00—$0.00—1
22$0.000—0—$0.00—$0.00—$0.00—0
23$0.000—0—$7.04—$0.00—$0.00—2

Dicey2 — $101.16 on this day

hrspendimpressionsCPMregistrationscost/regbookedbROASwithin-daywROAScohort@24hcROASbooked payers
00$3.81119$32.026$0.635$0.000.00$0.000.00$0.000.000
01$2.50206$12.142$1.250$0.000.00$0.000.00$0.000.000
02$3.95255$15.497$0.564$3.160.80$9.472.40$9.472.401
03$3.08221$13.940—$6.312.05$0.000.00$0.000.002
04$6.20257$24.127$0.886$0.000.00$3.350.54$3.350.540
05$4.57290$15.767$0.653$10.082.21$9.892.16$9.892.163
06$4.80244$19.673$1.600$2.940.61$0.000.00$0.000.001
07$3.29214$15.374$0.822$0.000.00$0.000.00$0.000.000
08$2.86253$11.304$0.715$3.561.24$0.000.00$0.000.001
09$5.97281$21.255$1.194$3.560.60$0.000.00$0.000.001
10$4.88241$20.252$2.440$0.000.00$0.000.00$0.000.000
11$4.31236$18.261$4.310$0.000.00$0.000.00$0.000.000
12$3.56204$17.454$0.890$0.000.00$0.000.00$0.000.000
13$4.12217$18.993$1.373$0.000.00$0.000.00$0.000.000
14$3.29213$15.454$0.823$0.000.00$3.160.96$3.160.960
15$2.0963$33.172$1.045$0.000.00$0.000.00$0.000.000
16$1.1244$25.451$1.120$0.000.00$0.000.00$0.000.000
17$5.52363$15.217$0.789$15.952.89$8.951.62$8.951.622
18$5.52364$15.166$0.920$9.111.65$0.000.00$0.000.003
19$5.11313$16.337$0.730$10.392.03$7.341.44$7.341.443
20$5.80269$21.568$0.725$10.711.85$17.863.08$17.863.082
21$4.06194$20.935$0.812$17.394.28$3.580.88$3.580.884
22$6.30225$28.007$0.900$0.000.00$0.000.00$0.000.000
23$4.45189$23.545$0.890$6.351.43$3.200.72$6.391.442

HR1 — $0.00 on this day

hrspendimpressionsCPMregistrationscost/regbookedbROASwithin-daywROAScohort@24hcROASbooked payers
00$0.000—0—$0.00—$0.00—$0.00—0
01$0.000—0—$0.00—$0.00—$0.00—0
02$0.000—0—$0.00—$0.00—$0.00—0
03$0.000—0—$0.00—$0.00—$0.00—0
04$0.000—0—$45.50—$0.00—$0.00—1
05$0.000—0—$0.00—$0.00—$0.00—0
06$0.000—0—$0.00—$0.00—$0.00—0
07$0.000—0—$29.22—$0.00—$0.00—2
08$0.000—0—$28.06—$0.00—$0.00—2
09$0.000—0—$0.00—$0.00—$0.00—0
10$0.000—0—$2.99—$0.00—$0.00—1
11$0.000—0—$0.00—$0.00—$0.00—0
12$0.000—0—$0.00—$0.00—$0.00—0
13$0.000—0—$14.61—$0.00—$0.00—1
14$0.000—0—$0.00—$0.00—$0.00—0
15$0.000—0—$0.00—$0.00—$0.00—0
16$0.000—0—$0.00—$0.00—$0.00—0
17$0.000—0—$28.65—$0.00—$0.00—1
18$0.000—0—$0.00—$0.00—$0.00—0
19$0.000—0—$13.66—$0.00—$0.00—2
20$0.000—0—$23.22—$0.00—$0.00—2
21$0.000—0—$0.00—$0.00—$0.00—0
22$0.000—0—$0.00—$0.00—$0.00—0
23$0.000—0—$0.00—$0.00—$0.00—0

HR2 — $0.00 on this day

hrspendimpressionsCPMregistrationscost/regbookedbROASwithin-daywROAScohort@24hcROASbooked payers
00$0.000—2$0.000$13.35—$0.00—$0.00—4
01$0.000—32$0.000$42.88—$62.54—$68.85—8
02$0.000—17$0.000$41.39—$17.48—$17.48—6
03$0.000—16$0.000$16.48—$28.05—$28.05—4
04$0.000—19$0.000$39.55—$28.93—$28.93—8
05$0.000—28$0.000$55.46—$45.43—$45.43—8
06$0.000—16$0.000$41.55—$15.57—$19.13—7
07$0.000—14$0.000$2.99—$10.82—$10.82—1
08$0.000—9$0.000$7.25—$0.00—$0.00—2
09$0.000—16$0.000$17.03—$3.16—$3.16—5
10$0.000—21$0.000$19.53—$2.99—$2.99—5
11$0.000—17$0.000$45.12—$6.87—$6.87—8
12$0.000—18$0.000$22.31—$18.44—$25.24—6
13$0.000—17$0.000$3.56—$0.00—$0.00—1
14$0.000—13$0.000$149.58—$0.00—$0.00—7
15$0.000—21$0.000$19.76—$19.94—$19.94—6
16$0.000—24$0.000$39.36—$17.08—$17.08—8
17$0.000—22$0.000$57.00—$9.14—$24.91—11
18$0.000—24$0.000$39.86—$32.12—$38.85—8
19$0.000—26$0.000$49.29—$6.79—$10.19—13
20$0.000—11$0.000$47.34—$0.00—$0.00—9
21$0.000—20$0.000$23.58—$2.99—$2.99—7
22$0.000—17$0.000$26.34—$0.00—$5.93—3
23$0.000—21$0.000$3.58—$0.00—$0.00—1

The prior day, 2026-09-03, same two tables

(i) The portfolio, hour by hour, 2026-09-03. Every account placeable on the report clock, summed; delivery on the delivery hour, our columns on the hour the appendix names.

hrspendimpressionsCPMregistrationscost/regbookedbROASwithin-daywROAScohort@24hcROASbooked payers
00$18.331,695$10.8132$0.573$163.188.90$13.340.73$13.340.7313
01$20.362,294$8.8826$0.783$18.990.93$3.180.16$3.180.166
02$25.492,585$9.8630$0.850$95.403.74$102.484.02$102.484.0210
03$21.542,395$8.9936$0.598$114.115.30$18.020.84$18.020.8419
04$22.422,591$8.6537$0.606$96.934.32$87.993.92$87.993.9219
05$20.412,504$8.1536$0.567$46.032.26$19.900.98$19.900.988
06$19.102,199$8.6947$0.406$85.344.47$7.480.39$7.480.3913
07$20.071,956$10.2639$0.515$110.845.52$104.835.22$104.835.2215
08$19.392,280$8.5028$0.693$112.845.82$19.871.02$19.871.0217
09$21.122,485$8.5029$0.728$217.7510.31$0.000.00$0.000.0023
10$23.572,787$8.4638$0.620$165.357.02$13.710.58$17.270.7317
11$25.352,837$8.9440$0.634$227.948.99$50.291.98$50.291.9824
12$27.882,715$10.2750$0.558$185.526.65$90.023.23$100.623.6123
13$31.952,737$11.6748$0.666$73.162.29$38.821.22$42.401.3312
14$30.422,864$10.6261$0.499$215.727.09$39.821.31$39.821.3114
15$31.393,321$9.4562$0.506$131.354.18$12.490.40$12.490.4015
16$31.162,823$11.0436$0.866$149.824.81$58.931.89$62.492.0125
17$21.292,181$9.7641$0.519$224.2210.53$26.461.24$45.202.1225
18$17.251,497$11.5223$0.750$71.874.17$0.000.00$12.960.7512
19$15.101,251$12.0715$1.007$102.286.77$7.230.48$18.131.2018
20$16.351,099$14.8818$0.908$94.685.79$3.740.23$3.740.2314
21$14.471,027$14.0915$0.965$83.435.77$0.000.00$0.000.0010
22$11.56828$13.9619$0.608$106.329.20$10.670.92$10.670.927
23$12.75881$14.4713$0.981$46.713.66$0.000.00$0.000.005

(ii) The same 24 hours per account, 2026-09-03.

