月度投放报告 — 2026 年 7 月

这个 App 的买量从 2026-07-23 开始,所以这里的 7 月只有九天,2026-07-23 至 2026-07-31,前面没有任何 周期可以对比。当时只有 B1 一个广告账户在买量,第二个账户从 2026-08-11 起才开始,在本周期之后。九天 里 B1 在八个广告系列上花了 $2,455.00,拿到 4,561 次安装、单次安装成本 $0.538、注册 5,069 次;入账 收入 $3,125.48,当日回收 $2,344.02,同期群 24 小时口径已经走满,为 $2,697.93。账户自己的投放导 出里,7 月的花费比本报告统计的多,因为账户上还跑过另一个 App 的测试和一笔代投;金额见复现一节。在 账户时区上,我们自己的记录为 Meta 入账口径的 1.07 倍,落在有记录的 0.91–1.06 区间之外。术语定义、 报表时区与三种收入口径见附录。

1总览

2026-07-23 → 2026-07-31
天数9
花费$2,455.00
展示次数347,058
CPM$7.07
链接点击量20,124
CTR5.80%
安装次数 (Meta)4,561
点击→安装22.7%
CPI$0.538
注册数 (自有)5,069
入账收入$3,125.48 (1.27) (Meta 1.18)
当日回收$2,344.02 (0.95)
同期群 24 小时$2,697.93 (1.10),覆盖率 100%
当日回收付费用户233

同期群价值 —— d0 / d3 / d7 的 ARPU 与 ARPPU

按安装日计算,从每个用户自己的安装时刻起算:dN 是他自己的前 (N+1) x 24 小时。ARPU 取自 Adjust,也就是合作方后台看到的那一套;ARPPU 取自我们自己的支付流水,因为 Adjust 没有累计去重付费人数这个字段。窗口还没走完的档位留空,不写 0。

应用ARPU (d0)ARPU (d3)ARPU (d7)ARPPU (d0)ARPPU (d3)ARPPU (d7)安装数
Vloom$0.4861$0.6372$0.7713$10.21$11.35$12.465,524

九天里各广告系列的花费占比: T3(7 月 25 日)55.2%,旧广告系列 14.6%,T2(7 月 25 日)10.9%, T1(7 月 25 日)8.3%,德语区(DE/AT/CH)4.1%,T1(7 月 23 日)2.9%,T2(7 月 23 日)2.3%,T3(7 月 23 日)1.7%。

这个月没有在当天把自己的钱赚回来。 当日回收回报 0.95;入账口径的 1.27 是继承下来的老同期群。

这是本系列里唯一一个不是下限的同期群读数。 H=24 的覆盖率是 100%,$2,697.93 已经冻结。

一个广告系列就是这个月的大部分。 T3(7 月 25 日)在 $2,455.00 里拿走了 $1,355.94,在 347,058 次展示里拿走了 284,604 次。账户级比率就是它自己的比率。

八笔买量之间,单次安装成本跨了两个数量级,$0.324 到 $18.098。合并的账户比率读的是当时在跑哪些 测试。

2分小时,2026-07-23 至 2026-07-31

本节是整个账户,按报表时区,把九天合并在一起看。横轴是一天中的小时,不是时间轴:每一格里装着九 个小时的买量,每天各出一个,这样做才让一天之内的形状读得出来,也让任何一行都不会被当成一次事件来读 。投放列的分子分母都取自 Meta 的计量口径。付费率、ARPU 与 ARPPU 两侧都是自有口径。两套安装数从不交 叉相除。ours 是该格里我们自己的注册数,cov 是该格同期群在 H=24 上的覆盖率。

小时花费展示CPM点击CTR安装点击→安装CPIourscov入账bROAS当日回收wROASwPay同期群cROAS
0$52.966,1008.684266.98%5212.2%$1.01853100%$38.710.73$0.000.000$0.000.00
1$74.9911,4896.536405.57%11117.3%$0.676123100%$88.031.17$52.960.719$52.960.71
2$85.6012,0617.106445.34%15824.5%$0.542165100%$40.620.47$68.160.808$68.160.80
3$78.7412,0056.566295.24%12920.5%$0.610140100%$156.151.98$166.432.1115$166.432.11
4$85.4315,3855.558145.29%13116.1%$0.652144100%$107.071.25$90.901.0613$100.221.17
5$109.1218,4255.921,0035.44%19819.7%$0.551218100%$117.021.07$131.531.2116$144.251.32
6$116.8321,5825.411,0795.00%25123.3%$0.465280100%$129.151.11$132.991.1413$136.781.17
7$106.5318,6955.709375.01%21923.4%$0.486247100%$235.862.21$183.851.7321$203.851.91
8$116.3621,0515.531,0494.98%22921.8%$0.508256100%$181.701.56$170.891.4716$174.451.50
9$115.8719,6985.881,0505.33%23922.8%$0.485272100%$144.231.24$41.770.365$41.770.36
10$119.2518,3976.481,0425.66%23622.6%$0.505271100%$52.020.44$39.470.335$46.590.39
11$116.3716,1217.228855.49%21324.1%$0.546236100%$196.421.69$225.411.9415$289.742.49
12$107.5013,4997.967925.87%18423.2%$0.584213100%$81.370.76$94.780.8810$113.421.06
13$111.1614,2187.828265.81%19323.4%$0.576202100%$177.241.59$63.930.589$63.930.58
14$129.3414,0179.238385.98%19022.7%$0.681211100%$225.721.75$211.671.6412$229.791.78
15$141.1317,7137.971,1166.30%26623.8%$0.531285100%$169.221.20$146.261.0413$152.881.08
16$165.9023,9926.911,3735.72%32023.3%$0.518365100%$129.420.78$110.240.667$121.000.73
17$149.5821,6816.901,3386.17%32124.0%$0.466363100%$203.441.36$143.320.9616$174.061.16
18$132.8918,5337.171,1466.18%30026.2%$0.443339100%$129.580.98$80.000.6010$151.091.14
19$93.4712,3407.578667.02%24127.8%$0.388274100%$126.281.35$61.930.667$77.110.82
20$83.636,53212.805167.90%15129.3%$0.554167100%$122.251.46$59.460.714$72.420.87
21$52.165,00810.423867.71%6817.6%$0.76777100%$108.372.08$25.220.483$41.320.79
22$47.693,49213.663229.22%6419.9%$0.74564100%$83.171.74$32.090.674$59.211.24
23$62.505,02412.444078.10%9723.8%$0.644104100%$82.451.32$10.750.172$16.510.26
合计$2,455.00347,0587.0720,1245.80%4,56122.7%$0.5385,069100%$3,125.481.27$2,344.020.95233$2,697.931.10

⚠ 每一格的当日回收付费用户数在 0 到 21 之间,只有四格到了 16。合计行看水平,分格只看形状。

每一格的覆盖率都是 100%:这个月已经过去五周才被读。

一天的两头买到的是两种东西。 千次展示成本第 6 小时 5.41、第 22 小时 13.66;点击率的方向正好相反。

最便宜的安装在下午晚些时候,最贵的在午夜前后:第 19 小时 $0.388,第 0 小时 $1.018;花费的高点 在第 16 小时。

四格过了 16 付费用户门槛,而它们说法不一: 第 5、7、8、17 小时的回报是 1.21、1.73、1.47 与 0.96。

与上一个周期对比

它前面没有周期。买量从 2026-07-23 开始,那正是本报告的第一天,所以逐小时配对、它的符号检验,以及把 上一个周期叠加上来的那张图,都没有对象可比,本月一个都不存在。本节全部是九天的水平读数。本系列能做 的第一次对比,是 8 月对 7 月。

2026-08-16T16:04:18.346126 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 3 6 9 12 15 18 21 hour (UTC) 0 20 40 60 80 100 120 140 160 USD hour of day, pooled across the 9 days of the period — a pooling, not a timeline Spend by hour 0 3 6 9 12 15 18 21 hour (UTC) 5000 10000 15000 20000 25000 impressions Impressions and installs by hour impressions installs 50 100 150 200 250 300 installs (Meta Leads) The period's motion — whole account, 2026-07-23..2026-07-31 (UTC)
本周期走势:分小时花费,以及展示与安装对照
2026-08-16T16:04:18.508556 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 4 8 12 16 20 hour (UTC) 5 6 7 8 9 CTR % CTR 0 4 8 12 16 20 hour (UTC) 6 8 10 12 14 USD per 1,000 impressions CPM 0 4 8 12 16 20 hour (UTC) 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install 0 4 8 12 16 20 hour (UTC) 0.4 0.5 0.6 0.7 0.8 0.9 1.0 USD Cost per install Delivery by hour — whole account, 2026-07-23..2026-07-31 (UTC)
全账户分小时的 CTR、CPM、点击→安装与 CPI

3分国家

2026-08-23T13:56:11.046309 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ IN MX US GB DE MY CH AU CA ID BR SA 0 200 400 600 800 1000 1200 USD delivery raked onto the report clock against two measured margins; revenue is the report period Spend against within-day revenue by country — 2026-07-01..2026-07-31 spend within-day revenue
分国家的花费与当日回收对照
2026-08-23T13:56:11.084106 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 — 2026-07-01..2026-07-31 (report clock)
分国家的单次安装成本

(一)大盘分国家,2026-07-01..2026-07-31。 按报表时区。 Meta 不提供小时 × 国家的交叉表,所以每一份国家切片都要从它自己的账户日折算到 UTC;自有口径本就在 UTC 上。

国家花费占比展示次数CPM注册数 (自有)单次注册成本当日回收wROAS付费用户
IN$1,474.6160.1%301,292$4.893,597$0.410$1,561.561.06215
MX$236.129.6%16,022$14.74295$0.800$162.290.6921
US$143.055.8%1,575$90.8072$1.987$126.800.898
GB$85.253.5%1,362$62.5939$2.186$0.000.000
DE$57.752.4%1,049$55.0426$2.221$101.661.763
MY$53.652.2%3,419$15.6950$1.073$5.380.101
CH$42.131.7%1,011$41.693$14.044$34.650.821
AU$36.031.5%578$62.3015$2.402$0.000.000
CA$32.051.3%669$47.9131$1.034$19.940.624
ID$27.231.1%3,947$6.90113$0.241$4.990.182
BR$19.730.8%1,369$14.41114$0.173$5.130.261
SA$19.650.8%1,073$18.3225$0.786$0.000.000

其余 163 个国家未列出, $227.74 合计(占本表 9.3%)。

上表中没有出现的账户: Nomad Node, KBM1, HR1. 它们的花费在每一张日粒度的表里都有;缺的只是国家切片。

(二)同一切片,每一个账户,2026-07-01..2026-07-31.

B1

国家花费占比展示次数CPM注册数 (自有)单次注册成本当日回收wROAS付费用户
IN$1,474.6160.1%301,292$4.893,597$0.410$1,561.561.06215
MX$236.129.6%16,022$14.74295$0.800$162.290.6921
US$143.055.8%1,575$90.8072$1.987$126.800.898
GB$85.253.5%1,362$62.5939$2.186$0.000.000
DE$57.752.4%1,049$55.0426$2.221$101.661.763
MY$53.652.2%3,419$15.6950$1.073$5.380.101
CH$42.131.7%1,011$41.693$14.044$34.650.821
AU$36.031.5%578$62.3015$2.402$0.000.000
CA$32.051.3%669$47.9131$1.034$19.940.624
ID$27.231.1%3,947$6.90113$0.241$4.990.182
BR$19.730.8%1,369$14.41114$0.173$5.130.261
SA$19.650.8%1,073$18.3225$0.786$0.000.000

其余 163 个国家未列出, $227.74 合计(占本表 9.3%)。

Nomad Node —— 没有分国家的表:它这一天什么都没有投出去,是实测的零。

KBM1 —— 没有分国家的表:它这一天什么都没有投出去,是实测的零。

HR1 —— 没有分国家的表:它这一天什么都没有投出去,是实测的零。

整个区间的对账。 每一个分国家单元格都是一次跨小时的分配;每一个总计都是实测的。每个账户折算出来的国家总计都等于它这些天买到的花费:B1 $2,455.00。逐日的对账——点名每个数字是从哪几个账户日折算来的——在各日报页上。

4汇总:本周期

本节按 UTC 时钟合并九天。本节平时会带的两样东西都需要第二个周期,而这里没有:一是带双比例检验的合 并逐小时漏斗,二是结构中性的账户行,它要把一个周期的各广告系列数字放到另一个周期的展示结构上重算。 两者都不出现。剩下的是整个月的水平读数,而这一部分本来就不需要对比才成立。

三个收入口径并列

口径2026-07-23 → 2026-07-31本期结束后还会动吗
入账收入$3,125.48 (1.27) (Meta 1.18)已结算
当日回收$2,344.02 (0.95)永远不动
同期群 24 小时$2,697.93 (1.10),覆盖率 100%已冻结,覆盖完整

三个口径对这个月的说法差了大约三分之一的收入。 入账口径会算上更早安装的人;同期群 24 小时口 径对一天中的时点中性;当日回收口径动不了。

只作参考、从不排次序:同一批人的迄今同期群收入是 $4,862.31,在 2026-08-16 的付款数据截止时间上 回报 1.98。它说明这批人的钱有多少是在第一天之后才到的,它与任何其他周期都不可比。

入账收入按付费用户已安装的天数拆分

安装于2026-07-23 → 2026-07-31
当日$2,344.02 (75.0%)
前 1 日$349.19 (11.2%)
前 2 日$285.86 (9.1%)
前 3 日$74.17 (2.4%)
前 4 日$45.44 (1.5%)
前 5 日$21.04 (0.7%)
前 6 日$5.76 (0.2%)

这张表的第一行就是当日回收的合计,只是换了一个方向看,这也说明入账与当日回收之间的差额恰好等于继 承下来的老同期群,别无其他。放到一个周期上,这个拆分就不再是钱落在哪一天,而变成了一条衰减曲线。

7 月入账的钱里有四分之三来自当天到达的人,尾巴在三天之内就没了 —— 默认视野取 H=24 依据的正 是这个形状。

2026-08-16T16:04:18.763779 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ booked within-day cohort @ 24h 0 500 1000 1500 2000 2500 3000 USD 1.27 0.95 1.10 the number above each bar is that measure's ROAS against the same period's spend The three revenue measures, 2026-07-23..2026-07-31 (UTC) 2026-07-23..2026-07-31
两个周期的三个收入口径,柱顶标注 ROAS
2026-08-16T16:04:18.806564 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 2026-07-23..2026-07-31 0 500 1000 1500 2000 2500 3000 USD the bottom segment IS the within-day figure: the same number seen twice Booked revenue by how long its payer had been installed (UTC) paid the day they installed one day after two days after three days after four or more days after
入账收入按付费用户已安装的天数拆分

5各广告系列并列

投放列的分子分母都取自 Meta 的计量口径,所有收入口径都是自有的。同期群视野 H=24h。报表时区为 UTC。只按 CPM、单次点击成本与 CPI 排次序 — 这里地域是被有意选择的处理变量,正是这一点让这 几项在广告系列之间可比,而收入仍然需要 16 个付费用户(DECISION_LOG.md #37、#48),未过门槛的广告 系列只能看方向,不能看幅度。