B1 — $337.39 on this day

hrspendimpressionsCPMregistrationscost/regbookedbROASwithin-daywROAScohort@24hcROASbooked payers
00$9.901,043$9.498$1.238$76.157.69$0.000.00$0.000.005
01$11.471,409$8.1419$0.604$12.671.11$3.180.28$3.180.284
02$15.051,559$9.6510$1.505$29.201.94$5.930.39$5.930.395
03$13.121,487$8.8212$1.093$38.022.90$5.800.44$5.800.449
04$12.691,602$7.9216$0.793$49.133.87$7.120.56$7.120.5612
05$12.831,593$8.0514$0.916$29.932.33$0.000.00$0.000.003
06$12.011,427$8.4226$0.462$32.452.70$7.480.62$7.480.626
07$12.421,216$10.2114$0.887$60.134.84$3.480.28$3.480.287
08$13.451,563$8.6113$1.035$52.383.89$0.000.00$0.000.008
09$13.961,754$7.9618$0.776$176.8812.67$0.000.00$0.000.0014
10$15.041,888$7.9718$0.836$40.342.68$10.530.70$14.090.948
11$16.071,789$8.9819$0.846$147.969.21$10.530.66$10.530.6614
12$16.831,690$9.9626$0.647$146.588.71$83.484.96$94.085.5915
13$18.871,693$11.1519$0.993$10.000.53$6.510.35$6.510.353
14$18.481,808$10.2219$0.973$162.448.79$2.940.16$2.940.166
15$18.682,098$8.9024$0.778$95.425.11$0.000.00$0.000.008
16$25.152,342$10.7419$1.324$82.663.29$23.360.93$23.360.9313
17$14.491,577$9.1925$0.580$121.498.38$7.120.49$25.861.7813
18$13.481,162$11.6016$0.843$42.863.18$0.000.00$0.000.006
19$11.54995$11.6013$0.888$29.122.52$7.230.63$18.131.577
20$12.72899$14.1514$0.909$67.115.28$3.740.29$3.740.296
21$10.83818$13.2413$0.833$21.862.02$0.000.00$0.000.003
22$8.64664$13.018$1.080$39.204.54$10.671.24$10.671.244
23$9.67704$13.747$1.381$3.560.37$0.000.00$0.000.001

Nomad Node — $0.00 on this day

hrspendimpressionsCPMregistrationscost/regbookedbROASwithin-daywROAScohort@24hcROASbooked payers
00$0.000—0—$0.00—$0.00—$0.00—0
01$0.000—0—$0.00—$0.00—$0.00—0
02$0.000—0—$0.00—$0.00—$0.00—0
03$0.000—0—$0.00—$0.00—$0.00—0
04$0.000—0—$0.00—$0.00—$0.00—0
05$0.000—0—$0.00—$0.00—$0.00—0
06$0.000—0—$0.00—$0.00—$0.00—0
07$0.000—0—$0.00—$0.00—$0.00—0
08$0.000—0—$0.00—$0.00—$0.00—0
09$0.000—0—$0.00—$0.00—$0.00—0
10$0.000—0—$0.00—$0.00—$0.00—0
11$0.000—0—$0.00—$0.00—$0.00—0
12$0.000—0—$0.00—$0.00—$0.00—0
13$0.000—0—$0.00—$0.00—$0.00—0
14$0.000—0—$0.00—$0.00—$0.00—0
15$0.000—0—$0.00—$0.00—$0.00—0
16$0.000—0—$0.00—$0.00—$0.00—0
17$0.000—0—$0.00—$0.00—$0.00—0
18$0.000—0—$0.00—$0.00—$0.00—0
19$0.000—0—$0.00—$0.00—$0.00—0
20$0.000—0—$0.00—$0.00—$0.00—0
21$0.000—0—$0.00—$0.00—$0.00—0
22$0.000—0—$0.00—$0.00—$0.00—0
23$0.000—0—$0.00—$0.00—$0.00—0

KBM1 — $0.00 on this day

hrspendimpressionsCPMregistrationscost/regbookedbROASwithin-daywROAScohort@24hcROASbooked payers
00$0.000—0—$48.34—$0.00—$0.00—1
01$0.000—0—$3.16—$0.00—$0.00—1
02$0.000—0—$0.00—$0.00—$0.00—0
03$0.000—0—$3.16—$0.00—$0.00—1
04$0.000—0—$0.00—$0.00—$0.00—0
05$0.000—0—$3.16—$0.00—$0.00—1
06$0.000—0—$9.80—$0.00—$0.00—2
07$0.000—0—$0.00—$0.00—$0.00—0
08$0.000—0—$20.61—$0.00—$0.00—3
09$0.000—0—$10.44—$0.00—$0.00—2
10$0.000—0—$10.06—$0.00—$0.00—2
11$0.000—0—$3.48—$0.00—$0.00—1
12$0.000—0—$22.32—$0.00—$0.00—3
13$0.000—0—$2.99—$0.00—$0.00—1
14$0.000—0—$15.46—$0.00—$0.00—1
15$0.000—0—$0.00—$0.00—$0.00—0
16$0.000—0—$14.37—$0.00—$0.00—3
17$0.000—0—$6.91—$0.00—$0.00—2
18$0.000—0—$0.00—$0.00—$0.00—0
19$0.000—0—$39.79—$0.00—$0.00—5
20$0.000—0—$0.00—$0.00—$0.00—0
21$0.000—0—$0.00—$0.00—$0.00—0
22$0.000—0—$48.34—$0.00—$0.00—1
23$0.000—0—$13.22—$0.00—$0.00—1

Dicey2 — $87.90 on this day

hrspendimpressionsCPMregistrationscost/regbookedbROASwithin-daywROAScohort@24hcROASbooked payers
00$3.46167$20.723$1.153$3.160.91$0.000.00$0.000.001
01$4.55312$14.582$2.275$3.160.69$0.000.00$0.000.001
02$4.42388$11.392$2.210$0.000.00$0.000.00$0.000.000
03$4.32353$12.245$0.864$9.282.15$9.282.15$9.282.152
04$4.25433$9.824$1.062$15.183.57$15.183.57$15.183.571
05$2.82334$8.444$0.705$0.000.00$0.000.00$0.000.000
06$3.04309$9.844$0.760$0.000.00$0.000.00$0.000.000
07$3.18234$13.593$1.060$12.693.99$76.3324.00$76.3324.001
08$1.96206$9.510—$12.696.48$0.000.00$0.000.001
09$3.70255$14.513$1.233$0.000.00$0.000.00$0.000.000
10$2.56249$10.282$1.280$12.694.96$0.000.00$0.000.001
11$3.30280$11.794$0.825$38.2511.59$26.448.01$26.448.011
12$3.44285$12.072$1.720$2.990.87$2.990.87$2.990.871
13$5.46312$17.504$1.365$5.931.09$9.491.74$9.491.741
14$3.59289$12.427$0.513$2.990.83$0.000.00$0.000.001
15$4.38293$14.954$1.095$2.990.68$0.000.00$0.000.001
16$4.42314$14.083$1.473$20.274.59$3.480.79$3.480.793
17$4.45306$14.544$1.113$16.553.72$3.560.80$3.560.802
18$3.77335$11.256$0.628$0.000.00$0.000.00$12.963.440
19$3.56256$13.912$1.780$13.223.71$0.000.00$0.000.001
20$3.63200$18.154$0.907$0.000.00$0.000.00$0.000.000
21$3.64209$17.422$1.820$0.000.00$0.000.00$0.000.000
22$2.92164$17.808$0.365$0.000.00$0.000.00$0.000.000
23$3.08177$17.406$0.513$3.161.02$0.000.00$0.000.001

HR1 — $0.00 on this day

hrspendimpressionsCPMregistrationscost/regbookedbROASwithin-daywROAScohort@24hcROASbooked payers
00$0.000—0—$19.51—$0.00—$0.00—2
01$0.000—0—$0.00—$0.00—$0.00—0
02$0.000—0—$0.00—$0.00—$0.00—0
03$0.000—0—$0.00—$0.00—$0.00—0
04$0.000—0—$0.00—$0.00—$0.00—0
05$0.000—0—$0.00—$0.00—$0.00—0
06$0.000—0—$3.56—$0.00—$0.00—1
07$0.000—0—$0.00—$0.00—$0.00—0
08$0.000—0—$0.00—$0.00—$0.00—0
09$0.000—0—$3.67—$0.00—$0.00—1
10$0.000—0—$0.00—$0.00—$0.00—0
11$0.000—0—$0.00—$0.00—$0.00—0
12$0.000—0—$0.00—$0.00—$0.00—0
13$0.000—0—$0.00—$0.00—$0.00—0
14$0.000—0—$18.76—$0.00—$0.00—2
15$0.000—0—$0.00—$0.00—$0.00—0
16$0.000—0—$0.00—$0.00—$0.00—0
17$0.000—0—$0.00—$0.00—$0.00—0
18$0.000—0—$0.00—$0.00—$0.00—0
19$0.000—0—$0.00—$0.00—$0.00—0
20$0.000—0—$3.56—$0.00—$0.00—1
21$0.000—0—$0.00—$0.00—$0.00—0
22$0.000—0—$0.00—$0.00—$0.00—0
23$0.000—0—$16.10—$0.00—$0.00—1

HR2 — $93.43 on this day

hrspendimpressionsCPMregistrationscost/regbookedbROASwithin-daywROAScohort@24hcROASbooked payers
00$4.97485$10.2521$0.237$16.043.23$13.342.68$13.342.684
01$4.34573$7.575$0.868$0.000.00$0.000.00$0.000.000
02$6.02638$9.4418$0.334$66.2011.00$96.5516.04$96.5516.045
03$4.10555$7.3919$0.216$63.6615.53$2.940.72$2.940.727
04$5.48556$9.8617$0.322$32.625.95$65.6911.99$65.6911.996
05$4.76577$8.2518$0.264$12.952.72$19.904.18$19.904.184
06$4.05463$8.7517$0.238$39.549.76$0.000.00$0.000.004
07$4.47506$8.8322$0.203$38.018.50$25.015.60$25.015.607
08$3.98511$7.7915$0.265$27.156.82$19.874.99$19.874.995
09$3.46476$7.278$0.432$26.767.73$0.000.00$0.000.006
10$5.97650$9.1818$0.332$102.2517.13$3.180.53$3.180.536
11$5.98768$7.7917$0.352$38.256.40$13.322.23$13.322.238
12$7.61740$10.2822$0.346$13.641.79$3.560.47$3.560.474
13$7.62732$10.4125$0.305$54.247.12$22.822.99$26.393.467
14$8.35767$10.8935$0.239$16.071.92$36.894.42$36.894.424
15$8.33930$8.9634$0.245$32.943.95$12.491.50$12.491.506
16$1.59167$9.5214$0.114$32.5120.45$32.0820.18$35.6422.426
17$2.35298$7.8912$0.196$79.2833.73$15.786.72$15.786.728
18$0.000—1$0.000$29.01—$0.00—$0.00—6
19$0.000—0—$20.15—$0.00—$0.00—5
20$0.000—0—$24.01—$0.00—$0.00—7
21$0.000—0—$61.57—$0.00—$0.00—7
22$0.000—3$0.000$18.79—$0.00—$0.00—2
23$0.000—0—$10.67—$0.00—$0.00—1

Against the prior day, hour matched to hour

Hours both days delivered in, on the UTC clock: 00, 01, 02, 03, 04, 05, 06, 07, 08, 09, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23. Only the shared hours are compared.

Counting which way each shared hour points, which needs no weighting and asks only for a direction:

metrichours 2026-09-04 ran higherhours it ran lowersign test
CPM21 of 243 of 24p = 0.000
CTR5 of 2419 of 24p = 0.007
click→install11 of 2413 of 24p = 0.839
CPI17 of 247 of 24p = 0.064

A sign test throws magnitude away and buys weight-free direction with it.

The full hour-by-hour pairing, 2026-09-03 → 2026-09-04

Every shared hour, 2026-09-03 → 2026-09-04. All columns are Meta's instrument on both sides of each ratio. Revenue is deliberately absent: per-hour payer counts run 0-3, so an hourly revenue column would be one person's basket read as a rate.

hrimpressionsCPMCTRclick→installCPI
001,210 → 85811.04 → 15.088.18% → 7.81%10.1% → 23.9%1.336 → 0.809
011,721 → 1,2949.31 → 9.727.73% → 7.34%14.3% → 10.5%0.843 → 1.258
021,947 → 1,81310.00 → 10.096.78% → 6.45%10.6% → 9.4%1.391 → 1.664
031,840 → 2,1959.48 → 9.417.66% → 6.70%11.3% → 12.9%1.090 → 1.087
042,035 → 2,2898.32 → 9.536.78% → 6.33%13.8% → 13.1%0.892 → 1.148
051,927 → 1,9868.12 → 11.057.89% → 6.29%11.8% → 18.4%0.869 → 0.954
061,736 → 1,7938.67 → 11.008.70% → 6.53%19.9% → 12.0%0.502 → 1.409
071,450 → 1,50910.76 → 10.017.79% → 5.63%14.2% → 23.5%0.975 → 0.755
081,769 → 1,7178.71 → 10.236.95% → 5.82%8.1% → 21.0%1.541 → 0.836
092,009 → 1,8568.79 → 11.457.17% → 4.74%16.0% → 11.4%0.768 → 2.125
102,137 → 1,7048.24 → 11.466.36% → 5.58%13.2% → 16.8%0.978 → 1.220
112,069 → 1,6639.36 → 11.147.30% → 6.67%13.9% → 9.9%0.922 → 1.685
121,975 → 1,52410.26 → 11.508.20% → 7.09%15.4% → 17.6%0.811 → 0.923
132,005 → 1,57612.13 → 12.597.83% → 7.17%17.2% → 8.8%0.901 → 1.984
142,097 → 1,57610.52 → 12.546.63% → 7.04%16.5% → 14.4%0.960 → 1.236
152,391 → 1,5209.64 → 11.297.78% → 6.97%14.5% → 19.8%0.854 → 0.817
162,656 → 1,26911.13 → 12.326.74% → 7.41%14.0% → 12.8%1.183 → 1.302
171,883 → 1,52110.06 → 13.627.91% → 8.48%16.8% → 14.0%0.758 → 1.151
181,497 → 1,52011.52 → 11.149.02% → 7.89%14.8% → 14.2%0.863 → 0.996
191,251 → 1,47212.07 → 13.348.95% → 9.10%15.2% → 20.1%0.888 → 0.727
201,099 → 1,33314.88 → 16.019.65% → 8.18%14.2% → 22.9%1.090 → 0.854
211,027 → 1,25614.09 → 15.9512.17% → 9.63%12.8% → 16.5%0.904 → 1.002
22828 → 1,06913.96 → 19.7110.99% → 12.16%16.5% → 14.6%0.771 → 1.109
23881 → 1,02514.47 → 20.3110.56% → 8.78%16.1% → 10.0%0.850 → 2.313

This pairing is B1 and Dicey2 on both sides, because HR2's hourly exports are not in the config these tables read, so no direction below is disturbed by its missing day.

CPM rose in 21 of 24 shared hours at p = 0.000 and the pooled read agrees, $10.26 to $12.06; the roster did not change between the days, so this is price and not mix.

CTR fell in 19 of 24 at p = 0.007 and pooled 7.84% to 7.12% at p = 0.000, both B1 creatives falling on their own, shirtb 10.21% to 8.57% and madapp 4.48% to 3.42%, while both of Dicey2's rose.

Click-to-install did not separate on either read, 11 hours of 24 at p = 0.839 and pooled 14.29% to 15.17% at p = 0.343.

CPI rose in 17 of 24 at p = 0.064 and pooled $0.917 to $1.118.

2026-09-05T17:24:32.406021 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 5 10 15 20 hour (UTC) 5 6 7 8 9 10 11 12 CTR % CTR 2026-09-03 2026-09-04 0 5 10 15 20 hour (UTC) 8 10 12 14 16 18 20 USD per 1,000 impressions CPM 0 5 10 15 20 hour (UTC) 8 10 12 14 16 18 20 22 24 installs / clicks % Click → install 0 5 10 15 20 hour (UTC) 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 USD Cost per install Delivery by hour — 2026-09-04 against 2026-09-03, shared hours only (UTC)
The same four ratios with the prior day laid over the day, shared hours only

Country, 2026-09-04

(i) The portfolio by country, 2026-09-04. On the report clock. Meta serves no hour x country grid, so each country cut is raked from its own Meta day onto UTC; our own columns were already there.

ccspendshareimpressionsCPMregistrations (ours)cost/regwithin-daywROASpayers
IN$108.9924.2%16,131$6.76113$0.965$109.611.0146
US$44.8810.0%616$72.8415$2.992$50.931.1320
MX$19.144.2%1,311$14.6017$1.126$0.000.008
ID$18.894.2%2,939$6.4330$0.630$11.750.6210
GB$18.794.2%445$42.255$3.758$0.000.007
TR$18.484.1%817$22.6123$0.803$35.761.949
AR$13.272.9%859$15.4615$0.885$8.970.685
BR$12.262.7%457$26.8417$0.721$18.941.544
CL$9.872.2%439$22.467$1.410$7.120.726
CO$9.572.1%674$14.1912$0.797$5.930.623
PH$9.482.1%1,947$4.8721$0.452$3.350.353
FR$8.631.9%222$38.802$4.315$0.000.000

154 further countries are not listed, $158.15 between them (35.1% of this block).

Reconciliation, what is measured and what is fitted. Every per-country cell is an allocation across hours; every total is measured. B1 (UTC-7) $99.05 from Meta 2026-09-03 and $250.19 from Meta 2026-09-04, a measured UTC-day total of $349.24; Dicey2 (UTC-7) $28.91 from Meta 2026-09-03 and $72.25 from Meta 2026-09-04, a measured UTC-day total of $101.16.

Absent from the table above, by name: Nomad Node, KBM1, HR1, HR2. Their spend is in every day-grain table; only their country split is missing.

(ii) The same cut per account, 2026-09-04.

B1

ccspendshareimpressionsCPMregistrations (ours)cost/regwithin-daywROASpayers
IN$84.7124.3%13,801$6.1488$0.963$106.051.2543
US$37.5310.7%517$72.6012$3.128$50.931.3620
MX$16.064.6%1,173$13.6916$1.004$0.000.008
GB$16.014.6%423$37.805$3.201$0.000.007
ID$15.894.6%2,635$6.0325$0.636$11.750.749
TR$12.973.7%632$20.5111$1.179$14.301.106
AR$10.583.0%707$14.955$2.115$2.990.283
CO$8.172.3%607$13.4510$0.817$2.970.362
PH$7.712.2%1,751$4.4018$0.428$0.000.002
FR$7.412.1%216$34.272$3.707$0.000.000
CL$7.262.1%336$21.623$2.419$7.120.986
MY$6.761.9%687$9.847$0.966$3.180.474

150 further countries are not listed, $118.19 between them (33.8% of this block).

Dicey2

ccspendshareimpressionsCPMregistrations (ours)cost/regwithin-daywROASpayers
IN$24.2924.0%2,330$10.4225$0.972$3.560.153
BR$12.2612.1%457$26.8417$0.721$18.941.544
US$7.357.3%99$74.053$2.451$0.000.000
TR$5.515.4%185$29.7712$0.459$21.463.893
MX$3.083.0%139$22.241$3.081$0.000.000
ID$2.993.0%304$9.855$0.599$0.000.001
GB$2.782.8%21$130.870—$0.000.000
AR$2.702.7%151$17.8210$0.270$5.982.222
CL$2.612.6%104$25.184$0.653$0.000.000
IT$2.402.4%58$41.483$0.801$0.000.000
PH$1.771.8%196$9.043$0.592$3.351.891
TH$1.731.7%139$12.475$0.346$0.000.000

107 further countries are not listed, $31.67 between them (31.3% of this block).

Nomad Node — no country block: it delivered nothing on this day, a measured zero.

KBM1 — no country block: its country export never arrived.

HR1 — no country block: its country export never arrived.

HR2 — no country block: its country export never arrived.