广告系列花费占比展示展示占比CPM点击CTR单次点击成本安装CPI当日回收wROAS付费用户
T3(7 月 25 日,印度/东南亚)$1,355.9455.2%284,60482.0%4.7614,2134.99%$0.0952,875$0.472$1,316.840.97142
旧广告系列(08-03 停投)$358.2014.6%29,5928.5%12.103,42311.57%$0.1051,106$0.324$591.311.6561
T2(7 月 25 日)$268.6310.9%18,2535.3%14.721,0925.98%$0.246298$0.901$144.060.5416
T1(7 月 25 日)$203.938.3%2,5910.7%78.7135413.66%$0.576102$1.999$87.870.436
德语区(DE/AT/CH)$99.984.1%2,1400.6%46.72833.88%$1.20523$4.347$127.001.272
T1(7 月 23 日)$72.392.9%1,2960.4%55.86715.48%$1.0204$18.098$0.000.000
T2(7 月 23 日)$55.252.3%2,6800.8%20.6229611.04%$0.18735$1.579$9.110.161
T3(7 月 23 日)$40.681.7%5,9021.7%6.8959210.03%$0.069118$0.345$67.821.675

表里每一个广告系列都有投放。当一个已经关掉的广告系列在它没有产生的花费上落下转化时,工具会打印一行 "有入账、无投放",这里它一行也没有打印,所以账户合计与八个标签页不用做任何剔除就能对上。

价格按 T1/T2/T3 的次序走,价值反着走 —— 千次展示成本 4.76 到 78.71,单次安装 $0.324 到 $18.098,两者都按分组排序。

三个广告系列到了 16 个付费用户,其中只有一个赚钱: 旧广告系列 1.65,对上 0.97 与 0.54。

旧广告系列是唯一赚回来的一个,而它只跑了九天里的两天。

T1(7 月 23 日)用 $72.39 买了 1,296 次展示,拿到四次安装,单次 $18.098。它拿到标签页靠的只是 花费占比。

本节的固定图表组画的是跨天的广告系列花费占比、CPM 与 CPI。图表生成器在这一层不出任何图,所以第 5 节只有表格。

6广告系列内部

本节是广告层级。先是账户全部素材的清单,按跨广告系列的花费排序;随后每个广告系列一个标签页,在它内 部把同一套表重新切一遍。标签页按花费从高到低排列,每一页都带同样的四节。

一个标签页带什么。 图限定在该广告系列上,表只到素材层;没有任何结论只存在于标签页里。

A. 分素材 — 投放

分子分母都取自 Meta 的计量口径。我们自己的注册数不进入这张表。

素材花费展示CPM点击CTR安装点击→安装CPI
earner/bodysuit$312.9326,15911.962,96811.35%93431.47%$0.335
T3b/utility_6s$305.3348,3536.312,6535.49%52119.64%$0.586
T3b/sasian_spice5_15s$282.8867,0804.222,1843.26%45320.74%$0.624
T3b/sasian_bodysuit_10s$281.6954,3875.182,8365.21%52718.58%$0.535
T3b/sasian_bikini_6s$189.2258,0643.263,1455.42%75423.97%$0.251
T3b/sasian_bikini_10s$120.1434,3033.501,8415.37%33218.03%$0.362
T1b/utility_6s#2$74.1887784.5814816.88%5738.51%$1.301
T1b/utility_15s$73.9888883.3112714.30%3527.56%$2.114
T2b/utility_6s#3$68.173,56819.112446.84%7229.51%$0.947
T2b/utility_15s#2$52.843,19916.521705.31%3721.76%$1.428
T2b/latina_es_bikini_10s$42.734,7069.082916.18%9934.02%$0.432
T3b/sasian_slip_dress$39.935,2577.604067.72%8119.95%$0.493
earner/pool5$36.632,73713.3837213.59%14438.71%$0.254
T1a/utility_sexy$30.9755955.40244.29%00.00%—
T1a/control_sexy$28.9534982.954011.46%410.00%$7.237
T3a/utility_sexy#2$28.904,1736.9347811.45%11423.85%$0.254
T2a/control_sexy#2$28.181,30321.631219.29%1310.74%$2.168
DE/de_german_mature_6s$27.4045660.09296.36%827.59%$3.425
T2a/utility_sexy#3$27.071,37719.6617512.71%2212.57%$1.230
T3b/sasian_fishnet_top_6s$25.154,1356.081674.04%4627.54%$0.547
DE/ch_arab_gulf_6s$21.7549843.67122.41%18.33%$21.750
T2b/latina_es_slip_dress$19.391,26215.36634.99%2742.86%$0.718
T2b/control_10s$14.5091015.93556.04%35.45%$4.833
T3b/sasian_bodysuit_6s$14.501,6878.60714.21%1521.13%$0.967
T1b/utility_10s$14.3123261.68166.90%00.00%—
T1b/control_6s$13.9918874.411910.11%15.26%$13.990
T1b/control_15s$13.8321364.93198.92%315.79%$4.610
T1b/control_10s#2$13.6419370.672512.95%624.00%$2.273
T2b/control_6s#2$13.3581716.34415.02%717.07%$1.907
T2b/control_15s#2$13.1482915.85536.39%611.32%$2.190
T3b/control_15s#3$13.011,4838.771248.36%1814.52%$0.723
T3b/utility_10s#2$12.791,2879.9413810.72%2719.57%$0.474
T2b/utility_10s#3$12.7962820.37447.01%920.45%$1.421
T3b/utility_15s#3$12.661,13711.13998.71%1616.16%$0.791
T3b/control_10s#3$12.551,2679.911058.29%109.52%$1.255
T3b/control_6s#3$12.181,5797.7116410.39%3219.51%$0.381
T3a/control_sexy#3$11.781,7296.811146.59%43.51%$2.945
T3b/sasian_corset_6s$11.622,0595.641326.41%2015.15%$0.581
T2b/latina_es_bodysuit_10s$11.0182213.39556.69%1323.64%$0.847
T3b/sasian_offshoulder$9.321,0608.79716.70%79.86%$1.331
earner/bikini$8.6469612.418311.93%2833.73%$0.309
T3b/sasian_jean_shorts$6.688288.07465.56%919.57%$0.742
T1a/control_sexy#4$6.5617836.8542.25%00.00%—
T2b/latina_es_offshoulder$6.0430819.61113.57%327.27%$2.013
T1a/utility_sexy#4$5.9121028.1431.43%00.00%—
DE/de_arab_gulf_10s$5.798369.7656.02%120.00%$5.790
DE/ch_german_10s$5.2010250.9821.96%00.00%—
DE/ch_latina_true_6s$5.1710648.7710.94%00.00%—
DE/de_turkish_mature_10s$4.7112637.3853.97%360.00%$1.570
T2b/latina_es_bikini_6s$4.154449.35214.73%733.33%$0.593
DE/german_6s$4.0710040.7066.00%00.00%—
DE/ch_german_mature_6s$4.0210339.0300.00%0——
DE/turkish_mature$3.955769.3047.02%4100.00%$0.987
DE/ch_arab_gulf_10s$3.6614625.0710.68%1100.00%$3.660
T3b/sasian_fishnet_top_10s$3.6335010.37123.43%325.00%$1.210
T2b/latina_es_jean_shorts$3.2623413.93135.56%646.15%$0.543
DE/german_mature$3.018933.8266.74%233.33%$1.505
T3b/sasian_corset_10s$2.662889.24196.60%421.05%$0.665
DE/de_arab_gulf_6s$2.512696.5427.69%150.00%$2.510
T2b/latina_es_corset_10s$2.2519211.72126.25%541.67%$0.450
T2b/latina_es_bodysuit_6s$2.0013714.60107.30%330.00%$0.667
DE/german$1.906031.6700.00%0——
T2b/latina_es_spice5_15s$1.467718.9622.60%150.00%$1.460
DE/latina_true$1.362164.7614.76%00.00%—
DE/de_turkish_10s$1.163038.67620.00%233.33%$0.580
DE/de_pool5_15s$1.042149.5200.00%0——
DE/de_asian_mature_6s$0.993726.7600.00%0——
T2b/latina_es_fishnet_top_6s$0.797510.5334.00%00.00%—
DE/ch_german_mature_10s$0.763025.3313.33%00.00%—
DE/turkish_6s$0.542422.5014.17%00.00%—
T2b/latina_es_corset_6s$0.52774.29228.57%00.00%—
DE/ch_latina_mature_10s$0.42760.0000.00%0——
DE/de_latina_true_6s$0.25550.0000.00%0——
T2b/latina_es_fishnet_top_10s$0.24386.3225.26%00.00%—
DE/ch_german_6s$0.14435.0000.00%0——
DE/ch_asian_mature_6s$0.11336.67133.33%00.00%—
DE/ch_pool5_15s$0.07417.5000.00%0——

六条素材拿走了这个月的大部分,其中四条来自南亚素材包。 七十七条素材里有四十六条各自拿到不足 $13。

这张表的上下两半是两个不同的竞价环境。 投印度的素材每千次展示 3.26 到 5.18;T1a/ 与 T1b/ 是 70.67 到 84.58。

2026-08-16T16:04:18.899904 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 4 6 8 10 12 14 16 CTR % hour of day, pooled across the 9 days of the period — a pooling, not a timeline CTR by hour — every asset, 2026-07-23..2026-07-31 (UTC) earner/bodysuit T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T3b/sasian_slip_dress earner/pool5 T3b/sasian_corset_6s
分素材的分小时 CTR
2026-08-16T16:04:19.057409 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 % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install by hour — every asset, 2026-07-23..2026-07-31 (UTC) earner/bodysuit T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T3b/sasian_slip_dress earner/pool5 T3b/sasian_corset_6s
分素材的分小时点击→安装
2026-08-16T16:04:19.133383 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.5 1.0 1.5 2.0 USD hour of day, pooled across the 9 days of the period — a pooling, not a timeline Cost per install by hour — every asset, 2026-07-23..2026-07-31 (UTC) earner/bodysuit T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T3b/sasian_slip_dress earner/pool5 T3b/sasian_corset_6s
分素材的分小时单次安装成本
2026-08-16T16:04:18.979701 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 4 6 8 10 12 14 16 USD per 1,000 impressions hour of day, pooled across the 9 days of the period — a pooling, not a timeline CPM by hour — every asset, 2026-07-23..2026-07-31 (UTC) earner/bodysuit T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T3b/sasian_slip_dress earner/pool5 T3b/sasian_corset_6s
分素材的分小时 CPM

B. 分素材 — 三种收入口径

每一列都是自有口径。Meta 只能表达 入账,因此它跟在那一列旁边的括号里,永远不单独成行。付费用户是 去重的人,不是交易笔数。同期群视野 H=24h,本周期覆盖率 100%。报表时区为 UTC。

素材入账bROAS (Meta)当日回收wROAS同期群@HcROAS付费用户付费率ARPUARPPU最大付费用户占比
earner/bodysuit$508.491.62 (1.71)$508.491.62$603.141.93505.0%0.50410.178.1%
T3b/utility_6s$483.711.58 (1.11)$226.490.74$314.431.03305.1%0.3837.5510.4%
T3b/sasian_spice5_15s$300.891.06 (0.96)$217.760.77$243.620.86315.9%0.4167.025.9%
T3b/sasian_bodysuit_10s$358.251.27 (1.35)$280.050.99$325.381.16335.4%0.4618.4914.8%
T3b/sasian_bikini_6s$401.492.12 (1.87)$283.401.50$316.271.67182.2%0.34115.7433.1%
T3b/sasian_bikini_10s$92.340.77 (0.99)$79.460.66$79.460.6671.9%0.21511.3552.0%
T1b/utility_6s#2$66.930.90 (0.90)$66.930.90$66.930.9035.2%1.15422.3152.3%
T1b/utility_15s$20.940.28 (0.28)$20.940.28$20.940.2837.9%0.5516.9852.4%
T2b/utility_6s#3$27.340.40 (0.40)$27.340.40$27.340.4034.0%0.3649.1158.2%
T2b/utility_15s#2$28.430.54 (0.64)$21.630.41$25.030.4725.6%0.60110.8173.6%
T2b/latina_es_bikini_10s$90.022.11 (2.09)$68.401.60$71.801.6899.2%0.6987.6033.2%
T3b/sasian_slip_dress$95.582.39 (1.83)$67.631.69$76.951.9377.0%0.6769.6631.0%
earner/pool5$77.782.12 (2.18)$77.782.12$77.782.12106.4%0.4997.7823.1%
T1a/utility_sexy$0.000.00 (0.00)$0.000.00$0.000.000————
T1a/control_sexy$9.390.32 (0.32)$0.000.00$9.390.3200.0%0.000——
T3a/utility_sexy#2$93.573.24 (2.23)$67.822.35$71.382.4753.7%0.50213.5651.7%
T2a/control_sexy#2$9.110.32 (0.32)$9.110.32$9.110.3217.7%0.7019.11100.0%
DE/de_german_mature_6s$101.663.71 (3.36)$92.363.37$98.103.58112.5%11.54592.36100.0%
T2a/utility_sexy#3$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T3b/sasian_fishnet_top_6s$94.543.76 (5.02)$88.783.53$94.543.76813.6%1.50511.1046.6%
DE/ch_arab_gulf_6s$47.022.16 (2.15)$34.651.59$47.022.16150.0%17.32334.65100.0%
T2b/latina_es_slip_dress$5.710.29 (0.29)$5.710.29$5.710.2913.1%0.1785.71100.0%
T2b/control_10s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T3b/sasian_bodysuit_6s$43.132.97 (2.95)$18.631.29$27.951.93211.8%1.0969.3269.1%
T1b/utility_10s$0.000.00 (0.00)$0.000.00$0.000.000————
T1b/control_6s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T1b/control_15s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T1b/control_10s#2$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T2b/control_6s#2$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T2b/control_15s#2$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T3b/control_15s#3$9.320.72 (1.25)$9.320.72$9.320.7214.8%0.4449.32100.0%
T3b/utility_10s#2$38.212.99 (0.89)$11.520.90$11.520.9026.9%0.3975.7650.0%
T2b/utility_10s#3$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T3b/utility_15s#3$10.750.85 (1.30)$0.000.00$4.990.3900.0%0.000——
T3b/control_10s#3$12.881.03 (0.00)$3.560.28$3.560.28110.0%0.3563.56100.0%
T3b/control_6s#3$12.881.06 (0.00)$5.760.47$5.760.4712.7%0.1565.76100.0%
T3a/control_sexy#3$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T3b/sasian_corset_6s$24.502.11 (0.79)$24.502.11$24.502.1114.5%1.11424.50100.0%
T2b/latina_es_bodysuit_10s$9.110.83 (0.00)$0.000.00$0.000.0000.0%0.000——
T3b/sasian_offshoulder$0.000.00 (2.12)$0.000.00$0.000.0000.0%0.000——
earner/bikini$5.040.58 (0.58)$5.040.58$5.040.5813.2%0.1635.04100.0%
T3b/sasian_jean_shorts$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T1a/control_sexy#4$0.000.00 (0.00)$0.000.00$0.000.000————
T2b/latina_es_offshoulder$46.487.70 (5.12)$20.993.48$20.993.48133.3%6.99720.99100.0%
T1a/utility_sexy#4$0.000.00 (0.00)$0.000.00$0.000.000————
DE/de_arab_gulf_10s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
DE/ch_german_10s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/ch_latina_true_6s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/de_turkish_mature_10s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T2b/latina_es_bikini_6s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
DE/german_6s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/ch_german_mature_6s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/turkish_mature$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
DE/ch_arab_gulf_10s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T3b/sasian_fishnet_top_10s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T2b/latina_es_jean_shorts$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
DE/german_mature$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T3b/sasian_corset_10s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
DE/de_arab_gulf_6s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T2b/latina_es_corset_10s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T2b/latina_es_bodysuit_6s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
DE/german$0.000.00 (0.00)$0.000.00$0.000.000————
T2b/latina_es_spice5_15s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
DE/latina_true$0.000.00 (0.00)$0.000.00$0.000.000————
DE/de_turkish_10s$0.000.00 (4.30)$0.000.00$0.000.0000.0%0.000——
DE/de_pool5_15s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/de_asian_mature_6s$0.000.00 (0.00)$0.000.00$0.000.000————
T2b/latina_es_fishnet_top_6s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/ch_german_mature_10s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/turkish_6s$0.000.00 (0.00)$0.000.00$0.000.000————
T2b/latina_es_corset_6s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/ch_latina_mature_10s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/de_latina_true_6s$0.000.00 (0.00)$0.000.00$0.000.000————
T2b/latina_es_fishnet_top_10s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/ch_german_6s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/ch_asian_mature_6s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/ch_pool5_15s$0.000.00 (0.00)$0.000.00$0.000.000————