This section is B1 and Dicey2 on both sides and every ratio in it is a return, covering $382.72 of the day's $711.05 of within-day revenue; HR2's $328.33 sits outside it, its country export never having arrived.

B1's own India cell is the comparison this section supports, 0.88 to 1.25 on 3.7% more spend, with 35 and 43 booked payers behind it, while a registration there went $0.704 to $0.963.

The United States is 10.0% of the buy at 1.13 on 20 booked payers, $44.88 against $30.55 the day before, at a CPM of $72.84 against the portfolio's $12.06.

The known-answer test had no account to run on. Raking is checked by requiring a UTC+0 account to come back unchanged, and both accounts in this cut are UTC-7, so the day rests on the reconciliation above.

The cost-per-install figure below draws nothing, and has not on any day of this series. The raked country cut carries spend, impressions and our own registrations, and Meta's per-country install count is not in it, so that chart has no denominator.

2026-09-05T17:24:32.502031 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ IN US MX ID GB TR AR BR CL CO PH FR 0 25 50 75 100 125 150 175 USD delivery raked onto the report clock against two measured margins; revenue is the report day Spend against within-day revenue by country — 2026-09-04 spend within-day revenue
Spend against within-day revenue, by country
2026-09-05T17:24:32.537672 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ −0.04 −0.02 0.00 0.02 0.04 USD per install Cost per install by country — 2026-09-04 (report clock)
Cost per install by country

Aggregate, both days

2026-09-032026-09-04change
spend$518.72$450.40−13.2%
impressions51,83237,338−28.0%
CPM$10.01$12.06+20.5%
registrations (ours)819876+7.0%
cost per registration$0.633$0.514−18.8%
booked$2,939.79 (5.67)$2,744.80 (6.09)−6.6%
Meta booked$3,023.60 (5.83)$2,327.86 (5.17)
within-day$729.28 (1.41)$711.05 (1.58)−2.5%
cohort @ 24h$793.18 (1.53)$804.62 (1.79)+1.4%
booked payers268288+20
payer rate (8h)8.7%10.4%+1.7 pts
payer rate (d0)9.4%11.4%+2.0 pts
ARPU (8h)$0.8866$0.7082−20.1%
ARPU (d0)$0.9685$0.9185−5.2%
ARPPU (8h)$10.23$6.82−33.3%
ARPPU (d0)$10.30$8.05−21.9%

(ii) The same three measures per account on both days are the one-table-per-account block under The day, and every account above, which is where the instrument emits them.

Pooled over the 24 hours both days delivered in, on the UTC clock. Meta's instrument on both sides of every ratio.

spendimpressionsCPMclicksCTRinstallsclick→installCPI
2026-09-03$425.2941,44010.263,2477.84%46414.29%$0.917
2026-09-04$450.4037,33812.062,6577.12%40315.17%$1.118
  • click → install 14.29% → 15.17% (+6.1%, p = 0.343) — did not separate.
  • CTR 7.84% → 7.12% (-9.2%, p = 0.000) — fell.

Pooling across hours weights each hour by what it delivered. The hourly pairing is the unweighted read of the same two days; where they disagree, the mix moved.

The pooled funnel reads $425.29 and $450.40 where the day tables read $518.72 and $450.40, and the difference is HR2 on 09-03, which this loader holds on neither side.

Both horizons are closed on both days. On the spend this loader holds, the 24 h cohort reads $793.18 at 1.87 against $804.62 at 1.79 and the 8 h cohort $726.11 at 1.71 against $620.36 at 1.38; the day tables put 09-03's 24 h return at 1.53 because their denominator carries HR2's $93.43.

Booked revenue split by the UTC day its payer installed. The top row is the within-day figure.

installedbooked on 2026-09-04share
2026-09-04$711.0525.9%
2026-09-03$257.319.4%
2026-09-02$108.103.9%
2026-09-01$166.676.1%
2026-08-31$143.165.2%
2026-08-30$84.893.1%

Six install days carry $1,471.18 of the day's money; the remaining $1,273.62 arrived from installs on or before 08-29, which the instrument does not list separately.

$2,033.75 of the day's money came from installs bought before it, 74.1%, against 75.2% on 09-03.

2026-09-05T17:24:32.590300 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ booked within-day cohort @ 24h 0 500 1000 1500 2000 2500 3000 USD 6.91 1.71 1.87 6.09 1.58 1.79 the number above each bar is that measure's ROAS against the same day's spend The three revenue measures, 2026-09-03 against 2026-09-04 (UTC) 2026-09-03 2026-09-04
The three revenue measures on both days, ROAS above each bar
2026-09-05T17:24:32.635653 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 2026-09-03 2026-09-04 0 500 1000 1500 2000 2500 3000 USD the bottom segment IS the within-day figure: the same number seen twice Booked revenue by the day its payer installed (UTC) the day itself one day earlier two days earlier three days earlier four or more
Booked revenue by the day its payer installed

The campaigns

(i) Every campaign the portfolio ran, 2026-09-04.

campaignaccountspendshareimpressionsCPMregistrationscost/regwithin-daywROASpayers
bailingxia_srfhjyd_sn_ww_cvr_260820B1$374.5078.6%32,826$11.41327$1.145$315.930.84152
amclqzqgtmnn_5l4ya1pifdk3_sn_ww_cvr_260826Dicey2$102.2621.4%5,628$18.17107$0.956$66.790.6518
dxu2a6iwzx5r_5l4ya1pifdk3_sn_ww_cvr_260826HR2$0.000.0%0—441$0.000$328.33—99
bailingxia_srfhjyd_sn_ww_cvr_260820KBM1$0.000.0%0—1$0.000$0.00—11

(ii) The same, per account, 2026-09-04.

B1

campaignaccountspendshareimpressionsCPMregistrationscost/regwithin-daywROASpayers
bailingxia_srfhjyd_sn_ww_cvr_260820B1$374.50100.0%32,826$11.41327$1.145$315.930.84152

Nomad Node

nothing delivered or registered.

KBM1

campaignaccountspendshareimpressionsCPMregistrationscost/regwithin-daywROASpayers
bailingxia_srfhjyd_sn_ww_cvr_260820KBM1$0.000.0%0—1$0.000$0.00—11

Dicey2

campaignaccountspendshareimpressionsCPMregistrationscost/regwithin-daywROASpayers
amclqzqgtmnn_5l4ya1pifdk3_sn_ww_cvr_260826Dicey2$102.26100.0%5,628$18.17107$0.956$66.790.6518

HR1

nothing delivered or registered.

HR2

campaignaccountspendshareimpressionsCPMregistrationscost/regwithin-daywROASpayers
dxu2a6iwzx5r_5l4ya1pifdk3_sn_ww_cvr_260826HR2$0.000.0%0—441$0.000$328.33—99

B1's campaign CPM rose on its own day and Dicey2's rose further, $9.90 to $11.41 and $15.10 to $18.17, so the dearer campaign per impression got dearer again.

KBM1 runs a campaign with the same name as B1's, and the split is keyed on campaign id, which is what keeps its eleven payers out of B1's row.

HR2's campaign reads $0.00 with 441 registrations against it, because this section is cut on each account's own day from the day export and HR2's was not pulled; the 78.6% and 21.4% shares are therefore B1 against Dicey2 alone.

Activity history, B1, requested over Sep 2 to Sep 6 and rendered as Sep 2 2026 to Sep 5 2026, asserted before any row of it was read. No budget change falls inside either reported day: the $400 daily budget set on 2026-08-27 has not moved, so both days sit inside the same regime, R3-400.

The only event in the rendered window falls before both days: vid_ww_laceb_11s_9x16_sn_v1 went from Active to Inactive at 22:01 UTC on 09-02, and nothing changed on 09-03 or on 09-04. This is the first pair of days in the series with an unchanged roster on both sides of the comparison.

The estimator reads both arms at H=24 on this run, shirtb 0.68 and madapp 1.61 at 83% coverage each, over the account days 09-03 and 09-04 pooled inside R3-400; the same read over 09-02 and 09-03 on the 09-03 page was 0.50 and 1.14.

Head to head inside R3-400, the two arms still do not separate on revenue, 0.68 against 1.61 with intervals of 0.44 to 0.97 and 0.66 to 2.71, while their audiences do differ in value at −33.3% with an interval of −45.6% to −15.0%; shirtb's hour-standardised CPM runs 54% over madapp's and it is the dearer in 22 of 23 shared hours.

Held to B1's own campaign on the META clock at H=6, click-to-install rose and the payer rate did not separate: pooled 14.72% to 17.09% at p = 0.027, shirtb 13.21% to 16.93% at p = 0.002 and madapp 19.05% to 17.66% at p = 0.572, against install-to-pay 8.36% to 9.84% at p = 0.485 on 31 and 36 payers, at full coverage on both days.

Revenue cannot carry a verdict on this account: 737 installs and 76 payers with a coefficient of variation of 4.58, and the largest regime-matched arm is 5% of the installs a 20% resolution needs.

Every asset

(i) Every asset the portfolio ran, 2026-09-04. One row per ad, the account named on it.

assetaccountspendimpressionsCPMregistrationscost/regwithin-daywROAScohort@24hpayers
vid_ww_shirtb_11s_9x16_sn_v1B1$278.4220,290$13.72247$1.127$189.130.68$227.9944
vid_ww_madapp_7s_9x16_sn_v1B1$96.0812,536$7.6669$1.392$126.801.32$129.7913
vid_ww_creamvloom_7s_9x16_sn_v1Dicey2$77.304,304$17.9694$0.822$57.280.74$57.2813
vid_ww_creamsolara_7s_9x16_sn_v1Dicey2$24.961,324$18.8513$1.920$9.510.38$12.703
vid_ww_editortabs_13s_9x16_sn_v1HR2$0.000—379$0.000$253.73—$283.3173
vid_ww_gym_12s_9x16_sn_v1 - CopyHR2$0.000—55$0.000$60.65—$79.5817
vid_ww_jetcone_7s_9x16_sn_v1HR2$0.000—7$0.000$13.96—$13.963
vid_ww_shirt_10s_9x16_sn_v1B1$0.000—4$0.000$0.00—$0.0017
vid_ww_lace_10s_9x16_sn_v1B1$0.000—2$0.000$0.00—$0.0018
vid_ww_pooldirect_15s_9x16_sn_v1B1$0.000—2$0.000$0.00—$0.0015
vid_ww_templatewall_7s_9x16_sn_v1B1$0.000—2$0.000$0.00—$0.0014
vid_ww_kimono_10s_9x16_sn_v1B1$0.000—1$0.000$0.00—$0.007
vid_ww_poolslow_15s_9x16_sn_v1KBM1$0.000—1$0.000$0.00—$0.005
vid_ww_kitchen_10s_9x16_sn_v1Dicey2$0.000—0—$0.00—$0.001
vid_ww_laceb_11s_9x16_sn_v1B1$0.000—0—$0.00—$0.001
vid_ww_poolphoto_15s_9x16_sn_v1B1$0.000—0—$0.00—$0.0011
vid_ww_poolphotob_15s_9x16_sn_v1B1$0.000—0—$0.00—$0.007
vid_ww_poolslow_15s_9x16_sn_v1B1$0.000—0—$0.00—$0.002
vid_ww_stockings_11s_9x16_sn_v1Dicey2$0.000—0—$0.00—$0.001
vid_ww_templatewallb_7s_9x16_sn_v1B1$0.000—0—$0.00—$0.002
vid_ww_trainpress2_15s_9x16_sn_v1B1$0.000—0—$0.00—$0.001

(ii) The same, per account, 2026-09-04.

B1

assetaccountspendimpressionsCPMregistrationscost/regwithin-daywROAScohort@24hpayers
vid_ww_shirtb_11s_9x16_sn_v1B1$278.4220,290$13.72247$1.127$189.130.68$227.9944
vid_ww_madapp_7s_9x16_sn_v1B1$96.0812,536$7.6669$1.392$126.801.32$129.7913
vid_ww_shirt_10s_9x16_sn_v1B1$0.000—4$0.000$0.00—$0.0017
vid_ww_lace_10s_9x16_sn_v1B1$0.000—2$0.000$0.00—$0.0018
vid_ww_pooldirect_15s_9x16_sn_v1B1$0.000—2$0.000$0.00—$0.0015
vid_ww_templatewall_7s_9x16_sn_v1B1$0.000—2$0.000$0.00—$0.0014
vid_ww_kimono_10s_9x16_sn_v1B1$0.000—1$0.000$0.00—$0.007
vid_ww_laceb_11s_9x16_sn_v1B1$0.000—0—$0.00—$0.001
vid_ww_poolphoto_15s_9x16_sn_v1B1$0.000—0—$0.00—$0.0011
vid_ww_poolphotob_15s_9x16_sn_v1B1$0.000—0—$0.00—$0.007
vid_ww_poolslow_15s_9x16_sn_v1B1$0.000—0—$0.00—$0.002
vid_ww_templatewallb_7s_9x16_sn_v1B1$0.000—0—$0.00—$0.002
vid_ww_trainpress2_15s_9x16_sn_v1B1$0.000—0—$0.00—$0.001

Nomad Node

nothing delivered or registered.

KBM1

assetaccountspendimpressionsCPMregistrationscost/regwithin-daywROAScohort@24hpayers
vid_ww_poolslow_15s_9x16_sn_v1KBM1$0.000—1$0.000$0.00—$0.005

Dicey2

assetaccountspendimpressionsCPMregistrationscost/regwithin-daywROAScohort@24hpayers
vid_ww_creamvloom_7s_9x16_sn_v1Dicey2$77.304,304$17.9694$0.822$57.280.74$57.2813
vid_ww_creamsolara_7s_9x16_sn_v1Dicey2$24.961,324$18.8513$1.920$9.510.38$12.703
vid_ww_kitchen_10s_9x16_sn_v1Dicey2$0.000—0—$0.00—$0.001
vid_ww_stockings_11s_9x16_sn_v1Dicey2$0.000—0—$0.00—$0.001

HR1

nothing delivered or registered.

HR2

assetaccountspendimpressionsCPMregistrationscost/regwithin-daywROAScohort@24hpayers
vid_ww_editortabs_13s_9x16_sn_v1HR2$0.000—379$0.000$253.73—$283.3173
vid_ww_gym_12s_9x16_sn_v1 - CopyHR2$0.000—55$0.000$60.65—$79.5817
vid_ww_jetcone_7s_9x16_sn_v1HR2$0.000—7$0.000$13.96—$13.963

This section is on each account's own Meta day, as the campaign section above is. B1's rows sum to $374.50 and Dicey2's to $102.26; HR2's read $0.00 for the reason the campaign section gives, and tables A and B below are the same assets on the report clock.

shirtb is $278.42 of B1's $374.50 on the account day and $253.85 of its $349.24 on the report clock, on 247 registrations and 44 booked payers; madapp took $96.08 and $95.39 on 69 and 13.

vid_ww_gym_12s_9x16_sn_v1 - Copy is two ad ids under one ledger name, which tables A and B carry separately as HR2/fittingroom and HR2/povpose; all 55 of its registrations and all $60.65 of its within-day revenue are povpose's.

Ten rows in table B booked money on no measured spend, $1,554.00 between them, 56.6% of Sinai's booked total: seven are creatives pulled or accounts stopped before the day, and three are HR2's, whose spend on the day is unmeasured rather than absent.

2026-09-05T17:24:32.687607 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 CTR % CTR by hour — every asset, 2026-09-04 (UTC) B1/shirtb B1/madapp Dicey2/creamvloom Dicey2/creamsolara
CTR by hour, one line per asset
2026-09-05T17:24:32.800627 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0 10 20 30 40 50 installs / clicks % Click → install by hour — every asset, 2026-09-04 (UTC) B1/shirtb B1/madapp Dicey2/creamvloom Dicey2/creamsolara
Click-to-install by hour, one line per asset
2026-09-05T17:24:32.853366 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 1 2 3 4 5 6 USD Cost per install by hour — every asset, 2026-09-04 (UTC) B1/shirtb B1/madapp Dicey2/creamvloom Dicey2/creamsolara
Cost per install by hour, one line per asset
2026-09-05T17:24:32.745167 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 5 10 15 20 25 30 35 USD per 1,000 impressions CPM by hour — every asset, 2026-09-04 (UTC) B1/shirtb B1/madapp Dicey2/creamvloom Dicey2/creamsolara
CPM by hour, one line per asset

The standing per-asset tables

One row per ad across the whole Sinai buy, the account named on each. Delivery, then the three revenue measures, then the country mix.

A. Per asset — delivery

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

assetspendimpressionsCPMclicksCTRinstallsclick→installCPI
B1/shirtb$253.8519,16513.251,6438.57%23814.49%$1.067
B1/madapp$95.3912,6987.514343.42%6815.67%$1.403
Dicey2/creamvloom$78.684,21918.6542910.17%8419.58%$0.937
Dicey2/creamsolara$22.481,25617.9015112.02%127.95%$1.873
B1/lace$0.000—0—0——
B1/shirt$0.000—0—0——
B1/pooldirect$0.000—0—0——
B1/kimono$0.000—0—0——
KBM1/poolslow$0.000—0—0——
HR2/fittingroom$0.000—0—0——
HR2/povpose$0.000—0—0——
HR2/editortabs$0.000—0—0——
B1/templatewall$0.000—0—0——
HR2/jetcone$0.000—0—0——

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, 100% covered on this day. Report clock UTC.

assetbookedbROAS (Meta)within-daywROAScohort@HcROASpayers (d0)payer% (d0)ARPU (d0)ARPPU (d0)largest payer
B1/shirtb$316.411.25 (2.61)$189.130.75$227.990.903213.0%0.9237.129.9%
B1/madapp$232.842.44 (5.01)$126.801.33$129.791.36710.1%1.88118.5439.3%
Dicey2/creamvloom$76.950.98 (8.66)$57.280.73$57.280.731010.6%0.6095.7325.0%
Dicey2/creamsolara$16.630.74 (5.01)$9.510.42$12.700.57215.4%0.9776.3566.4%
B1/lace$123.38— (—)$0.00—$0.00—00.0%0.000——
B1/shirt$156.58— (—)$0.00—$0.00—00.0%0.000——
B1/pooldirect$116.96— (—)$0.00—$0.00—00.0%0.000——
B1/kimono$111.92— (—)$0.00—$0.00—00.0%0.000——
KBM1/poolslow$59.72— (—)$0.00—$0.00—00.0%0.000——
HR2/fittingroom$24.29— (—)$0.00—$0.00—00.0%0.000——
HR2/povpose$91.49— (—)$60.65—$79.58—713.2%1.50211.3761.8%
HR2/editortabs$628.29— (—)$253.73—$283.31—4010.6%0.7487.088.2%
B1/templatewall$221.10— (—)$0.00—$0.00—00.0%0.000——
HR2/jetcone$20.27— (—)$13.96—$13.96—228.6%1.9946.9851.2%

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.

assetspendIndia %US %other %US % of its revenue
B1/shirtb$278.4219.7%13.0%67.3%3.2%
B1/madapp$96.0831.9%13.8%54.3%35.4%
Dicey2/creamvloom$77.3023.3%7.6%69.1%0.0%
Dicey2/creamsolara$24.9625.3%6.6%68.1%0.0%

Two rows clear the 16-payer bar at d0, HR2/editortabs at 40 payers and 10.6% and B1/shirtb at 32 and 13.0%; every other row in table B is a direction.