16 付费用户门槛(#37):上表中付费用户数不足 16 的行只能看方向,不能看幅度。

整个月只有五条素材过了门槛,其中四条在同一个广告系列里。 它们的当日回收回报在 0.74 到 1.62 之间。

earner/bodysuit 是本月最不集中的一行收入,也是唯一一个读得出来、又在 1.00 以上的回报,最大 的单个付费用户占 8.1%,对上 33.1%。

最大付费用户占比这一列,把本页大多数高回报判了出局 —— 3.37、3.48 与 3.53 各自只站在一到八个 付费用户上。

2026-08-16T16:04:19.700541 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0 2 4 6 8 10 12 14 16 payers / installs % Payer rate by day — every asset (UTC) earner/bodysuit T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T1b/utility_6s#2 T1b/utility_15s T2b/utility_6s#3 T2b/utility_15s#2 T2b/latina_es_bikini_10s T3b/sasian_slip_dress earner/pool5 T3a/utility_sexy#2 T2a/utility_sexy#3 T3b/sasian_fishnet_top_6s T2b/latina_es_slip_dress T3b/control_15s#3 T3b/utility_10s#2 T3b/control_6s#3 earner/bikini
分素材的每日付费率
2026-08-16T16:04:19.824816 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 within-day USD per install ARPU by day — every asset (UTC) earner/bodysuit T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T1b/utility_6s#2 T1b/utility_15s T2b/utility_6s#3 T2b/utility_15s#2 T2b/latina_es_bikini_10s T3b/sasian_slip_dress earner/pool5 T3a/utility_sexy#2 T2a/utility_sexy#3 T3b/sasian_fishnet_top_6s T2b/latina_es_slip_dress T3b/control_15s#3 T3b/utility_10s#2 T3b/control_6s#3 earner/bikini
分素材的每日 ARPU
2026-08-16T16:04:19.942814 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0 5 10 15 20 25 within-day USD per payer ARPPU by day — every asset (UTC) earner/bodysuit T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T1b/utility_6s#2 T1b/utility_15s T2b/utility_6s#3 T2b/utility_15s#2 T2b/latina_es_bikini_10s T3b/sasian_slip_dress earner/pool5 T3a/utility_sexy#2 T3b/sasian_fishnet_top_6s T2b/latina_es_slip_dress T3b/control_15s#3 T3b/utility_10s#2 T3b/control_6s#3 earner/bikini
分素材的每日 ARPPU
2026-08-16T16:04:20.070843 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0 1 2 3 4 revenue / spend Within-day ROAS by day — every asset (UTC) earner/bodysuit T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T1b/utility_6s#2 T1b/utility_15s T2b/utility_6s#3 T2b/utility_15s#2 T2b/latina_es_bikini_10s T3b/sasian_slip_dress earner/pool5 T3a/utility_sexy#2 T2a/utility_sexy#3 T3b/sasian_fishnet_top_6s T2b/latina_es_slip_dress T3b/control_15s#3 T3b/utility_10s#2 T3b/control_6s#3 earner/bikini
分素材的每日当日回收 ROAS
2026-08-16T16:04:20.224112 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 period (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 period are not plotted. Composite — every asset across every measure, 2026-07-23..2026-07-31 earner/bodysuit T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T1b/utility_6s#2 T1b/utility_15s T2b/utility_6s#3 T2b/utility_15s#2 T2b/latina_es_bikini_10s T3b/sasian_slip_dress earner/pool5 T3a/utility_sexy#2 T2a/utility_sexy#3 T3b/sasian_fishnet_top_6s T2b/latina_es_slip_dress T3b/utility_10s#2 T3b/control_6s#3 T3b/sasian_corset_6s earner/bikini
综合图:各素材在各口径上按本周期归一化后的表现

收入类图表中,落在 16 付费用户以下的点画成空心。当日安装不足 20 次的素材从综合图和地域图中剔除。

C. 分素材 — 国家结构

⚠ 本表按账户日切,而本页其余各表都在 UTC 时钟上。 Meta 不提供小时 × 国家的交叉表,因此分国家的投放数据只存 在于广告账户自己的 UTC-7 日上,无法重新拼接。它与上面的 A、B 两表相差七小时;本表只读结构,不与它 们比水平。

B 表里某一行读不读得出来,看这里。最后一列是该素材自身当日回收中来自美国的占比:当它远高于 该素材的美国花费占比时,这条素材的回报是它买到的那块地域的属性。

素材花费印度占比美国占比其他占比美国占其自身收入
earner/bodysuit$408.9932.8%4.7%62.6%8.6%
T3b/utility_6s$305.3398.9%0.0%1.1%0.0%
T3b/sasian_spice5_15s$282.88100.0%0.0%0.0%0.0%
T3b/sasian_bodysuit_10s$281.69100.0%0.0%0.0%0.0%
T3b/sasian_bikini_6s$189.22100.0%0.0%0.0%0.0%
T3b/sasian_bikini_10s$120.14100.0%0.0%0.0%0.0%
T1b/utility_6s#2$74.180.0%48.4%51.6%100.0%
T1b/utility_15s$73.980.0%42.5%57.5%52.4%
T2b/utility_6s#3$68.170.0%0.0%100.0%0.0%
T2b/utility_15s#2$52.840.0%0.0%100.0%0.0%
T2b/latina_es_bikini_10s$42.730.0%0.0%100.0%0.0%
T3b/sasian_slip_dress$39.93100.0%0.0%0.0%0.0%
earner/pool5$36.6327.3%8.2%64.5%6.4%
T1a/utility_sexy$30.970.0%51.2%48.8%—
T1a/control_sexy$28.950.0%57.5%42.5%—
T3a/utility_sexy#2$28.9079.4%0.0%20.6%0.0%
T2a/control_sexy#2$28.180.0%0.0%100.0%0.0%
DE/de_german_mature_6s$27.400.0%0.0%100.0%0.0%
T2a/utility_sexy#3$27.070.0%0.0%100.0%—
T3b/sasian_fishnet_top_6s$25.15100.0%0.0%0.0%0.0%
DE/ch_arab_gulf_6s$21.750.0%0.0%100.0%0.0%
T2b/latina_es_slip_dress$19.390.0%0.0%100.0%0.0%
T2b/control_10s$14.500.0%0.0%100.0%—
T3b/sasian_bodysuit_6s$14.50100.0%0.0%0.0%0.0%
T1b/utility_10s$14.310.0%46.3%53.7%—
T1b/control_6s$13.990.0%39.5%60.5%—
T1b/control_15s$13.830.0%37.4%62.6%—
T1b/control_10s#2$13.640.0%38.8%61.2%—
T2b/control_6s#2$13.350.0%0.0%100.0%—
T2b/control_15s#2$13.140.0%0.0%100.0%—
T3b/control_15s#3$13.0181.9%0.0%18.1%0.0%
T3b/utility_10s#2$12.7982.5%0.0%17.5%0.0%
T2b/utility_10s#3$12.790.0%0.0%100.0%—
T3b/utility_15s#3$12.6683.9%0.0%16.1%—
T3b/control_10s#3$12.5576.6%0.0%23.4%0.0%
T3b/control_6s#3$12.1861.7%0.0%38.3%0.0%
T3a/control_sexy#3$11.7857.8%0.0%42.2%—
T3b/sasian_corset_6s$11.62100.0%0.0%0.0%0.0%
T2b/latina_es_bodysuit_10s$11.010.0%0.0%100.0%—
T3b/sasian_offshoulder$9.32100.0%0.0%0.0%—
earner/bikini$150.3329.0%5.0%65.9%0.0%
T3b/sasian_jean_shorts$6.68100.0%0.0%0.0%—
T1a/control_sexy#4$6.560.0%21.2%78.8%—
T2b/latina_es_offshoulder$6.040.0%0.0%100.0%0.0%
T1a/utility_sexy#4$5.910.0%32.0%68.0%—
DE/de_arab_gulf_10s$5.790.0%0.0%100.0%—
DE/ch_german_10s$5.200.0%0.0%100.0%—
DE/ch_latina_true_6s$5.170.0%0.0%100.0%—
DE/de_turkish_mature_10s$4.710.0%0.0%100.0%—
T2b/latina_es_bikini_6s$4.150.0%0.0%100.0%—
DE/german_6s$4.070.0%0.0%100.0%—
DE/ch_german_mature_6s$4.020.0%0.0%100.0%—
DE/turkish_mature$3.950.0%0.0%100.0%—
DE/ch_arab_gulf_10s$3.660.0%0.0%100.0%—
T3b/sasian_fishnet_top_10s$3.63100.0%0.0%0.0%—
T2b/latina_es_jean_shorts$3.260.0%0.0%100.0%—
DE/german_mature$3.010.0%0.0%100.0%—
T3b/sasian_corset_10s$2.66100.0%0.0%0.0%—
DE/de_arab_gulf_6s$2.510.0%0.0%100.0%—
T2b/latina_es_corset_10s$2.250.0%0.0%100.0%—
T2b/latina_es_bodysuit_6s$2.000.0%0.0%100.0%—
DE/german$1.900.0%0.0%100.0%—
T2b/latina_es_spice5_15s$1.460.0%0.0%100.0%—
DE/latina_true$1.360.0%0.0%100.0%—
DE/de_turkish_10s$1.160.0%0.0%100.0%—
DE/de_pool5_15s$1.040.0%0.0%100.0%—
DE/de_asian_mature_6s$0.990.0%0.0%100.0%—
T2b/latina_es_fishnet_top_6s$0.790.0%0.0%100.0%—
DE/ch_german_mature_10s$0.760.0%0.0%100.0%—
DE/turkish_6s$0.540.0%0.0%100.0%—
T2b/latina_es_corset_6s$0.520.0%0.0%100.0%—
DE/ch_latina_mature_10s$0.420.0%0.0%100.0%—
DE/de_latina_true_6s$0.250.0%0.0%100.0%—
T2b/latina_es_fishnet_top_10s$0.240.0%0.0%100.0%—
DE/ch_german_6s$0.140.0%0.0%100.0%—
DE/ch_asian_mature_6s$0.110.0%0.0%100.0%—
DE/ch_pool5_15s$0.070.0%0.0%100.0%—

⚠ 在旧广告系列这几行上,七小时的偏移不是一处小毛病。 本表的窗口是账户自己的日,一直伸进 2026-08-01 UTC,而那时这个广告系列的预算又被调高了。它给出的是 earner/bodysuit $408.99 与 earner/bikini $150.33,A 表给出的是 $312.93 与 $8.64。只读这里的百分比,不要读这里的金额,更不要 拿它去对 A、B 两表。

T1b/utility_6s#2 买到一块地域并在那里赚到了钱,背后只有三个付费用户 —— 48.4% 的钱投在美 国,收入的 100.0% 来自美国。

真正有量的都是印度买量 —— T3b/sasian_ 包 100.0%、T3b/utility_6s 98.9%,里面没有一行是美 国结果。

2026-08-16T16:04:19.297510 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ earner/bodysuit T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s earner/bikini T3b/sasian_bikini_10s T1b/utility_6s#2 T1b/utility_15s T2b/utility_6s#3 T2b/utility_15s#2 T2b/latina_es_bikini_10s T3b/sasian_slip_dress earner/pool5 T3a/utility_sexy#2 T2a/utility_sexy#3 T3b/sasian_fishnet_top_6s T2b/latina_es_slip_dress T3b/utility_10s#2 T3b/control_6s#3 T3b/sasian_corset_6s 0 20 40 60 80 100 % of that asset's spend Where each asset bought — share of its own spend by region, 2026-07-23..2026-07-31 (META days) IN MX US GB DE MY other
各素材自身花费的地域结构
2026-08-16T16:04:19.423341 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ earner/bodysuit T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s earner/bikini T3b/sasian_bikini_10s T1b/utility_6s#2 T1b/utility_15s T2b/utility_6s#3 T2b/utility_15s#2 T2b/latina_es_bikini_10s T3b/sasian_slip_dress earner/pool5 T3a/utility_sexy#2 T2a/utility_sexy#3 T3b/sasian_fishnet_top_6s T2b/latina_es_slip_dress T3b/utility_10s#2 T3b/control_6s#3 T3b/sasian_corset_6s 0.0 0.5 1.0 1.5 2.0 2.5 3.0 USD per install Cost per install by region — every asset, 2026-07-23..2026-07-31 (META days) IN MX US GB DE MY
各素材分地域的单次安装成本
2026-08-16T16:04:19.513240 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ earner/bodysuit T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s earner/bikini T3b/sasian_bikini_10s T1b/utility_6s#2 T1b/utility_15s T2b/utility_6s#3 T2b/utility_15s#2 T2b/latina_es_bikini_10s T3b/sasian_slip_dress earner/pool5 T3a/utility_sexy#2 T2a/utility_sexy#3 T3b/sasian_fishnet_top_6s T2b/latina_es_slip_dress T3b/utility_10s#2 T3b/control_6s#3 T3b/sasian_corset_6s 0 50 100 150 200 250 USD Within-day revenue by region — every asset, 2026-07-23..2026-07-31 (META days) IN MX US DE MY
各素材分地域的当日回收
当前显示

T3(7 月 25 日,印度/东南亚) —— 这一行以下的每一节都是 T3(7 月 25 日,印度/东南亚) 的:当日、分小时、分国家、分广告系列、分素材,都记在 T3(7 月 25 日,印度/东南亚) 自己那本账、自己那些账户上。用上面的按钮切换。

2026-07-25 那一轮里投印度与东南亚的一条,也是这个月的主体:账户 $2,455.00 里的 $1,355.94,347,058 次展示里的 284,604 次,2,875 次安装、单次 $0.472,142 个当日回收付费用户,回报 0.97。它是唯一一个 付费用户数明显过了 16 付费用户门槛的广告系列,而即使在它内部,也只有四条素材过得了。