B1/shirtb took 72.7% of B1's report-clock spend and returned 0.90 at 24 h, ARPU (d0) $0.923 on 32 payers with 9.9% of its money in one basket; on 09-03 the same row read 0.37 on exactly 16 payers, so this is the one per-asset comparison on the page with both sides at the bar.

B1/madapp reads 1.36 at 24 h on 7 payers with 39.3% of its money in one basket, and is not readable at this sample size; it read 1.36 on 14 the day before.

Installs cost $0.937 to $1.873 across the four assets that spent, on the report clock, Dicey2/creamvloom at the bottom and Dicey2/creamsolara at the top on 12 installs; the spread is delivery, and cost per install does not rank an asset here.

⚠ HR2/editortabs is the ad's NAME and not its clip. Ad 120251516018270671 plays vid_ww_uploadpanel_7s_9x16_sn_v1.mp4, read off its own media editor on 2026-09-01, so this row is that clip's record and the editor-tabs concept has never been live anywhere.

B1/madapp bought 13.8% of its spend in the United States and 35.4% of its within-day money came from it, against B1/shirtb at 13.0% and 3.2%, and neither Dicey2 asset took a US dollar of revenue.

The three per-user charts below carry d0 in their captions and the tables above are now at d0, so the 2026-09-04 point on each is a closed horizon.

2026-09-05T17:24:32.986683 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ B1/shirtb B1/madapp Dicey2/creamvloom 0 20 40 60 80 100 % of that asset's spend Where each asset bought — share of its own spend by region, 2026-09-04 (META day) IN US ID GB MX TR other
Where each asset bought: share of its own spend by region
2026-09-05T17:24:33.033808 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ B1/shirtb B1/madapp Dicey2/creamvloom 0 1 2 3 4 5 6 USD per install Cost per install by region — every asset, 2026-09-04 (META day) IN US ID GB MX TR
Cost per install by region, every asset
2026-09-05T17:24:33.081988 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ B1/shirtb B1/madapp Dicey2/creamvloom 0 10 20 30 40 50 60 70 80 USD Within-day revenue by region — every asset, 2026-09-04 (META day) IN US ID TR
Within-day revenue by region, every asset
2026-09-05T17:24:33.203067 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 09-01 09-02 09-03 09-04 day (UTC) 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 cohort payers / installs % Payer rate d0 by day — every asset (UTC) B1/shirtb B1/poolphotob B1/madapp Dicey2/creamvloom B1/laceb Dicey2/creamsolara B1/templatewallb HR2/editortabs Dicey2/creamseed HR2/gym HR2/fittingroom HR2/povpose
Payer rate d0 by day, one line per asset
2026-09-05T17:24:33.281426 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 09-01 09-02 09-03 09-04 day (UTC) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 cohort USD per install ARPU d0 by day — every asset (UTC) B1/shirtb B1/poolphotob B1/madapp Dicey2/creamvloom B1/laceb Dicey2/creamsolara B1/templatewallb HR2/editortabs Dicey2/creamseed HR2/gym HR2/fittingroom HR2/povpose
ARPU d0 by day, one line per asset
2026-09-05T17:24:33.353194 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 09-01 09-02 09-03 09-04 day (UTC) 0 5 10 15 20 25 cohort USD per payer ARPPU d0 by day — every asset (UTC) B1/shirtb B1/poolphotob B1/madapp Dicey2/creamvloom B1/laceb Dicey2/creamsolara B1/templatewallb HR2/editortabs Dicey2/creamseed HR2/gym HR2/fittingroom HR2/povpose
ARPPU d0 by day, one line per asset
2026-09-05T17:24:33.418358 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 09-01 09-02 09-03 09-04 day (UTC) 0 2 4 6 8 revenue / spend Within-day ROAS by day — every asset (UTC) B1/shirtb B1/poolphotob B1/madapp Dicey2/creamvloom B1/laceb Dicey2/creamsolara B1/templatewallb HR2/editortabs Dicey2/creamseed HR2/fittingroom
Within-day ROAS by day, one line per asset
2026-09-05T17:24:33.487065 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ CTR click→install CPI payer rate * ARPU * wROAS * 0.0 0.2 0.4 0.6 0.8 1.0 normalised across the day (1.0 = best of the pack) * axis sits under the 16-payer bar on most days: direction only, never a level. Assets under 20 installs on the day are not plotted. Composite — every asset across every measure, 2026-09-04 B1/shirtb B1/madapp Dicey2/creamvloom
Composite: every asset across every measure, normalised across the day

Vloom — every section below this line is Vloom's: the day, the hours, the countries, the campaigns and the assets, counted on Vloom's own ledger and its own accounts. Use the buttons above to switch.

Vloom's tab. Every section below is Vloom's: the day, the hours, the countries, the campaigns and the assets, on the lum ledger and the accounts that carried it.

Vloom bought nothing on 2026-09-04, as it bought nothing on 2026-09-03, and $1,084.43 still arrived from people who installed before the day.

The day, and every account

2026-09-032026-09-04change
spend$0.00$0.00—
impressions00—
CPM$0.00$0.00—
registrations (ours)2420−16.7%
cost per registration$0.000$0.000—
booked$561.34$1,084.43+93.2%
Meta booked——
within-day$0.00$10.62new
cohort @ 24h$0.00$10.62new
booked payers5459+5
payer rate (8h)0.0%10.0%+10.0 pts
payer rate (d0)0.0%10.0%+10.0 pts
ARPU (8h)$0.0000$0.5311new
ARPU (d0)$0.0000$0.5311new
ARPPU (8h)$0.00$5.31new
ARPPU (d0)$0.00$5.31new

2026-09-04 (the day) — delivery

accountspendshareimpressionsCPMregistrationscost/reg
Portfolio$0.000%0—20$0.000
B1$0.000%0—16$0.000
Nomad Node$0.000%0—4$0.000
KBM1$0.000%0—0—
Dicey2$0.000%0—0—
HR1$0.000%0—0—
HR2$0.000%0—0—

2026-09-04 (the day) — the three measures

accountbookedbROASwithin-daywROAScohort@24hcROASpayers (d0)payer% (d0)ARPU (d0)ARPPU (d0)
Portfolio$1,084.43—$10.62—$10.62—210.0%$0.5311$5.31
B1$893.61—$10.62—$10.62—212.5%$0.6639$5.31
Nomad Node$187.83—$0.00—$0.00—00.0%$0.0000—
KBM1$2.99—$0.00—$0.00—0———
Dicey2$0.00—$0.00—$0.00—0———
HR1$0.00—$0.00—$0.00—0———
HR2$0.00—$0.00—$0.00—0———

Each account, 2026-09-03 against 2026-09-04, one table each.

B1

2026-09-032026-09-04change
spend$0.00$0.00—
impressions00—
CPM$0.00$0.00—
registrations (ours)2216−27.3%
cost per registration$0.000$0.000—
booked$537.08$893.61+66.4%
Meta booked——
within-day$0.00$10.62new
cohort @ 24h$0.00$10.62new
booked payers4952+3
payer rate (8h)0.0%12.5%+12.5 pts
payer rate (d0)0.0%12.5%+12.5 pts
ARPU (8h)$0.0000$0.6639new
ARPU (d0)$0.0000$0.6639new
ARPPU (8h)$0.00$5.31new
ARPPU (d0)$0.00$5.31new

Nomad Node

2026-09-032026-09-04change
spend$0.00$0.00—
impressions00—
CPM$0.00$0.00—
registrations (ours)24+100.0%
cost per registration$0.000$0.000—
booked$15.30$187.83+1127.9%
Meta booked——
within-day$0.00$0.00—
cohort @ 24h$0.00$0.00—
booked payers46+2
payer rate (8h)0.0%0.0%+0.0 pts
payer rate (d0)0.0%0.0%+0.0 pts
ARPU (8h)$0.0000$0.0000—
ARPU (d0)$0.0000$0.0000—
ARPPU (8h)$0.00$0.00—
ARPPU (d0)$0.00$0.00—

KBM1 — nothing delivered and nobody registered on either day.

Dicey2 — nothing delivered and nobody registered on either day.

HR1 — nothing delivered and nobody registered on either day.

HR2 — nothing delivered and nobody registered on either day.

Two people who arrived on 09-04 paid inside the day, $10.62 between them against no spend anywhere, which is the first within-day revenue this app has recorded since it stopped buying; the 24 h cohort closed at the same $10.62.

Read the booked column for what the base is still paying and nothing else here. Its $1,084.43 is inherited but for that $10.62.

Hourly, 2026-09-04

No hourly section: nothing delivered in any hour of this day, in any of the seven accounts, so every cell of the 24-row table and every one of its per-account blocks is a measured zero. The last Vloom delivery was the shared Nomad Node pair on 08-23. The charts this section would carry are not drawn, because a chart of a zero day draws nothing, and no figure on this page is Vloom's: the figure directory for the day holds Sinai's nineteen and nothing else.

Country, 2026-09-04

No country section: a country split of no spend has no rows. The country export for this day returns nothing for any Vloom campaign, and the raking that puts a country cut on the report clock has nothing to rake.