分小时、分国家与汇总

2026-08-16T16:05:15.434676 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 3 6 9 12 15 18 21 hour (UTC) 0 20 40 60 80 USD hour of day, pooled across the 9 days of the period — a pooling, not a timeline Spend by hour 0 3 6 9 12 15 18 21 hour (UTC) 2500 5000 7500 10000 12500 15000 17500 20000 impressions Impressions and installs by hour impressions installs 0 50 100 150 200 installs (Meta Leads) The period's motion — whole account, 2026-07-23..2026-07-31 (UTC)
本周期走势:分小时花费,以及展示与安装对照
2026-08-16T16:05:15.607966 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 4 8 12 16 20 hour (UTC) 4.50 4.75 5.00 5.25 5.50 5.75 6.00 CTR % CTR 0 4 8 12 16 20 hour (UTC) 4.5 5.0 5.5 6.0 USD per 1,000 impressions CPM 0 4 8 12 16 20 hour (UTC) 5 10 15 20 25 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install 0 4 8 12 16 20 hour (UTC) 0.4 0.6 0.8 1.0 1.2 1.4 1.6 USD Cost per install Delivery by hour — T3 (Jul 25, India/SEA), 2026-07-23..2026-07-31 (UTC)
T3 (Jul 25, India/SEA) 的分小时 CTR、CPM、点击→安装与 CPI
2026-08-16T16:05:15.731284 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ IN ID PH PK NG 0 200 400 600 800 1000 1200 1400 USD spend and installs are a META-day cut (no hour x country grid exists); revenue is the report period Spend against within-day revenue by country — 2026-07-23..2026-07-31 spend within-day revenue
分国家的花费与当日回收对照
2026-08-16T16:05:15.771785 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ IN ID PH PK 0.0 0.1 0.2 0.3 0.4 0.5 0.6 USD per install Cost per install by country — 2026-07-23..2026-07-31 (META days)
分国家的单次安装成本
2026-08-16T16:05:15.837023 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ booked within-day cohort @ 24h 0 250 500 750 1000 1250 1500 1750 2000 USD 1.46 0.97 1.13 the number above each bar is that measure's ROAS against the same period's spend The three revenue measures, 2026-07-23..2026-07-31 (UTC) 2026-07-23..2026-07-31
两个周期的三个收入口径,柱顶标注 ROAS
2026-08-16T16:05:15.891785 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 2026-07-23..2026-07-31 0 250 500 750 1000 1250 1500 1750 2000 USD the bottom segment IS the within-day figure: the same number seen twice Booked revenue by how long its payer had been installed (UTC) paid the day they installed one day after two days after three days after four or more days after
入账收入按付费用户已安装的天数拆分

A. 分素材 — 投放

分子分母都取自 Meta 的计量口径。我们自己的注册数不进入这张表。

素材花费展示CPM点击CTR安装点击→安装CPI
T3b/utility_6s$305.3348,3536.312,6535.49%52119.64%$0.586
T3b/sasian_spice5_15s$282.8867,0804.222,1843.26%45320.74%$0.624
T3b/sasian_bodysuit_10s$281.6954,3875.182,8365.21%52718.58%$0.535
T3b/sasian_bikini_6s$189.2258,0643.263,1455.42%75423.97%$0.251
T3b/sasian_bikini_10s$120.1434,3033.501,8415.37%33218.03%$0.362
T3b/sasian_slip_dress$39.935,2577.604067.72%8119.95%$0.493
T3b/sasian_fishnet_top_6s$25.154,1356.081674.04%4627.54%$0.547
T3b/sasian_bodysuit_6s$14.501,6878.60714.21%1521.13%$0.967
T3b/control_15s#3$13.011,4838.771248.36%1814.52%$0.723
T3b/utility_10s#2$12.791,2879.9413810.72%2719.57%$0.474
T3b/utility_15s#3$12.661,13711.13998.71%1616.16%$0.791
T3b/control_10s#3$12.551,2679.911058.29%109.52%$1.255
T3b/control_6s#3$12.181,5797.7116410.39%3219.51%$0.381
T3b/sasian_corset_6s$11.622,0595.641326.41%2015.15%$0.581
T3b/sasian_offshoulder$9.321,0608.79716.70%79.86%$1.331
T3b/sasian_jean_shorts$6.688288.07465.56%919.57%$0.742
T3b/sasian_fishnet_top_10s$3.6335010.37123.43%325.00%$1.210
T3b/sasian_corset_10s$2.662889.24196.60%421.05%$0.665

五条素材几乎拿走了全部,而这个广告系列内部的成本跨度很大:T3b/sasian_bikini_6s 的千次展示成本是 3.26、单次安装 $0.251,T3b/sasian_offshoulder 是 8.79 与 $1.331。六个 utility_/control_ 格是 07-25 那一轮建的;所有以 sasian_ 命名的,是 2026-07-26 加进一个合并广告组的十二条南亚包素材,所 以这两组之间不构成一次对等的比较。

2026-08-16T16:05:15.971696 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 3 4 5 6 7 8 CTR % hour of day, pooled across the 9 days of the period — a pooling, not a timeline CTR by hour — every asset, 2026-07-23..2026-07-31 (UTC) T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T3b/sasian_slip_dress T3b/sasian_corset_6s
分素材的分小时 CTR
2026-08-16T16:05:16.112715 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 5 10 15 20 25 30 35 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install by hour — every asset, 2026-07-23..2026-07-31 (UTC) T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T3b/sasian_slip_dress T3b/sasian_corset_6s
分素材的分小时点击→安装
2026-08-16T16:05:16.179466 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.5 1.0 1.5 2.0 USD hour of day, pooled across the 9 days of the period — a pooling, not a timeline Cost per install by hour — every asset, 2026-07-23..2026-07-31 (UTC) T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T3b/sasian_slip_dress T3b/sasian_corset_6s
分素材的分小时单次安装成本
2026-08-16T16:05:16.044907 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 3 4 5 6 7 8 USD per 1,000 impressions hour of day, pooled across the 9 days of the period — a pooling, not a timeline CPM by hour — every asset, 2026-07-23..2026-07-31 (UTC) T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T3b/sasian_slip_dress T3b/sasian_corset_6s
分素材的分小时 CPM

B. 分素材 — 三种收入口径

每一列都是自有口径。Meta 只能表达 入账,因此它跟在那一列旁边的括号里,永远不单独成行。付费用户是 去重的人,不是交易笔数。同期群视野 H=24h,本周期覆盖率 100%。报表时区为 UTC。

素材入账bROAS (Meta)当日回收wROAS同期群@HcROAS付费用户付费率ARPUARPPU最大付费用户占比
T3b/utility_6s$483.711.58 (1.11)$226.490.74$314.431.03305.1%0.3837.5510.4%
T3b/sasian_spice5_15s$300.891.06 (0.96)$217.760.77$243.620.86315.9%0.4167.025.9%
T3b/sasian_bodysuit_10s$358.251.27 (1.35)$280.050.99$325.381.16335.4%0.4618.4914.8%
T3b/sasian_bikini_6s$401.492.12 (1.87)$283.401.50$316.271.67182.2%0.34115.7433.1%
T3b/sasian_bikini_10s$92.340.77 (0.99)$79.460.66$79.460.6671.9%0.21511.3552.0%
T3b/sasian_slip_dress$95.582.39 (1.83)$67.631.69$76.951.9377.0%0.6769.6631.0%
T3b/sasian_fishnet_top_6s$94.543.76 (5.02)$88.783.53$94.543.76813.6%1.50511.1046.6%
T3b/sasian_bodysuit_6s$43.132.97 (2.95)$18.631.29$27.951.93211.8%1.0969.3269.1%
T3b/control_15s#3$9.320.72 (1.25)$9.320.72$9.320.7214.8%0.4449.32100.0%
T3b/utility_10s#2$38.212.99 (0.89)$11.520.90$11.520.9026.9%0.3975.7650.0%
T3b/utility_15s#3$10.750.85 (1.30)$0.000.00$4.990.3900.0%0.000——
T3b/control_10s#3$12.881.03 (0.00)$3.560.28$3.560.28110.0%0.3563.56100.0%
T3b/control_6s#3$12.881.06 (0.00)$5.760.47$5.760.4712.7%0.1565.76100.0%
T3b/sasian_corset_6s$24.502.11 (0.79)$24.502.11$24.502.1114.5%1.11424.50100.0%
T3b/sasian_offshoulder$0.000.00 (2.12)$0.000.00$0.000.0000.0%0.000——
T3b/sasian_jean_shorts$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T3b/sasian_fishnet_top_10s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T3b/sasian_corset_10s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——

16 付费用户门槛(#37):上表中付费用户数不足 16 的行只能看方向,不能看幅度。

四行过了门槛:T3b/sasian_bodysuit_10s 33 个付费用户、0.99,T3b/sasian_spice5_15s 31 个、0.77, T3b/utility_6s 30 个、0.74,T3b/sasian_bikini_6s 18 个、1.50。其中最高的那一行也是最集中的,当 日回收的 33.1% 来自一个人,T3b/sasian_spice5_15s 是 5.9%,所以这几行之间的次序,靠这些数字定 不下来。T3b/sasian_fishnet_top_6s 读出 3.53,背后八个付费用户,该样本量下读不出来。

2026-08-16T16:05:16.551191 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0 2 4 6 8 10 12 14 16 payers / installs % Payer rate by day — every asset (UTC) T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T3b/sasian_slip_dress T3b/sasian_fishnet_top_6s T3b/control_15s#3 T3b/utility_10s#2 T3b/control_6s#3
分素材的每日付费率
2026-08-16T16:05:16.650961 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 within-day USD per install ARPU by day — every asset (UTC) T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T3b/sasian_slip_dress T3b/sasian_fishnet_top_6s T3b/control_15s#3 T3b/utility_10s#2 T3b/control_6s#3
分素材的每日 ARPU
2026-08-16T16:05:16.729755 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0 5 10 15 20 25 within-day USD per payer ARPPU by day — every asset (UTC) T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T3b/sasian_slip_dress T3b/sasian_fishnet_top_6s T3b/control_15s#3 T3b/utility_10s#2 T3b/control_6s#3
分素材的每日 ARPPU
2026-08-16T16:05:16.810507 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0 1 2 3 4 revenue / spend Within-day ROAS by day — every asset (UTC) T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T3b/sasian_slip_dress T3b/sasian_fishnet_top_6s T3b/control_15s#3 T3b/utility_10s#2 T3b/control_6s#3
分素材的每日当日回收 ROAS
2026-08-16T16:05:16.936102 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 period (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 period are not plotted. Composite — every asset across every measure, 2026-07-23..2026-07-31 T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T3b/sasian_slip_dress T3b/sasian_fishnet_top_6s T3b/utility_10s#2 T3b/control_6s#3 T3b/sasian_corset_6s
综合图:各素材在各口径上按本周期归一化后的表现

C. 分素材 — 国家结构

⚠ 本表按账户日切,而本页其余各表都在 UTC 时钟上。 Meta 不提供小时 × 国家的交叉表,因此分国家的投放数据只存 在于广告账户自己的 UTC-7 日上,无法重新拼接。它与上面的 A、B 两表相差七小时;本表只读结构,不与它 们比水平。

B 表里某一行读不读得出来,看这里。最后一列是该素材自身当日回收中来自美国的占比:当它远高于 该素材的美国花费占比时,这条素材的回报是它买到的那块地域的属性。

素材花费印度占比美国占比其他占比美国占其自身收入
T3b/utility_6s$305.3398.9%0.0%1.1%0.0%
T3b/sasian_spice5_15s$282.88100.0%0.0%0.0%0.0%
T3b/sasian_bodysuit_10s$281.69100.0%0.0%0.0%0.0%
T3b/sasian_bikini_6s$189.22100.0%0.0%0.0%0.0%
T3b/sasian_bikini_10s$120.14100.0%0.0%0.0%0.0%
T3b/sasian_slip_dress$39.93100.0%0.0%0.0%0.0%
T3b/sasian_fishnet_top_6s$25.15100.0%0.0%0.0%0.0%
T3b/sasian_bodysuit_6s$14.50100.0%0.0%0.0%0.0%
T3b/control_15s#3$13.0181.9%0.0%18.1%0.0%
T3b/utility_10s#2$12.7982.5%0.0%17.5%0.0%
T3b/utility_15s#3$12.6683.9%0.0%16.1%—
T3b/control_10s#3$12.5576.6%0.0%23.4%0.0%
T3b/control_6s#3$12.1861.7%0.0%38.3%0.0%
T3b/sasian_corset_6s$11.62100.0%0.0%0.0%0.0%
T3b/sasian_offshoulder$9.32100.0%0.0%0.0%—
T3b/sasian_jean_shorts$6.68100.0%0.0%0.0%—
T3b/sasian_fishnet_top_10s$3.63100.0%0.0%0.0%—
T3b/sasian_corset_10s$2.66100.0%0.0%0.0%—

sasian_ 包是 100.0% 印度,每一行的美国占比都是 0.0%,所以这里 B 表中没有一行是买地域买来的。五个 control_/utility_ 格的印度占比在 61.7% 到 98.9% 之间,其余落在 T3 的其他国家,这是本表能显示的 唯一一点结构差异,而它很小。

2026-08-16T16:05:16.305100 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T3b/sasian_slip_dress T3b/sasian_fishnet_top_6s T3b/utility_10s#2 T3b/control_6s#3 T3b/sasian_corset_6s 0 20 40 60 80 100 % of that asset's spend Where each asset bought — share of its own spend by region, 2026-07-23..2026-07-31 (META days) IN ID PH PK NG other
各素材自身花费的地域结构
2026-08-16T16:05:16.374089 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T3b/sasian_slip_dress T3b/sasian_fishnet_top_6s T3b/utility_10s#2 T3b/control_6s#3 T3b/sasian_corset_6s 0.0 0.2 0.4 0.6 0.8 1.0 1.2 USD per install Cost per install by region — every asset, 2026-07-23..2026-07-31 (META days) IN ID PH PK
各素材分地域的单次安装成本
2026-08-16T16:05:16.428482 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ T3b/utility_6s T3b/sasian_spice5_15s T3b/sasian_bodysuit_10s T3b/sasian_bikini_6s T3b/sasian_bikini_10s T3b/sasian_slip_dress T3b/sasian_fishnet_top_6s T3b/utility_10s#2 T3b/control_6s#3 T3b/sasian_corset_6s 0 50 100 150 200 250 USD Within-day revenue by region — every asset, 2026-07-23..2026-07-31 (META days) IN
各素材分地域的当日回收

旧广告系列(08-03 停投) —— 这一行以下的每一节都是 旧广告系列(08-03 停投) 的:当日、分小时、分国家、分广告系列、分素材,都记在 旧广告系列(08-03 停投) 自己那本账、自己那些账户上。用上面的按钮切换。

2026-07-30 开投、2026-08-03 关掉的全球 CBO 买量。在本月它只跑了九天里的两天,花费 $358.20:1,106 次安装、单次 $0.324,是本周期最便宜的;当日回收 $591.31,回报 1.65,背后 61 个付费用户。 DECISION_LOG.md #45 把它定为一条赚钱的投放,一个广告组、CBO、三条素材,所以它的分素材行不是一份 排名。

分小时、分国家与汇总

2026-08-16T16:05:18.803444 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 25 USD hour of day, pooled across the 9 days of the period — a pooling, not a timeline Spend by hour 0 3 6 9 12 15 18 21 hour (UTC) 0 500 1000 1500 2000 2500 3000 impressions Impressions and installs by hour impressions installs 10 20 30 40 50 60 70 80 90 installs (Meta Leads) The period's motion — whole account, 2026-07-23..2026-07-31 (UTC)
本周期走势:分小时花费,以及展示与安装对照
2026-08-16T16:05:18.971417 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 4 8 12 16 20 hour (UTC) 9 10 11 12 13 14 15 CTR % CTR 0 4 8 12 16 20 hour (UTC) 8 10 12 14 16 USD per 1,000 impressions CPM 0 4 8 12 16 20 hour (UTC) 25 30 35 40 45 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install 0 4 8 12 16 20 hour (UTC) 0.250 0.275 0.300 0.325 0.350 0.375 0.400 USD Cost per install Delivery by hour — Earner (retired 08-03), 2026-07-23..2026-07-31 (UTC)
Earner (retired 08-03) 的分小时 CTR、CPM、点击→安装与 CPI
2026-08-16T16:05:19.104743 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ IN BR US TR AR ID MX MY PH PE AU GB 0 25 50 75 100 125 150 175 USD spend and installs are a META-day cut (no hour x country grid exists); revenue is the report period Spend against within-day revenue by country — 2026-07-23..2026-07-31 spend within-day revenue
分国家的花费与当日回收对照
2026-08-16T16:05:19.155896 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ IN BR US TR AR ID MX MY PH PE AU GB 0.0 0.2 0.4 0.6 0.8 1.0 USD per install Cost per install by country — 2026-07-23..2026-07-31 (META days)
分国家的单次安装成本
2026-08-16T16:05:19.216836 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ booked within-day cohort @ 24h 0 100 200 300 400 500 600 700 USD 1.65 1.65 1.92 the number above each bar is that measure's ROAS against the same period's spend The three revenue measures, 2026-07-23..2026-07-31 (UTC) 2026-07-23..2026-07-31
两个周期的三个收入口径,柱顶标注 ROAS
2026-08-16T16:05:19.267444 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 2026-07-23..2026-07-31 0 100 200 300 400 500 USD the bottom segment IS the within-day figure: the same number seen twice Booked revenue by how long its payer had been installed (UTC) paid the day they installed one day after two days after three days after four or more days after
入账收入按付费用户已安装的天数拆分

A. 分素材 — 投放

分子分母都取自 Meta 的计量口径。我们自己的注册数不进入这张表。

素材花费展示CPM点击CTR安装点击→安装CPI
earner/bodysuit$312.9326,15911.962,96811.35%93431.47%$0.335
earner/pool5$36.632,73713.3837213.59%14438.71%$0.254
earner/bikini$8.6469612.418311.93%2833.73%$0.309

CBO 把这个广告系列 $358.20 里的 $312.93 放在 earner/bodysuit 上,earner/bikini 只拿到 $8.64。 这正是本格式在广告组层级提醒过的集中:花费这么不均的一个广告组,是很早就挑定了一条素材,三条从来没 有排出过次序。三条在每一项投放比率上都很接近,CPM 是 11.96、13.38 与 12.41,单次安装成本是 $0.335 、$0.254 与 $0.309,所以这个分配从表上看不出任何依据。

2026-08-16T16:05:19.330952 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 9 10 11 12 13 14 15 CTR % hour of day, pooled across the 9 days of the period — a pooling, not a timeline CTR by hour — every asset, 2026-07-23..2026-07-31 (UTC) earner/bodysuit earner/pool5
分素材的分小时 CTR
2026-08-16T16:05:19.436126 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 25 30 35 40 45 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install by hour — every asset, 2026-07-23..2026-07-31 (UTC) earner/bodysuit earner/pool5
分素材的分小时点击→安装
2026-08-16T16:05:19.490250 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.225 0.250 0.275 0.300 0.325 0.350 0.375 0.400 USD hour of day, pooled across the 9 days of the period — a pooling, not a timeline Cost per install by hour — every asset, 2026-07-23..2026-07-31 (UTC) earner/bodysuit earner/pool5
分素材的分小时单次安装成本
2026-08-16T16:05:19.382198 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 8 10 12 14 16 USD per 1,000 impressions hour of day, pooled across the 9 days of the period — a pooling, not a timeline CPM by hour — every asset, 2026-07-23..2026-07-31 (UTC) earner/bodysuit earner/pool5
分素材的分小时 CPM

B. 分素材 — 三种收入口径

每一列都是自有口径。Meta 只能表达 入账,因此它跟在那一列旁边的括号里,永远不单独成行。付费用户是 去重的人,不是交易笔数。同期群视野 H=24h,本周期覆盖率 100%。报表时区为 UTC。

素材入账bROAS (Meta)当日回收wROAS同期群@HcROAS付费用户付费率ARPUARPPU最大付费用户占比
earner/bodysuit$508.491.62 (1.71)$508.491.62$603.141.93505.0%0.50410.178.1%
earner/pool5$77.782.12 (2.18)$77.782.12$77.782.12106.4%0.4997.7823.1%
earner/bikini$5.040.58 (0.58)$5.040.58$5.040.5813.2%0.1635.04100.0%

16 付费用户门槛(#37):上表中付费用户数不足 16 的行只能看方向,不能看幅度。

只有 earner/bodysuit 过了门槛,50 个付费用户,最大的单个付费用户占其当日回收的 8.1%。这两点 加在一起,让它的 1.62 成为本月唯一一个可以当成水平来读的分素材回报。earner/pool5 读出 2.12,背后 十个付费用户;earner/bikini 读出 0.58,背后一个人;两者在该样本量下都读不出来。它们跑的那两天里 ,三条素材下面的预算从每日 $150 一路动到 $250、$500 与 $875,所以这里也没有任何一处是同预算区间的 。

2026-08-16T16:05:19.797499 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0 1 2 3 4 5 6 7 8 payers / installs % Payer rate by day — every asset (UTC) earner/bodysuit earner/pool5 earner/bikini
分素材的每日付费率
2026-08-16T16:05:19.851812 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 within-day USD per install ARPU by day — every asset (UTC) earner/bodysuit earner/pool5 earner/bikini
分素材的每日 ARPU
2026-08-16T16:05:19.905335 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0 2 4 6 8 10 within-day USD per payer ARPPU by day — every asset (UTC) earner/bodysuit earner/pool5 earner/bikini
分素材的每日 ARPPU
2026-08-16T16:05:19.971070 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.5 1.0 1.5 2.0 2.5 3.0 revenue / spend Within-day ROAS by day — every asset (UTC) earner/bodysuit earner/pool5 earner/bikini
分素材的每日当日回收 ROAS
2026-08-16T16:05:20.050742 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 period (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 period are not plotted. Composite — every asset across every measure, 2026-07-23..2026-07-31 earner/bodysuit earner/pool5 earner/bikini
综合图:各素材在各口径上按本周期归一化后的表现

C. 分素材 — 国家结构

⚠ 本表按账户日切,而本页其余各表都在 UTC 时钟上。 Meta 不提供小时 × 国家的交叉表,因此分国家的投放数据只存 在于广告账户自己的 UTC-7 日上,无法重新拼接。它与上面的 A、B 两表相差七小时;本表只读结构,不与它 们比水平。

B 表里某一行读不读得出来,看这里。最后一列是该素材自身当日回收中来自美国的占比:当它远高于 该素材的美国花费占比时,这条素材的回报是它买到的那块地域的属性。

素材花费印度占比美国占比其他占比美国占其自身收入
earner/bodysuit$408.9932.8%4.7%62.6%8.6%
earner/pool5$36.6327.3%8.2%64.5%6.4%
earner/bikini$150.3329.0%5.0%65.9%0.0%

⚠ 时钟偏移在这里造成的是实打实的偏差。 账户自己的日比报告的日晚七小时,一直伸进 2026-08-01 UTC ,而那时这个广告系列的预算又被调高了。这里的花费列读出 $408.99 与 $150.33,A 表读出的是 $312.93 与 $8.64,所以这些金额属于另一个窗口,只有百分比能对得过来。

结构本身就是一次不设限的全球买量要看的东西:三条素材都是大约三分之一印度、二十分之一美国,其余分散 。earner/bodysuit 自身收入里的美国占比是 8.6%,它的美国花费占比是 4.7%,这是本页上这个差距最温和 的一次。

2026-08-16T16:05:19.599159 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ earner/bodysuit earner/bikini earner/pool5 0 20 40 60 80 100 % of that asset's spend Where each asset bought — share of its own spend by region, 2026-07-23..2026-07-31 (META days) IN BR US TR AR ID other
各素材自身花费的地域结构
2026-08-16T16:05:19.649613 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ earner/bodysuit earner/bikini earner/pool5 0.0 0.2 0.4 0.6 0.8 USD per install Cost per install by region — every asset, 2026-07-23..2026-07-31 (META days) IN BR US TR AR ID
各素材分地域的单次安装成本
2026-08-16T16:05:19.699085 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ earner/bodysuit earner/bikini earner/pool5 0 20 40 60 80 100 120 140 160 USD Within-day revenue by region — every asset, 2026-07-23..2026-07-31 (META days) IN BR US TR AR ID
各素材分地域的当日回收

T2(7 月 25 日) —— 这一行以下的每一节都是 T2(7 月 25 日) 的:当日、分小时、分国家、分广告系列、分素材,都记在 T2(7 月 25 日) 自己那本账、自己那些账户上。用上面的按钮切换。

07-25 那一轮的中间一档,买阿联酋、沙特、马来西亚、南非与墨西哥,另外还有 2026-07-26 加进来的西语拉 丁裔素材包。$268.63 买了 18,253 次展示,298 次安装、单次 $0.901,当日回收 $144.06、回报 0.54, 背后正好 16 个付费用户,刚好卡在门槛上。

分小时、分国家与汇总

2026-08-16T16:05:21.942672 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 hour of day, pooled across the 9 days of the period — a pooling, not a timeline Spend by hour 0 3 6 9 12 15 18 21 hour (UTC) 200 400 600 800 1000 1200 1400 1600 impressions Impressions and installs by hour impressions installs 0 5 10 15 20 25 installs (Meta Leads) The period's motion — whole account, 2026-07-23..2026-07-31 (UTC)
本周期走势:分小时花费,以及展示与安装对照
2026-08-16T16:05:22.104345 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 4 8 12 16 20 hour (UTC) 4 5 6 7 8 CTR % CTR 0 4 8 12 16 20 hour (UTC) 12 13 14 15 16 17 18 19 USD per 1,000 impressions CPM 0 4 8 12 16 20 hour (UTC) 0 10 20 30 40 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install 0 4 8 12 16 20 hour (UTC) 0.6 0.8 1.0 1.2 1.4 1.6 USD Cost per install Delivery by hour — T2 (Jul 25), 2026-07-23..2026-07-31 (UTC)
T2 (Jul 25) 的分小时 CTR、CPM、点击→安装与 CPI
2026-08-16T16:05:22.226177 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ MX MY SA AE ZA 0 25 50 75 100 125 150 175 200 USD spend and installs are a META-day cut (no hour x country grid exists); revenue is the report period Spend against within-day revenue by country — 2026-07-23..2026-07-31 spend within-day revenue
分国家的花费与当日回收对照
2026-08-16T16:05:22.267280 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ MX MY SA AE ZA 0.0 0.5 1.0 1.5 2.0 USD per install Cost per install by country — 2026-07-23..2026-07-31 (META days)
分国家的单次安装成本
2026-08-16T16:05:22.327391 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ booked within-day cohort @ 24h 0 25 50 75 100 125 150 175 200 USD 0.77 0.54 0.56 the number above each bar is that measure's ROAS against the same period's spend The three revenue measures, 2026-07-23..2026-07-31 (UTC) 2026-07-23..2026-07-31
两个周期的三个收入口径,柱顶标注 ROAS
2026-08-16T16:05:22.374436 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 2026-07-23..2026-07-31 0 25 50 75 100 125 150 175 200 USD the bottom segment IS the within-day figure: the same number seen twice Booked revenue by how long its payer had been installed (UTC) paid the day they installed one day after two days after three days after four or more days after
入账收入按付费用户已安装的天数拆分

A. 分素材 — 投放

分子分母都取自 Meta 的计量口径。我们自己的注册数不进入这张表。

素材花费展示CPM点击CTR安装点击→安装CPI
T2b/utility_6s#3$68.173,56819.112446.84%7229.51%$0.947
T2b/utility_15s#2$52.843,19916.521705.31%3721.76%$1.428
T2b/latina_es_bikini_10s$42.734,7069.082916.18%9934.02%$0.432
T2b/latina_es_slip_dress$19.391,26215.36634.99%2742.86%$0.718
T2b/control_10s$14.5091015.93556.04%35.45%$4.833
T2b/control_6s#2$13.3581716.34415.02%717.07%$1.907
T2b/control_15s#2$13.1482915.85536.39%611.32%$2.190
T2b/utility_10s#3$12.7962820.37447.01%920.45%$1.421
T2b/latina_es_bodysuit_10s$11.0182213.39556.69%1323.64%$0.847
T2b/latina_es_offshoulder$6.0430819.61113.57%327.27%$2.013
T2b/latina_es_bikini_6s$4.154449.35214.73%733.33%$0.593
T2b/latina_es_jean_shorts$3.2623413.93135.56%646.15%$0.543
T2b/latina_es_corset_10s$2.2519211.72126.25%541.67%$0.450
T2b/latina_es_bodysuit_6s$2.0013714.60107.30%330.00%$0.667
T2b/latina_es_spice5_15s$1.467718.9622.60%150.00%$1.460
T2b/latina_es_fishnet_top_6s$0.797510.5334.00%00.00%—
T2b/latina_es_corset_6s$0.52774.29228.57%00.00%—
T2b/latina_es_fishnet_top_10s$0.24386.3225.26%00.00%—

加进来的拉丁裔素材包,比它所在的那一轮买得更好 —— T2b/latina_es_bikini_10s 9.08 与 $0.432,T2b/utility_6s#3 19.11 与 $0.947。

2026-08-16T16:05:22.439755 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 CTR % hour of day, pooled across the 9 days of the period — a pooling, not a timeline CTR by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时 CTR
2026-08-16T16:05:22.538289 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时点击→安装
2026-08-16T16:05:22.588017 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 USD hour of day, pooled across the 9 days of the period — a pooling, not a timeline Cost per install by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时单次安装成本
2026-08-16T16:05:22.487696 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 USD per 1,000 impressions hour of day, pooled across the 9 days of the period — a pooling, not a timeline CPM by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时 CPM

B. 分素材 — 三种收入口径

每一列都是自有口径。Meta 只能表达 入账,因此它跟在那一列旁边的括号里,永远不单独成行。付费用户是 去重的人,不是交易笔数。同期群视野 H=24h,本周期覆盖率 100%。报表时区为 UTC。

素材入账bROAS (Meta)当日回收wROAS同期群@HcROAS付费用户付费率ARPUARPPU最大付费用户占比
T2b/utility_6s#3$27.340.40 (0.40)$27.340.40$27.340.4034.0%0.3649.1158.2%
T2b/utility_15s#2$28.430.54 (0.64)$21.630.41$25.030.4725.6%0.60110.8173.6%
T2b/latina_es_bikini_10s$90.022.11 (2.09)$68.401.60$71.801.6899.2%0.6987.6033.2%
T2b/latina_es_slip_dress$5.710.29 (0.29)$5.710.29$5.710.2913.1%0.1785.71100.0%
T2b/control_10s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T2b/control_6s#2$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T2b/control_15s#2$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T2b/utility_10s#3$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T2b/latina_es_bodysuit_10s$9.110.83 (0.00)$0.000.00$0.000.0000.0%0.000——
T2b/latina_es_offshoulder$46.487.70 (5.12)$20.993.48$20.993.48133.3%6.99720.99100.0%
T2b/latina_es_bikini_6s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T2b/latina_es_jean_shorts$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T2b/latina_es_corset_10s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T2b/latina_es_bodysuit_6s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T2b/latina_es_spice5_15s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T2b/latina_es_fishnet_top_6s$0.000.00 (0.00)$0.000.00$0.000.000————
T2b/latina_es_corset_6s$0.000.00 (0.00)$0.000.00$0.000.000————
T2b/latina_es_fishnet_top_10s$0.000.00 (0.00)$0.000.00$0.000.000————