Aggregate, both days

2026-09-032026-09-04change
spend$0.00$0.00—
impressions00—
CPM$0.00$0.00—
registrations (ours)2420−16.7%
cost per registration$0.000$0.000—
booked$561.34$1,084.43+93.2%
Meta booked——
within-day$0.00$10.62new
cohort @ 24h$0.00$10.62new
booked payers5459+5
payer rate (8h)0.0%10.0%+10.0 pts
payer rate (d0)0.0%10.0%+10.0 pts
ARPU (8h)$0.0000$0.5311new
ARPU (d0)$0.0000$0.5311new
ARPPU (8h)$0.00$5.31new
ARPPU (d0)$0.00$5.31new

The 24 h rows are closed on both days and rest on two payers, which is a long way under the bar.

The campaigns

(i) Every campaign the portfolio ran, 2026-09-04.

campaignaccountspendshareimpressionsCPMregistrationscost/regwithin-daywROASpayers
bailingxia_meituan_ww_cvr_260802B1$0.000.0%0—11$0.000$10.62—44
bailingxia_meituan_t3_cvr_260725B1$0.000.0%0—4$0.000$0.00—2
0811_vl_bailingxia_srfhjyd-rain-chuvaNomad Node$0.000.0%0—1$0.000$0.00—2
0814_vl_bailingxia_srfhjyd-rain-chuvaNomad Node$0.000.0%0—1$0.000$0.00—2
0814_vl_bailingxia_srfhjyd_maimaiNomad Node$0.000.0%0—1$0.000$0.00—2
0818_vl_bailingxia_srfhjyd_maimaiNomad Node$0.000.0%0—1$0.000$0.00—0
bailingxia_meituan_ww_cvr_260730B1$0.000.0%0—1$0.000$0.00—5
bailingxia_meituan_t2_cvr_260725B1$0.000.0%0—0—$0.00—1

(ii) The same, per account, 2026-09-04.

B1

campaignaccountspendshareimpressionsCPMregistrationscost/regwithin-daywROASpayers
bailingxia_meituan_ww_cvr_260802B1$0.000.0%0—11$0.000$10.62—44
bailingxia_meituan_t3_cvr_260725B1$0.000.0%0—4$0.000$0.00—2
bailingxia_meituan_ww_cvr_260730B1$0.000.0%0—1$0.000$0.00—5
bailingxia_meituan_t2_cvr_260725B1$0.000.0%0—0—$0.00—1

Nomad Node

campaignaccountspendshareimpressionsCPMregistrationscost/regwithin-daywROASpayers
0811_vl_bailingxia_srfhjyd-rain-chuvaNomad Node$0.000.0%0—1$0.000$0.00—2
0814_vl_bailingxia_srfhjyd-rain-chuvaNomad Node$0.000.0%0—1$0.000$0.00—2
0814_vl_bailingxia_srfhjyd_maimaiNomad Node$0.000.0%0—1$0.000$0.00—2
0818_vl_bailingxia_srfhjyd_maimaiNomad Node$0.000.0%0—1$0.000$0.00—0

KBM1

nothing delivered or registered.

Dicey2

nothing delivered or registered.

HR1

nothing delivered or registered.

HR2

nothing delivered or registered.

No campaign delivered. Every Vloom campaign in every account spent $0.00 on this day; the money in the tables above is booked against install days that precede it, and Meta stamps a conversion on the conversion hour, so those rows keep landing on days with no delivery behind them.

Every asset

(i) Every asset the portfolio ran, 2026-09-04. One row per ad, the account named on it.

assetaccountspendimpressionsCPMregistrationscost/regwithin-daywROAScohort@24hpayers
vid_ww_pool5_15s_9x16_v1B1$0.000—7$0.000$0.00—$0.0023
vid_ww_bodysuit_10s_9x16_v1B1$0.000—4$0.000$5.63—$5.6314
vid_sasian_bodysuit_10s_9x16_v1B1$0.000—2$0.000$0.00—$0.001
nomad-0811-03-rain-chuva-video-01Nomad Node$0.000—1$0.000$0.00—$0.002
nomad-0814-03-maimaiNomad Node$0.000—1$0.000$0.00—$0.002
nomad-0814-03-rain-chuva-video-01Nomad Node$0.000—1$0.000$0.00—$0.002
nomad-0818-03-maimaiNomad Node$0.000—1$0.000$0.00—$0.000
vid_control_6s_9x16_v1B1$0.000—1$0.000$0.00—$0.000
vid_sasian_spice5_15s_9x16_v1B1$0.000—1$0.000$0.00—$0.000
vid_ww_bikini_10s_9x16_v1B1$0.000—1$0.000$4.99—$4.996
vid_latina_es_bikini_10s_9x16_v1B1$0.000—0—$0.00—$0.001
vid_sasian_bikini_6s_9x16_v1B1$0.000—0—$0.00—$0.001
vid_ww_lounge_8s_9x16_v1B1$0.000—0—$0.00—$0.002
vid_ww_spill_8s_9x16_v1 - CopyB1$0.000—0—$0.00—$0.004

(ii) The same, per account, 2026-09-04.

B1

assetaccountspendimpressionsCPMregistrationscost/regwithin-daywROAScohort@24hpayers
vid_ww_pool5_15s_9x16_v1B1$0.000—7$0.000$0.00—$0.0023
vid_ww_bodysuit_10s_9x16_v1B1$0.000—4$0.000$5.63—$5.6314
vid_sasian_bodysuit_10s_9x16_v1B1$0.000—2$0.000$0.00—$0.001
vid_control_6s_9x16_v1B1$0.000—1$0.000$0.00—$0.000
vid_sasian_spice5_15s_9x16_v1B1$0.000—1$0.000$0.00—$0.000
vid_ww_bikini_10s_9x16_v1B1$0.000—1$0.000$4.99—$4.996
vid_latina_es_bikini_10s_9x16_v1B1$0.000—0—$0.00—$0.001
vid_sasian_bikini_6s_9x16_v1B1$0.000—0—$0.00—$0.001
vid_ww_lounge_8s_9x16_v1B1$0.000—0—$0.00—$0.002
vid_ww_spill_8s_9x16_v1 - CopyB1$0.000—0—$0.00—$0.004

Nomad Node

assetaccountspendimpressionsCPMregistrationscost/regwithin-daywROAScohort@24hpayers
nomad-0811-03-rain-chuva-video-01Nomad Node$0.000—1$0.000$0.00—$0.002
nomad-0814-03-maimaiNomad Node$0.000—1$0.000$0.00—$0.002
nomad-0814-03-rain-chuva-video-01Nomad Node$0.000—1$0.000$0.00—$0.002
nomad-0818-03-maimaiNomad Node$0.000—1$0.000$0.00—$0.000

KBM1

nothing delivered or registered.

Dicey2

nothing delivered or registered.

HR1

nothing delivered or registered.

HR2

nothing delivered or registered.

No asset delivered, so every spend cell above is $0.00 and no ratio in the block is defined. Registrations still arrived against no impression anywhere, 16 in B1 and 4 in Nomad Node, and vid_ww_bodysuit_10s_9x16_v1 and vid_ww_bikini_10s_9x16_v1 carry the day's $5.63 and $4.99.

3Caveats

This is the 24 h publication of 2026-09-04 and it completes the day. The horizon closed at 00:00 UTC on 09-06 and the payment cut was taken at 00:19, so every (d0) and cohort @ 24h cell on both days is at 100% coverage and frozen. Day against day at 24 h is the comparison this page supports.

The 8 h page and this one disagree about the size of the per-user fall, and the difference is the horizon. Sinai's 09-03 cohort had paid 91.5% of its 24 h money by 8 h and its 09-04 cohort 77.1%, so the morning page's 18.2% fall in ARPU was mostly 09-03 paying early; at d0 the fall is 3.3% at the portfolio line and 5.2% for Sinai. An 8 h change is a direction until the evening page closes it.

HR2 spent on 2026-09-04 and none of it is measured. Its hourly, day and country exports were all skipped from that day after an identity checkpoint took the account out of the automated pull, so its $0.00 is an absence. Its campaign registered 441 people in every hour of the day, more than the 341 it registered in eighteen delivering hours on 09-03, and $376.85 of the day's 24 h cohort revenue is its.

Two instruments disagree about HR2's first seven hours, and the run cannot settle it. UTC hours 00 to 06 of 09-04 are Meta hours 17 to 23 of HR2's 09-03 account day, whose archived hourly file reads $0.00 there; our registration ledger puts 130 registrations through HR2's campaign in those same hours, 32 of them at 01:00, the largest hour any account had on the day. Either that file is a snapshot taken before those hours delivered, or the registrations arrived through HR2's link with no delivery behind them; the size and the hour-by-hour spread point to the first, and only HR2's own export can close it. Hours 07 to 23 come from no file at all.

So the portfolio's headline fall of 13.2% is a coverage change and not a buying decision. The 09-03 figure carries HR2's $93.43 and the 09-04 figure carries no HR2 at all. The comparison this page supports is B1 and Dicey2 on both sides, $425.29 to $450.40, which is the pooled funnel.

Cost per registration falls 18.3% at the portfolio line for the same reason. Its numerator on 09-04 is two accounts' spend and its denominator is three accounts' registrations. B1's own figure rose 23.5% to $1.068 and Dicey2's fell 5.4% to $0.945.