16 付费用户门槛(#37):上表中付费用户数不足 16 的行只能看方向,不能看幅度。

这里没有一条素材过得了门槛。这个广告系列的 16 个付费用户分散在五条素材上,最多的一条占九个,十八条 素材里有十二条在任何口径上都没有收入。两个看起来像结果的数字,T2b/latina_es_offshoulder 的 3.48 与 T2b/latina_es_bikini_10s 的 1.60,一个建立在一个付费用户上,另一个建立在九个上、其中一个人占 了三分之一。该样本量下读不出来。

2026-08-16T16:05:22.877429 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0 2 4 6 8 10 payers / installs % Payer rate by day — every asset (UTC) T2b/utility_6s#3 T2b/utility_15s#2 T2b/latina_es_bikini_10s T2b/latina_es_slip_dress
分素材的每日付费率
2026-08-16T16:05:22.931663 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 within-day USD per install ARPU by day — every asset (UTC) T2b/utility_6s#3 T2b/utility_15s#2 T2b/latina_es_bikini_10s T2b/latina_es_slip_dress
分素材的每日 ARPU
2026-08-16T16:05:22.987148 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0 2 4 6 8 10 within-day USD per payer ARPPU by day — every asset (UTC) T2b/utility_6s#3 T2b/utility_15s#2 T2b/latina_es_bikini_10s T2b/latina_es_slip_dress
分素材的每日 ARPPU
2026-08-16T16:05:23.047184 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 revenue / spend Within-day ROAS by day — every asset (UTC) T2b/utility_6s#3 T2b/utility_15s#2 T2b/latina_es_bikini_10s T2b/latina_es_slip_dress
分素材的每日当日回收 ROAS
2026-08-16T16:05:23.144719 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 period (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 period are not plotted. Composite — every asset across every measure, 2026-07-23..2026-07-31 T2b/utility_6s#3 T2b/utility_15s#2 T2b/latina_es_bikini_10s T2b/latina_es_slip_dress
综合图:各素材在各口径上按本周期归一化后的表现

C. 分素材 — 国家结构

⚠ 本表按账户日切,而本页其余各表都在 UTC 时钟上。 Meta 不提供小时 × 国家的交叉表,因此分国家的投放数据只存 在于广告账户自己的 UTC-7 日上,无法重新拼接。它与上面的 A、B 两表相差七小时;本表只读结构,不与它 们比水平。

B 表里某一行读不读得出来,看这里。最后一列是该素材自身当日回收中来自美国的占比:当它远高于 该素材的美国花费占比时,这条素材的回报是它买到的那块地域的属性。

素材花费印度占比美国占比其他占比美国占其自身收入
T2b/utility_6s#3$68.170.0%0.0%100.0%0.0%
T2b/utility_15s#2$52.840.0%0.0%100.0%0.0%
T2b/latina_es_bikini_10s$42.730.0%0.0%100.0%0.0%
T2b/latina_es_slip_dress$19.390.0%0.0%100.0%0.0%
T2b/control_10s$14.500.0%0.0%100.0%—
T2b/control_6s#2$13.350.0%0.0%100.0%—
T2b/control_15s#2$13.140.0%0.0%100.0%—
T2b/utility_10s#3$12.790.0%0.0%100.0%—
T2b/latina_es_bodysuit_10s$11.010.0%0.0%100.0%—
T2b/latina_es_offshoulder$6.040.0%0.0%100.0%0.0%
T2b/latina_es_bikini_6s$4.150.0%0.0%100.0%—
T2b/latina_es_jean_shorts$3.260.0%0.0%100.0%—
T2b/latina_es_corset_10s$2.250.0%0.0%100.0%—
T2b/latina_es_bodysuit_6s$2.000.0%0.0%100.0%—
T2b/latina_es_spice5_15s$1.460.0%0.0%100.0%—
T2b/latina_es_fishnet_top_6s$0.790.0%0.0%100.0%—
T2b/latina_es_corset_6s$0.520.0%0.0%100.0%—
T2b/latina_es_fishnet_top_10s$0.240.0%0.0%100.0%—

每一行都是 0.0% 印度、0.0% 美国、100.0% 其他,这正是这个广告系列的地域设置在起作用。这一列分不开墨 西哥与沙特,所以这里的结构问题由第 3 节回答:墨西哥 $243.10、回报 0.67,马来西亚 $59.46、回报 0.09 。

2026-08-16T16:05:22.684185 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ T2b/utility_6s#3 T2b/utility_15s#2 T2b/latina_es_bikini_10s T2b/latina_es_slip_dress 0 20 40 60 80 100 % of that asset's spend Where each asset bought — share of its own spend by region, 2026-07-23..2026-07-31 (META days) MX MY SA AE ZA other
各素材自身花费的地域结构
2026-08-16T16:05:22.735823 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ T2b/utility_6s#3 T2b/utility_15s#2 T2b/latina_es_bikini_10s T2b/latina_es_slip_dress 0.0 0.5 1.0 1.5 2.0 2.5 3.0 USD per install Cost per install by region — every asset, 2026-07-23..2026-07-31 (META days) MX MY SA AE ZA
各素材分地域的单次安装成本
2026-08-16T16:05:22.786891 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ T2b/utility_6s#3 T2b/utility_15s#2 T2b/latina_es_bikini_10s T2b/latina_es_slip_dress 0 10 20 30 40 50 60 70 USD Within-day revenue by region — every asset, 2026-07-23..2026-07-31 (META days) MX
各素材分地域的当日回收

T1(7 月 25 日) —— 这一行以下的每一节都是 T1(7 月 25 日) 的:当日、分小时、分国家、分广告系列、分素材,都记在 T1(7 月 25 日) 自己那本账、自己那些账户上。用上面的按钮切换。

07-25 那一轮的英语一档:美国、加拿大、英国、澳大利亚与新西兰。$203.93 买了 2,591 次展示、CPM 78.71 ,102 次安装、单次 $1.999;当日回收 $87.87、回报 0.43,背后六个付费用户。

分小时、分国家与汇总

2026-08-16T16:05:25.054190 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 3 6 9 12 15 18 21 hour (UTC) 0 2 4 6 8 10 12 USD hour of day, pooled across the 9 days of the period — a pooling, not a timeline Spend by hour 0 3 6 9 12 15 18 21 hour (UTC) 40 60 80 100 120 140 160 180 impressions Impressions and installs by hour impressions installs 0 2 4 6 8 10 installs (Meta Leads) The period's motion — whole account, 2026-07-23..2026-07-31 (UTC)
本周期走势:分小时花费,以及展示与安装对照
2026-08-16T16:05:25.203202 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 4 8 12 16 20 hour (UTC) −0.04 −0.02 0.00 0.02 0.04 CTR % CTR 0 4 8 12 16 20 hour (UTC) −0.04 −0.02 0.00 0.02 0.04 USD per 1,000 impressions CPM 0 4 8 12 16 20 hour (UTC) −0.04 −0.02 0.00 0.02 0.04 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install 0 4 8 12 16 20 hour (UTC) −0.04 −0.02 0.00 0.02 0.04 USD Cost per install Delivery by hour — T1 (Jul 25), 2026-07-23..2026-07-31 (UTC)
T1 (Jul 25) 的分小时 CTR、CPM、点击→安装与 CPI
2026-08-16T16:05:25.326782 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ US GB AU CA NZ VI 0 20 40 60 80 USD spend and installs are a META-day cut (no hour x country grid exists); revenue is the report period Spend against within-day revenue by country — 2026-07-23..2026-07-31 spend within-day revenue
分国家的花费与当日回收对照
2026-08-16T16:05:25.366829 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ US GB AU CA NZ 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 USD per install Cost per install by country — 2026-07-23..2026-07-31 (META days)
分国家的单次安装成本
2026-08-16T16:05:25.423860 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ booked within-day cohort @ 24h 0 20 40 60 80 USD 0.43 0.43 0.43 the number above each bar is that measure's ROAS against the same period's spend The three revenue measures, 2026-07-23..2026-07-31 (UTC) 2026-07-23..2026-07-31
两个周期的三个收入口径,柱顶标注 ROAS
2026-08-16T16:05:25.467753 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 2026-07-23..2026-07-31 0 10 20 30 40 50 60 70 80 USD the bottom segment IS the within-day figure: the same number seen twice Booked revenue by how long its payer had been installed (UTC) paid the day they installed one day after two days after three days after four or more days after
入账收入按付费用户已安装的天数拆分

A. 分素材 — 投放

分子分母都取自 Meta 的计量口径。我们自己的注册数不进入这张表。

素材花费展示CPM点击CTR安装点击→安装CPI
T1b/utility_6s#2$74.1887784.5814816.88%5738.51%$1.301
T1b/utility_15s$73.9888883.3112714.30%3527.56%$2.114
T1b/utility_10s$14.3123261.68166.90%00.00%—
T1b/control_6s$13.9918874.411910.11%15.26%$13.990
T1b/control_15s$13.8321364.93198.92%315.79%$4.610
T1b/control_10s#2$13.6419370.672512.95%624.00%$2.273

这里的点击率是本月最好的,而它们什么也没买到: T1b/utility_6s#2 点击率 16.88%,单次安装仍 要 $1.301,因为千次展示成本是 84.58。

2026-08-16T16:05:25.533743 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 CTR % hour of day, pooled across the 9 days of the period — a pooling, not a timeline CTR by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时 CTR
2026-08-16T16:05:25.630395 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时点击→安装
2026-08-16T16:05:25.679537 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 USD hour of day, pooled across the 9 days of the period — a pooling, not a timeline Cost per install by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时单次安装成本
2026-08-16T16:05:25.584847 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 USD per 1,000 impressions hour of day, pooled across the 9 days of the period — a pooling, not a timeline CPM by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时 CPM

B. 分素材 — 三种收入口径

每一列都是自有口径。Meta 只能表达 入账,因此它跟在那一列旁边的括号里,永远不单独成行。付费用户是 去重的人,不是交易笔数。同期群视野 H=24h,本周期覆盖率 100%。报表时区为 UTC。

素材入账bROAS (Meta)当日回收wROAS同期群@HcROAS付费用户付费率ARPUARPPU最大付费用户占比
T1b/utility_6s#2$66.930.90 (0.90)$66.930.90$66.930.9035.2%1.15422.3152.3%
T1b/utility_15s$20.940.28 (0.28)$20.940.28$20.940.2837.9%0.5516.9852.4%
T1b/utility_10s$0.000.00 (0.00)$0.000.00$0.000.000————
T1b/control_6s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T1b/control_15s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
T1b/control_10s#2$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——

16 付费用户门槛(#37):上表中付费用户数不足 16 的行只能看方向,不能看幅度。

整个广告系列六个付费用户,两个 utility_ 格各三个,其他地方一个都没有。它们的 0.90 与 0.28 各自靠 一个人撑着,那个人分别占该格收入的 52.3% 与 52.4%,所以这两格之间的次序就是一个人。该样本量下读不 出来。

2026-08-16T16:05:25.951947 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0 1 2 3 4 5 6 7 payers / installs % Payer rate by day — every asset (UTC) T1b/utility_6s#2 T1b/utility_15s
分素材的每日付费率
2026-08-16T16:05:26.002909 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 within-day USD per install ARPU by day — every asset (UTC) T1b/utility_6s#2 T1b/utility_15s
分素材的每日 ARPU
2026-08-16T16:05:26.054512 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0 5 10 15 20 within-day USD per payer ARPPU by day — every asset (UTC) T1b/utility_6s#2 T1b/utility_15s
分素材的每日 ARPPU
2026-08-16T16:05:26.096480 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.4 0.6 0.8 1.0 revenue / spend Within-day ROAS by day — every asset (UTC) T1b/utility_6s#2 T1b/utility_15s
分素材的每日当日回收 ROAS
2026-08-16T16:05:26.162865 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 period (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 period are not plotted. Composite — every asset across every measure, 2026-07-23..2026-07-31 T1b/utility_6s#2 T1b/utility_15s
综合图:各素材在各口径上按本周期归一化后的表现

C. 分素材 — 国家结构

⚠ 本表按账户日切,而本页其余各表都在 UTC 时钟上。 Meta 不提供小时 × 国家的交叉表,因此分国家的投放数据只存 在于广告账户自己的 UTC-7 日上,无法重新拼接。它与上面的 A、B 两表相差七小时;本表只读结构,不与它 们比水平。

B 表里某一行读不读得出来,看这里。最后一列是该素材自身当日回收中来自美国的占比:当它远高于 该素材的美国花费占比时,这条素材的回报是它买到的那块地域的属性。

素材花费印度占比美国占比其他占比美国占其自身收入
T1b/utility_6s#2$74.180.0%48.4%51.6%100.0%
T1b/utility_15s$73.980.0%42.5%57.5%52.4%
T1b/utility_10s$14.310.0%46.3%53.7%—
T1b/control_6s$13.990.0%39.5%60.5%—
T1b/control_15s$13.830.0%37.4%62.6%—
T1b/control_10s#2$13.640.0%38.8%61.2%—

T1b/utility_6s#2 把 48.4% 的钱花在美国,当日回收的 100.0% 也来自美国。三个付费用户落在一个国 家里,另外四个国家一个都没有,这点样本读不出一个关于美国市场的结论。

2026-08-16T16:05:25.773655 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ T1b/utility_6s#2 T1b/utility_15s 0 20 40 60 80 100 % of that asset's spend Where each asset bought — share of its own spend by region, 2026-07-23..2026-07-31 (META days) US GB AU CA NZ VI other
各素材自身花费的地域结构
2026-08-16T16:05:25.824352 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ T1b/utility_6s#2 T1b/utility_15s 0.0 0.5 1.0 1.5 2.0 2.5 3.0 USD per install Cost per install by region — every asset, 2026-07-23..2026-07-31 (META days) US GB AU CA NZ
各素材分地域的单次安装成本
2026-08-16T16:05:25.871881 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ T1b/utility_6s#2 T1b/utility_15s 0 10 20 30 40 50 60 70 USD Within-day revenue by region — every asset, 2026-07-23..2026-07-31 (META days) US CA
各素材分地域的当日回收

德语区(DE/AT/CH) —— 这一行以下的每一节都是 德语区(DE/AT/CH) 的:当日、分小时、分国家、分广告系列、分素材,都记在 德语区(DE/AT/CH) 自己那本账、自己那些账户上。用上面的按钮切换。

德语区广告系列,2026-07-30 发布并开投,DE+AT 与 CH 各一个广告组,各每日 $75。在本月它只跑了九天里 的两天,花费 $99.98:2,140 次展示、CPM 46.72,23 次安装、单次 $4.347,当日回收 $127.00、回报 1.27,背后两个付费用户。它是本周期里唯一一个回报在 1.00 以上、同时又读不出来的广告系列。

分小时、分国家与汇总

2026-08-16T16:05:28.033077 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 25 USD hour of day, pooled across the 9 days of the period — a pooling, not a timeline Spend by hour 0 3 6 9 12 15 18 21 hour (UTC) 0 100 200 300 400 500 impressions Impressions and installs by hour impressions installs 0 2 4 6 8 installs (Meta Leads) The period's motion — whole account, 2026-07-23..2026-07-31 (UTC)
本周期走势:分小时花费,以及展示与安装对照
2026-08-16T16:05:28.188350 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 4 8 12 16 20 hour (UTC) 2.90 2.95 3.00 3.05 3.10 3.15 3.20 CTR % CTR 0 4 8 12 16 20 hour (UTC) 36 37 38 39 USD per 1,000 impressions CPM 0 4 8 12 16 20 hour (UTC) 25.5 26.0 26.5 27.0 27.5 28.0 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install 0 4 8 12 16 20 hour (UTC) 4.4 4.5 4.6 4.7 4.8 USD Cost per install Delivery by hour — German (DE/AT/CH), 2026-07-23..2026-07-31 (UTC)
German (DE/AT/CH) 的分小时 CTR、CPM、点击→安装与 CPI
2026-08-16T16:05:28.306768 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ DE CH AT 0 20 40 60 80 USD spend and installs are a META-day cut (no hour x country grid exists); revenue is the report period Spend against within-day revenue by country — 2026-07-23..2026-07-31 spend within-day revenue
分国家的花费与当日回收对照
2026-08-16T16:05:28.345711 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ DE CH AT 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 USD per install Cost per install by country — 2026-07-23..2026-07-31 (META days)
分国家的单次安装成本
2026-08-16T16:05:28.404446 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ booked within-day cohort @ 24h 0 20 40 60 80 100 120 140 USD 1.49 1.27 1.45 the number above each bar is that measure's ROAS against the same period's spend The three revenue measures, 2026-07-23..2026-07-31 (UTC) 2026-07-23..2026-07-31
两个周期的三个收入口径,柱顶标注 ROAS
2026-08-16T16:05:28.451361 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 2026-07-23..2026-07-31 0 20 40 60 80 100 120 140 USD the bottom segment IS the within-day figure: the same number seen twice Booked revenue by how long its payer had been installed (UTC) paid the day they installed one day after two days after three days after four or more days after
入账收入按付费用户已安装的天数拆分

A. 分素材 — 投放

分子分母都取自 Meta 的计量口径。我们自己的注册数不进入这张表。

素材花费展示CPM点击CTR安装点击→安装CPI
DE/de_german_mature_6s$27.4045660.09296.36%827.59%$3.425
DE/ch_arab_gulf_6s$21.7549843.67122.41%18.33%$21.750
DE/de_arab_gulf_10s$5.798369.7656.02%120.00%$5.790
DE/ch_german_10s$5.2010250.9821.96%00.00%—
DE/ch_latina_true_6s$5.1710648.7710.94%00.00%—
DE/de_turkish_mature_10s$4.7112637.3853.97%360.00%$1.570
DE/german_6s$4.0710040.7066.00%00.00%—
DE/ch_german_mature_6s$4.0210339.0300.00%0——
DE/turkish_mature$3.955769.3047.02%4100.00%$0.987
DE/ch_arab_gulf_10s$3.6614625.0710.68%1100.00%$3.660
DE/german_mature$3.018933.8266.74%233.33%$1.505
DE/de_arab_gulf_6s$2.512696.5427.69%150.00%$2.510
DE/german$1.906031.6700.00%0——
DE/latina_true$1.362164.7614.76%00.00%—
DE/de_turkish_10s$1.163038.67620.00%233.33%$0.580
DE/de_pool5_15s$1.042149.5200.00%0——
DE/de_asian_mature_6s$0.993726.7600.00%0——
DE/ch_german_mature_10s$0.763025.3313.33%00.00%—
DE/turkish_6s$0.542422.5014.17%00.00%—
DE/ch_latina_mature_10s$0.42760.0000.00%0——
DE/de_latina_true_6s$0.25550.0000.00%0——
DE/ch_german_6s$0.14435.0000.00%0——
DE/ch_asian_mature_6s$0.11336.67133.33%00.00%—
DE/ch_pool5_15s$0.07417.5000.00%0——

二十四条素材分掉了 $99.98,其中十九条拿到不足 $5。 只有最大的两条点击率读得出来,6.36% 对 2.41%。

2026-08-16T16:05:28.516384 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 CTR % hour of day, pooled across the 9 days of the period — a pooling, not a timeline CTR by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时 CTR
2026-08-16T16:05:28.610175 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时点击→安装
2026-08-16T16:05:28.655760 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 USD hour of day, pooled across the 9 days of the period — a pooling, not a timeline Cost per install by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时单次安装成本
2026-08-16T16:05:28.565166 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 USD per 1,000 impressions hour of day, pooled across the 9 days of the period — a pooling, not a timeline CPM by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时 CPM

B. 分素材 — 三种收入口径

每一列都是自有口径。Meta 只能表达 入账,因此它跟在那一列旁边的括号里,永远不单独成行。付费用户是 去重的人,不是交易笔数。同期群视野 H=24h,本周期覆盖率 100%。报表时区为 UTC。

素材入账bROAS (Meta)当日回收wROAS同期群@HcROAS付费用户付费率ARPUARPPU最大付费用户占比
DE/de_german_mature_6s$101.663.71 (3.36)$92.363.37$98.103.58112.5%11.54592.36100.0%
DE/ch_arab_gulf_6s$47.022.16 (2.15)$34.651.59$47.022.16150.0%17.32334.65100.0%
DE/de_arab_gulf_10s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
DE/ch_german_10s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/ch_latina_true_6s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/de_turkish_mature_10s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
DE/german_6s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/ch_german_mature_6s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/turkish_mature$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
DE/ch_arab_gulf_10s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
DE/german_mature$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
DE/de_arab_gulf_6s$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——
DE/german$0.000.00 (0.00)$0.000.00$0.000.000————
DE/latina_true$0.000.00 (0.00)$0.000.00$0.000.000————
DE/de_turkish_10s$0.000.00 (4.30)$0.000.00$0.000.0000.0%0.000——
DE/de_pool5_15s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/de_asian_mature_6s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/ch_german_mature_10s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/turkish_6s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/ch_latina_mature_10s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/de_latina_true_6s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/ch_german_6s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/ch_asian_mature_6s$0.000.00 (0.00)$0.000.00$0.000.000————
DE/ch_pool5_15s$0.000.00 (0.00)$0.000.00$0.000.000————

16 付费用户门槛(#37):上表中付费用户数不足 16 的行只能看方向,不能看幅度。

整个广告系列两个付费用户,分别落在两条素材上,各占该素材收入的 100.0%。3.37 与 1.59 就是两个人,第 5 节里这个广告系列的 1.27 是同样这两个人高了一层。该样本量下读不出来。7 月这次德语区读数真正立住的 东西写在 DECISION_LOG.md #44 里,靠的是投放数据:德语区的展示价格是印度的几倍,而这块地域每次安 装赚到的钱足以撑住它,只是素材包里没有一条到得了盈亏平衡。

2026-08-16T16:05:28.865775 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 payers / installs % Payer rate by day — every asset (UTC)
分素材的每日付费率
2026-08-16T16:05:28.903336 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 within-day USD per install ARPU by day — every asset (UTC)
分素材的每日 ARPU
2026-08-16T16:05:28.942014 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 within-day USD per payer ARPPU by day — every asset (UTC)
分素材的每日 ARPPU
2026-08-16T16:05:28.981124 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 revenue / spend Within-day ROAS by day — every asset (UTC)
分素材的每日当日回收 ROAS
2026-08-16T16:05:29.043539 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 period (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 period are not plotted. Composite — every asset across every measure, 2026-07-23..2026-07-31
综合图:各素材在各口径上按本周期归一化后的表现

C. 分素材 — 国家结构

⚠ 本表按账户日切,而本页其余各表都在 UTC 时钟上。 Meta 不提供小时 × 国家的交叉表,因此分国家的投放数据只存 在于广告账户自己的 UTC-7 日上,无法重新拼接。它与上面的 A、B 两表相差七小时;本表只读结构,不与它 们比水平。

B 表里某一行读不读得出来,看这里。最后一列是该素材自身当日回收中来自美国的占比:当它远高于 该素材的美国花费占比时,这条素材的回报是它买到的那块地域的属性。

素材花费印度占比美国占比其他占比美国占其自身收入
DE/de_german_mature_6s$27.400.0%0.0%100.0%0.0%
DE/ch_arab_gulf_6s$21.750.0%0.0%100.0%0.0%
DE/de_arab_gulf_10s$5.790.0%0.0%100.0%—
DE/ch_german_10s$5.200.0%0.0%100.0%—
DE/ch_latina_true_6s$5.170.0%0.0%100.0%—
DE/de_turkish_mature_10s$4.710.0%0.0%100.0%—
DE/german_6s$4.070.0%0.0%100.0%—
DE/ch_german_mature_6s$4.020.0%0.0%100.0%—
DE/turkish_mature$3.950.0%0.0%100.0%—
DE/ch_arab_gulf_10s$3.660.0%0.0%100.0%—
DE/german_mature$3.010.0%0.0%100.0%—
DE/de_arab_gulf_6s$2.510.0%0.0%100.0%—
DE/german$1.900.0%0.0%100.0%—
DE/latina_true$1.360.0%0.0%100.0%—
DE/de_turkish_10s$1.160.0%0.0%100.0%—
DE/de_pool5_15s$1.040.0%0.0%100.0%—
DE/de_asian_mature_6s$0.990.0%0.0%100.0%—
DE/ch_german_mature_10s$0.760.0%0.0%100.0%—
DE/turkish_6s$0.540.0%0.0%100.0%—
DE/ch_latina_mature_10s$0.420.0%0.0%100.0%—
DE/de_latina_true_6s$0.250.0%0.0%100.0%—
DE/ch_german_6s$0.140.0%0.0%100.0%—
DE/ch_asian_mature_6s$0.110.0%0.0%100.0%—
DE/ch_pool5_15s$0.070.0%0.0%100.0%—

每一行都是 100.0% 其他,这是这个广告系列的定向要求的。本表分不出来的部分在第 3 节:德国 $60.82、回 报 1.67,瑞士 $43.05、回报 0.80,各自背后一个付费用户。

2026-08-16T16:05:28.738532 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 20 40 60 80 100 % of that asset's spend Where each asset bought — share of its own spend by region, 2026-07-23..2026-07-31 (META days) DE CH AT other
各素材自身花费的地域结构
2026-08-16T16:05:28.766757 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0.0 0.2 0.4 0.6 0.8 1.0 USD per install Cost per install by region — every asset, 2026-07-23..2026-07-31 (META days)
各素材分地域的单次安装成本
2026-08-16T16:05:28.791976 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0.0 0.2 0.4 0.6 0.8 1.0 USD Within-day revenue by region — every asset, 2026-07-23..2026-07-31 (META days)
各素材分地域的当日回收

T1(7 月 23 日) —— 这一行以下的每一节都是 T1(7 月 23 日) 的:当日、分小时、分国家、分广告系列、分素材,都记在 T1(7 月 23 日) 自己那本账、自己那些账户上。用上面的按钮切换。

第一轮的英语一档,来自 2026-07-23 那一轮。它花了 $72.39,买了 1,296 次展示、CPM 55.86,产出四次安 装,单次 $18.098。它没有付费用户,三个收入口径都是 $0.00。它凭花费占比拿到了一个标签页,此外什么也 没拿到:这是一个买到了展示、之后几乎什么都没买到的广告系列。

分小时、分国家与汇总

2026-08-16T16:05:30.903806 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 hour of day, pooled across the 9 days of the period — a pooling, not a timeline Spend by hour 0 3 6 9 12 15 18 21 hour (UTC) 0 50 100 150 200 250 300 impressions Impressions and installs by hour impressions installs 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 installs (Meta Leads) The period's motion — whole account, 2026-07-23..2026-07-31 (UTC)
本周期走势:分小时花费,以及展示与安装对照
2026-08-16T16:05:31.051686 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 4 8 12 16 20 hour (UTC) −0.04 −0.02 0.00 0.02 0.04 CTR % CTR 0 4 8 12 16 20 hour (UTC) −0.04 −0.02 0.00 0.02 0.04 USD per 1,000 impressions CPM 0 4 8 12 16 20 hour (UTC) −0.04 −0.02 0.00 0.02 0.04 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install 0 4 8 12 16 20 hour (UTC) −0.04 −0.02 0.00 0.02 0.04 USD Cost per install Delivery by hour — T1 (Jul 23), 2026-07-23..2026-07-31 (UTC)
T1 (Jul 23) 的分小时 CTR、CPM、点击→安装与 CPI
2026-08-16T16:05:31.171172 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ US GB CA AU NZ VI 0 5 10 15 20 25 30 35 USD spend and installs are a META-day cut (no hour x country grid exists); revenue is the report period Spend against within-day revenue by country — 2026-07-23..2026-07-31 spend within-day revenue
分国家的花费与当日回收对照
2026-08-16T16:05:31.208378 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ GB CA NZ 0 5 10 15 20 USD per install Cost per install by country — 2026-07-23..2026-07-31 (META days)
分国家的单次安装成本
2026-08-16T16:05:31.261730 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ booked within-day cohort @ 24h 0 2 4 6 8 USD 0.13 0.00 0.13 the number above each bar is that measure's ROAS against the same period's spend The three revenue measures, 2026-07-23..2026-07-31 (UTC) 2026-07-23..2026-07-31
两个周期的三个收入口径,柱顶标注 ROAS
2026-08-16T16:05:31.302518 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 2026-07-23..2026-07-31 0 2 4 6 8 USD the bottom segment IS the within-day figure: the same number seen twice Booked revenue by how long its payer had been installed (UTC) paid the day they installed one day after two days after three days after four or more days after
入账收入按付费用户已安装的天数拆分

A. 分素材 — 投放

分子分母都取自 Meta 的计量口径。我们自己的注册数不进入这张表。

素材花费展示CPM点击CTR安装点击→安装CPI
T1a/utility_sexy$30.9755955.40244.29%00.00%—
T1a/control_sexy$28.9534982.954011.46%410.00%$7.237
T1a/control_sexy#4$6.5617836.8542.25%00.00%—
T1a/utility_sexy#4$5.9121028.1431.43%00.00%—

四条素材里有三条一次安装也没产出。第四条 T1a/control_sexy 产出四次,单次 $7.237,背后是 349 次以 82.95 千次成本买来的展示。本该被拿来比较的两条实验臂 utility_sexy 与 control_sexy 花了 $30.97 与 $28.95,产出零次对四次安装,这是四个事件的差别。

2026-08-16T16:05:31.364444 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 CTR % hour of day, pooled across the 9 days of the period — a pooling, not a timeline CTR by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时 CTR
2026-08-16T16:05:31.457497 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时点击→安装
2026-08-16T16:05:31.504632 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 USD hour of day, pooled across the 9 days of the period — a pooling, not a timeline Cost per install by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时单次安装成本
2026-08-16T16:05:31.409699 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 USD per 1,000 impressions hour of day, pooled across the 9 days of the period — a pooling, not a timeline CPM by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时 CPM

B. 分素材 — 三种收入口径

每一列都是自有口径。Meta 只能表达 入账,因此它跟在那一列旁边的括号里,永远不单独成行。付费用户是 去重的人,不是交易笔数。同期群视野 H=24h,本周期覆盖率 100%。报表时区为 UTC。

素材入账bROAS (Meta)当日回收wROAS同期群@HcROAS付费用户付费率ARPUARPPU最大付费用户占比
T1a/utility_sexy$0.000.00 (0.00)$0.000.00$0.000.000————
T1a/control_sexy$9.390.32 (0.32)$0.000.00$9.390.3200.0%0.000——
T1a/control_sexy#4$0.000.00 (0.00)$0.000.00$0.000.000————
T1a/utility_sexy#4$0.000.00 (0.00)$0.000.00$0.000.000————