Every portfolio-level return on this page divides three accounts' revenue by two accounts' spend. That is true of within-day, of both cohort horizons and of booked, and it was true of 09-03 only for the country section. The per-account tables are the cut where cost and revenue name the same account, and summed over B1 and Dicey2 the 24 h cohort return is 0.95 on both days.

The money ledger names thirty-two account-days it cannot measure, and one of them delivered. sinai_ww/HR2 2026-09-04 is the one; the other thirty-one are NN2, KBM1 and HR1 across 08-29 to 09-04, none of which delivered. The ledger's $450.40 for this day matches the page and is short by HR2's unmeasured spend.

The figures moved between the day's two publications by the settling the exports do after midnight. Spend reads $450.40 against $448.34 on the 8 h page, B1 $349.24 against $347.68 and Dicey2 $101.16 against $100.66; the 09-03 column reads $518.72 where the 09-03 page reads $518.47, 25 cents between B1 and Dicey2. No revenue figure moved.

The pooled funnel and the hour pairing carry no HR2 on either day. Their loader holds B1 and Dicey2 for both days, so 09-03 reads $425.29 there against $518.72 in the day tables; both sides are the same two accounts, which is what makes the directions comparable.

The campaign and per-asset sections read HR2 at $0.00 on both days. Those two sections cut each account on its own day from the day export, and HR2's day export was not pulled on this run, so its $0.00 there is an absence on 09-03 as well as on 09-04.

The country section carries no HR2 block at all on 09-04. Unlike 09-03, where it appeared with registrations and payers against no spend, it is absent here by name, so the section's spend and its revenue are both B1's and Dicey2's and its ratios are returns. It reaches $382.72 of the day's $711.05 of within-day revenue.

The payer counts in three tables are not the payer counts in table B. The country, campaign and per-asset sections emitted by the portfolio instrument count people who paid inside the day whatever day they installed; table B counts the day's own install cohort at d0. So B1/shirtb reads 44 in one and 32 in the other, and the 16-payer bar is being applied to two different populations.

within-day is sealed and does not move. It is the column to read on this page, and the 24 h cohort is the cross-check on it, closed on both days.

The strike test is one-sided and an unstruck day is not a clean one. Meta's purchase count stayed under the app's transaction count on both days, so no line is struck, but on Dicey2's own row Meta books $838.89 against $175.20 the day before and $99.51 on this one. The excess is not split per account or per asset, so Dicey2's (Meta) brackets in table B are not that account's revenue.

There is no ours / Meta instrument reading for Sinai in this run. That gate is valid only on the META clock, and the META-clock pass was configured against the Vloom ledger, which bought nothing. The 1.18 ratio the UTC pass prints is stamped CROSS-CLOCK, NOT A VALID CHECK by the instrument itself and is not the gate.

There is no mix-neutral account CPM line on this page. The instrument that renders it ran against the Vloom config, which delivered nothing, so the line it emitted reads $0.00 to $0.00. The campaign mix is stated instead, as each campaign's share of spend on both days.

Two different kinds of zero sit in the account tables. NN1 exported at every grain and served no impression, which is a measured zero. NN2, KBM1 and HR1 were stopped by Meta on 08-23, 08-24 and 08-24 and HR2 left the pull on 09-04, so their $0.00 is an absence of measurement; all four still book revenue through the payment ledger, which needs no export.

The estimator's H=24 read pools two account days and is not this UTC day. Its window is the account days 09-03 and 09-04, so its 0.68 and 1.61 sit on 431 and 166 installs across both, at 83% coverage; the 09-03 page's 0.50 and 1.14 were the window one day earlier. The day-level per-asset figures are table B's.

The ads map labels ad 120251516018270671 HR2/editortabs, which is its name in Ads Manager and not the file it plays. The local correction to HR2/uploadpanel recorded in daily_pipeline.py is not in this run's configs: the map is auto-derived from the ad name for any id it does not already carry, and a map rebuilt from scratch loses the entry.

128 ad names in the sweep do not follow the naming convention and can only be matched exactly. No rejected asset is attached in another account, so nothing was pulled across.

Dicey2's budget history is unverified for this run. The staged history is B1's, so the $100 daily budget its regime table records for 2026-08-28 onward has not been re-checked here. Dicey2 spent $102.26 on its own day against that figure.

The per-account coverage note the instrument emits states the wrong targeting for two accounts. It reads KBM1 and HR1 as United States only at 100% US delivery, which is their Vloom targeting, attached to the account rather than to the campaign. Neither account delivered on this day, so nothing above rests on it.

Every figure is priced through an FX table stamped 2026-08-01, which reads 0.8% low against Google Play's own record.

4What this hands to the next read

HR2 needs either a restored pull or an explicit exclusion from the portfolio denominator. While it registers 441 people a day against $0.00 of measured spend and earns $376.85 at 24 h, every blended figure on the page is two accounts' cost against three accounts' revenue.

B1's per-user figures rose at d0 while its delivery got dearer. Payer rate 7.7% to 11.9% and ARPU (d0) $0.6228 to $1.0941 on 30 and 39 payers, against a CPM up 13.0% and a registration up 23.5% to $1.068; its cohort return went 0.72 to 1.02.

shirtb took 72.7% of B1's report-clock spend and is the one per-asset row above the bar on both days, 0.37 on 16 payers to 0.90 on 32, with 9.9% of its money in one basket; madapp reads 1.36 on 7.

HR2's per-user fall is the first of its that clears the bar on both sides. ARPU (d0) $1.1453 to $0.8545 on 36 and 49 payers, its payer rate 10.6% to 11.1% and its ARPPU $10.85 to $7.69, so the whole of that fall is basket size.

The estimator reads both arms at H=24 and they do not separate. shirtb 0.68 and madapp 1.61 at 83% coverage, intervals 0.44 to 0.97 and 0.66 to 2.71; the audiences differ in value by a third, and shirtb pays 54% more per impression.

Dicey2 still cannot answer a revenue question. Twelve payers at d0 on $101.16, a CPM up 40.0% to $18.48 and a cohort return of 0.69 against 1.82.

An 8 h page reads the per-user change as a direction, and this day is the measure of why. The morning read an 18.2% fall in ARPU that closed at 3.3%, because the two cohorts paid on different clocks inside their first day.

Nothing changed in B1's campaign inside either day. Both sit in R3-400 on the same two creatives, so this pair is the cleanest roster-matched comparison the series has had.

Vloom recorded within-day revenue for the first time since it stopped buying, $10.62 from two people who arrived on 09-04 against no spend anywhere, and the 24 h cohort closed at the same figure.

5Reproduce

Staged in _runs/2026-09-06_2026-09-04_daily_auto/ at 00:24 UTC on 09-06, after the 24 h horizon closed at 00:00 UTC; this is the day's 24 h publication and it rewrites the 8 h page issued from _runs/2026-09-05_2026-09-04_daily_auto/ at 08:24 UTC on 09-05.

# the pull, both ledgers, every config and every instrument, in one command
python -m ad_ops.daily_pipeline --day 2026-09-04

# the country cut the figures are drawn from, then the figures, as the pipeline issues them
python -m ad_ops.country_json --run <run> --days 2026-09-03,2026-09-04
python -m ad_ops.report_charts --config <run>/run_config_sinai_full.json \
    --day 2026-09-04 --prev 2026-09-03 --days 2026-09-01,2026-09-02,2026-09-03,2026-09-04 \
    --app sinai_ww --country-json <run>/country_sinai_ww.json \
    --out ad_ops/figures/2026-09-04_sinai --md <run>/chartblock_sinai_ww.md \
    --relpath figures/2026-09-04_sinai

Payment cut 2026-09-06 00:19 UTC for Sinai and 00:13 UTC for Vloom; the matched-hours funnel on the META clock runs on a 2026-09-05 17:19 PT cut. B1's campaign activity history was requested over Sep 2 to Sep 6 and the range the page actually rendered, Sep 2 2026 to Sep 5 2026, was asserted before any row of it was read. HR2 exported nothing to this run at any grain, which is why it is in every day-grain and hourly table out of the archive and in none of the others.

6Appendix — definitions

The report clock is UTC. Meta stamps its exports in each ad account's own zone, B1, KBM1, Dicey2 and HR2 on UTC-7 and HR1 and the Nomad Node pair on UTC+0, so every delivery row on this page is restitched from the hourly exports onto UTC. The campaign section, the per-asset section and table C are the exceptions and say so where they appear.

The three revenue measures, all on our own payment records:

measurecountsmoves after the day closes?
booked (bROAS)revenue that arrived inside the report day, whatever day its payer installedsettles a few hours after midnight, then fixed
within-day (wROAS)installed that day and paid before that day closednever. Sealed at midnight
cohort @ 24h (cROAS)that day's installs, counted within 24 hours of each person's own installrises until every install has lived 24 hours, then frozen

A day is published twice. The morning page carries the 8 h cohort, the only horizon closed that early, and the evening page carries the 24 h cohort and completes it. This is the evening page: both horizons are closed on both days, and every per-user row appears at each of them.

Meta booked is Meta's own conversion value over Meta's own spend, both restitched onto the report clock from the same hourly export. It is struck through on any day where Meta booked more purchases than the payment processor processed, and nothing is written in its place: the true figure is the booked row directly above it. Neither day on this page is struck.

A regime is a stretch over which the ad set's daily budget did not change. It is in a daily report because the budget is a geography dial on this account: raising it buys wider and cheaper countries, so two days either side of a change are not buying the same audience.

The 16-payer bar is decision #37: a row carrying fewer than sixteen distinct payers is a direction and never a magnitude.

Internal — noindex. Not for distribution outside the team.