16 付费用户门槛(#37):上表中付费用户数不足 16 的行只能看方向,不能看幅度。

$9.39 的入账收入,对上 $72.39 的花费,其中没有一分是当日回收的,整个广告系列没有付费用户。平时那句 提醒在这里用不上,因为这里没有任何数字需要提醒。

2026-08-16T16:05:31.720812 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 payers / installs % Payer rate by day — every asset (UTC)
分素材的每日付费率
2026-08-16T16:05:31.758779 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 within-day USD per install ARPU by day — every asset (UTC)
分素材的每日 ARPU
2026-08-16T16:05:31.797515 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 within-day USD per payer ARPPU by day — every asset (UTC)
分素材的每日 ARPPU
2026-08-16T16:05:31.844937 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 revenue / spend Within-day ROAS by day — every asset (UTC)
分素材的每日当日回收 ROAS
2026-08-16T16:05:31.907908 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 period (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 period are not plotted. Composite — every asset across every measure, 2026-07-23..2026-07-31
综合图:各素材在各口径上按本周期归一化后的表现

C. 分素材 — 国家结构

⚠ 本表按账户日切,而本页其余各表都在 UTC 时钟上。 Meta 不提供小时 × 国家的交叉表,因此分国家的投放数据只存 在于广告账户自己的 UTC-7 日上,无法重新拼接。它与上面的 A、B 两表相差七小时;本表只读结构,不与它 们比水平。

B 表里某一行读不读得出来,看这里。最后一列是该素材自身当日回收中来自美国的占比:当它远高于 该素材的美国花费占比时,这条素材的回报是它买到的那块地域的属性。

素材花费印度占比美国占比其他占比美国占其自身收入
T1a/utility_sexy$30.970.0%51.2%48.8%—
T1a/control_sexy$28.950.0%57.5%42.5%—
T1a/control_sexy#4$6.560.0%21.2%78.8%—
T1a/utility_sexy#4$5.910.0%32.0%68.0%—

每条素材有五分之一到五分之三的钱花在美国,而最后一列每一行都是空的,因为根本没有收入可以归。

2026-08-16T16:05:31.594153 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 20 40 60 80 100 % of that asset's spend Where each asset bought — share of its own spend by region, 2026-07-23..2026-07-31 (META days) US GB CA AU NZ VI other
各素材自身花费的地域结构
2026-08-16T16:05:31.623785 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0.0 0.2 0.4 0.6 0.8 1.0 USD per install Cost per install by region — every asset, 2026-07-23..2026-07-31 (META days)
各素材分地域的单次安装成本
2026-08-16T16:05:31.650038 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0.0 0.2 0.4 0.6 0.8 1.0 USD Within-day revenue by region — every asset, 2026-07-23..2026-07-31 (META days)
各素材分地域的当日回收

T2(7 月 23 日) —— 这一行以下的每一节都是 T2(7 月 23 日) 的:当日、分小时、分国家、分广告系列、分素材,都记在 T2(7 月 23 日) 自己那本账、自己那些账户上。用上面的按钮切换。

第一轮的中间一档。$55.25 买了 2,680 次展示、CPM 20.62,35 次安装、单次 $1.579,当日回收 $9.11 、回报 0.16,背后一个付费用户。

分小时、分国家与汇总

2026-08-16T16:05:33.743508 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 3 6 9 12 15 18 21 hour (UTC) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 USD hour of day, pooled across the 9 days of the period — a pooling, not a timeline Spend by hour 0 3 6 9 12 15 18 21 hour (UTC) 80 100 120 140 160 180 impressions Impressions and installs by hour impressions installs 0 1 2 3 4 5 installs (Meta Leads) The period's motion — whole account, 2026-07-23..2026-07-31 (UTC)
本周期走势:分小时花费,以及展示与安装对照
2026-08-16T16:05:33.890951 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 4 8 12 16 20 hour (UTC) −0.04 −0.02 0.00 0.02 0.04 CTR % CTR 0 4 8 12 16 20 hour (UTC) −0.04 −0.02 0.00 0.02 0.04 USD per 1,000 impressions CPM 0 4 8 12 16 20 hour (UTC) −0.04 −0.02 0.00 0.02 0.04 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install 0 4 8 12 16 20 hour (UTC) −0.04 −0.02 0.00 0.02 0.04 USD Cost per install Delivery by hour — T2 (Jul 23), 2026-07-23..2026-07-31 (UTC)
T2 (Jul 23) 的分小时 CTR、CPM、点击→安装与 CPI
2026-08-16T16:05:34.010271 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ MY MX AE SA ZA 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 USD spend and installs are a META-day cut (no hour x country grid exists); revenue is the report period Spend against within-day revenue by country — 2026-07-23..2026-07-31 spend within-day revenue
分国家的花费与当日回收对照
2026-08-16T16:05:34.053362 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ MY MX AE SA ZA 0 1 2 3 4 5 6 7 8 USD per install Cost per install by country — 2026-07-23..2026-07-31 (META days)
分国家的单次安装成本
2026-08-16T16:05:34.111758 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ booked within-day cohort @ 24h 0 2 4 6 8 USD 0.16 0.16 0.16 the number above each bar is that measure's ROAS against the same period's spend The three revenue measures, 2026-07-23..2026-07-31 (UTC) 2026-07-23..2026-07-31
两个周期的三个收入口径,柱顶标注 ROAS
2026-08-16T16:05:34.151417 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 2026-07-23..2026-07-31 0 2 4 6 8 USD the bottom segment IS the within-day figure: the same number seen twice Booked revenue by how long its payer had been installed (UTC) paid the day they installed one day after two days after three days after four or more days after
入账收入按付费用户已安装的天数拆分

A. 分素材 — 投放

分子分母都取自 Meta 的计量口径。我们自己的注册数不进入这张表。

素材花费展示CPM点击CTR安装点击→安装CPI
T2a/control_sexy#2$28.181,30321.631219.29%1310.74%$2.168
T2a/utility_sexy#3$27.071,37719.6617512.71%2212.57%$1.230

这是本月唯一一处两条实验臂花钱相当、并且两条都有投放的地方。 utility 的点击率 12.71%, control 9.29%,后面站着的一共三十五次安装。

2026-08-16T16:05:34.213453 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 CTR % hour of day, pooled across the 9 days of the period — a pooling, not a timeline CTR by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时 CTR
2026-08-16T16:05:34.307435 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时点击→安装
2026-08-16T16:05:34.359676 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 USD hour of day, pooled across the 9 days of the period — a pooling, not a timeline Cost per install by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时单次安装成本
2026-08-16T16:05:34.261179 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 USD per 1,000 impressions hour of day, pooled across the 9 days of the period — a pooling, not a timeline CPM by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时 CPM

B. 分素材 — 三种收入口径

每一列都是自有口径。Meta 只能表达 入账,因此它跟在那一列旁边的括号里,永远不单独成行。付费用户是 去重的人,不是交易笔数。同期群视野 H=24h,本周期覆盖率 100%。报表时区为 UTC。

素材入账bROAS (Meta)当日回收wROAS同期群@HcROAS付费用户付费率ARPUARPPU最大付费用户占比
T2a/control_sexy#2$9.110.32 (0.32)$9.110.32$9.110.3217.7%0.7019.11100.0%
T2a/utility_sexy#3$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——

16 付费用户门槛(#37):上表中付费用户数不足 16 的行只能看方向,不能看幅度。

整个广告系列一个付费用户,而它落在投放较差的那一条上。该样本量下读不出来,两个方向都读不出来:这么 大的一个投放优势,和一个人带来的收入劣势,是两个不同的问题,而只有第一个有样本。

2026-08-16T16:05:34.605644 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.00 0.01 0.02 0.03 0.04 0.05 payers / installs % Payer rate by day — every asset (UTC) T2a/utility_sexy#3
分素材的每日付费率
2026-08-16T16:05:34.647788 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.00 0.01 0.02 0.03 0.04 0.05 within-day USD per install ARPU by day — every asset (UTC) T2a/utility_sexy#3
分素材的每日 ARPU
2026-08-16T16:05:34.689960 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 within-day USD per payer ARPPU by day — every asset (UTC)
分素材的每日 ARPPU
2026-08-16T16:05:34.728990 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) −0.04 −0.02 0.00 0.02 0.04 revenue / spend Within-day ROAS by day — every asset (UTC) T2a/utility_sexy#3
分素材的每日当日回收 ROAS
2026-08-16T16:05:34.792308 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 period (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 period are not plotted. Composite — every asset across every measure, 2026-07-23..2026-07-31 T2a/utility_sexy#3
综合图:各素材在各口径上按本周期归一化后的表现

C. 分素材 — 国家结构

⚠ 本表按账户日切,而本页其余各表都在 UTC 时钟上。 Meta 不提供小时 × 国家的交叉表,因此分国家的投放数据只存 在于广告账户自己的 UTC-7 日上,无法重新拼接。它与上面的 A、B 两表相差七小时;本表只读结构,不与它 们比水平。

B 表里某一行读不读得出来,看这里。最后一列是该素材自身当日回收中来自美国的占比:当它远高于 该素材的美国花费占比时,这条素材的回报是它买到的那块地域的属性。

素材花费印度占比美国占比其他占比美国占其自身收入
T2a/control_sexy#2$28.180.0%0.0%100.0%0.0%
T2a/utility_sexy#3$27.070.0%0.0%100.0%—

两行都是 100.0% 其他,没有印度也没有美国,所以两条实验臂买的是同一块地域,上面那处投放差异不是结构 差异。

2026-08-16T16:05:34.448643 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ T2a/utility_sexy#3 0 20 40 60 80 100 % of that asset's spend Where each asset bought — share of its own spend by region, 2026-07-23..2026-07-31 (META days) MY MX AE SA ZA other
各素材自身花费的地域结构
2026-08-16T16:05:34.489802 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ T2a/utility_sexy#3 0.0 0.5 1.0 1.5 2.0 2.5 3.0 USD per install Cost per install by region — every asset, 2026-07-23..2026-07-31 (META days) MY MX AE SA ZA
各素材分地域的单次安装成本
2026-08-16T16:05:34.523460 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ T2a/utility_sexy#3 0.0 0.2 0.4 0.6 0.8 1.0 USD Within-day revenue by region — every asset, 2026-07-23..2026-07-31 (META days)
各素材分地域的当日回收

T3(7 月 23 日) —— 这一行以下的每一节都是 T3(7 月 23 日) 的:当日、分小时、分国家、分广告系列、分素材,都记在 T3(7 月 23 日) 自己那本账、自己那些账户上。用上面的按钮切换。

第一轮的走量一档,买印度与东南亚。$40.68 买了 5,902 次展示、CPM 6.89,是本月第二便宜的;118 次安装 、单次 $0.345,当日回收 $67.82、回报 1.67,背后五个付费用户。

分小时、分国家与汇总

2026-08-16T16:05:36.665921 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 3 6 9 12 15 18 21 hour (UTC) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 USD hour of day, pooled across the 9 days of the period — a pooling, not a timeline Spend by hour 0 3 6 9 12 15 18 21 hour (UTC) 100 200 300 400 500 600 impressions Impressions and installs by hour impressions installs 0 2 4 6 8 10 installs (Meta Leads) The period's motion — whole account, 2026-07-23..2026-07-31 (UTC)
本周期走势:分小时花费,以及展示与安装对照
2026-08-16T16:05:36.824455 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 4 8 12 16 20 hour (UTC) 4.90 4.95 5.00 5.05 5.10 5.15 CTR % CTR 0 4 8 12 16 20 hour (UTC) 3.5 3.6 3.7 3.8 3.9 4.0 4.1 USD per 1,000 impressions CPM 0 4 8 12 16 20 hour (UTC) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install 0 4 8 12 16 20 hour (UTC) 2.40 2.45 2.50 2.55 2.60 USD Cost per install Delivery by hour — T3 (Jul 23), 2026-07-23..2026-07-31 (UTC)
T3 (Jul 23) 的分小时 CTR、CPM、点击→安装与 CPI
2026-08-16T16:05:36.947022 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ IN ID PH PK NG 0 10 20 30 40 50 60 USD spend and installs are a META-day cut (no hour x country grid exists); revenue is the report period Spend against within-day revenue by country — 2026-07-23..2026-07-31 spend within-day revenue
分国家的花费与当日回收对照
2026-08-16T16:05:36.984542 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ IN ID PH PK 0.0 0.1 0.2 0.3 0.4 0.5 USD per install Cost per install by country — 2026-07-23..2026-07-31 (META days)
分国家的单次安装成本
2026-08-16T16:05:37.038836 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ booked within-day cohort @ 24h 0 20 40 60 80 USD 2.30 1.67 1.75 the number above each bar is that measure's ROAS against the same period's spend The three revenue measures, 2026-07-23..2026-07-31 (UTC) 2026-07-23..2026-07-31
两个周期的三个收入口径,柱顶标注 ROAS
2026-08-16T16:05:37.083271 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 2026-07-23..2026-07-31 0 20 40 60 80 USD the bottom segment IS the within-day figure: the same number seen twice Booked revenue by how long its payer had been installed (UTC) paid the day they installed one day after two days after three days after four or more days after
入账收入按付费用户已安装的天数拆分

A. 分素材 — 投放

分子分母都取自 Meta 的计量口径。我们自己的注册数不进入这张表。

素材花费展示CPM点击CTR安装点击→安装CPI
T3a/utility_sexy#2$28.904,1736.9347811.45%11423.85%$0.254
T3a/control_sexy#3$11.781,7296.811146.59%43.51%$2.945

与 07-23 那一轮的中间一档形状相同、方向相同,只是量更大。在千次成本基本一样的展示上,6.93 对 6.81 ,utility 的点击率是 11.45%,control 是 6.59%;点击转安装 23.85% 对 3.51%,最后落到单次安装 $0.254 对 $2.945。这里花费没有对齐,$28.90 对 $11.78,所以读比率,不要读合计。

2026-08-16T16:05:37.143613 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 CTR % hour of day, pooled across the 9 days of the period — a pooling, not a timeline CTR by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时 CTR
2026-08-16T16:05:37.234196 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 installs / clicks % hour of day, pooled across the 9 days of the period — a pooling, not a timeline Click → install by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时点击→安装
2026-08-16T16:05:37.280557 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 USD hour of day, pooled across the 9 days of the period — a pooling, not a timeline Cost per install by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时单次安装成本
2026-08-16T16:05:37.189417 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 0 2 4 6 8 10 12 14 16 18 20 22 hour (UTC) 0.0 0.2 0.4 0.6 0.8 1.0 USD per 1,000 impressions hour of day, pooled across the 9 days of the period — a pooling, not a timeline CPM by hour — every asset, 2026-07-23..2026-07-31 (UTC)
分素材的分小时 CPM

B. 分素材 — 三种收入口径

每一列都是自有口径。Meta 只能表达 入账,因此它跟在那一列旁边的括号里,永远不单独成行。付费用户是 去重的人,不是交易笔数。同期群视野 H=24h,本周期覆盖率 100%。报表时区为 UTC。

素材入账bROAS (Meta)当日回收wROAS同期群@HcROAS付费用户付费率ARPUARPPU最大付费用户占比
T3a/utility_sexy#2$93.573.24 (2.23)$67.822.35$71.382.4753.7%0.50213.5651.7%
T3a/control_sexy#3$0.000.00 (0.00)$0.000.00$0.000.0000.0%0.000——

16 付费用户门槛(#37):上表中付费用户数不足 16 的行只能看方向,不能看幅度。

五个付费用户上的 2.35,其中 51.7% 来自一个人。该样本量下读不出来;第 5 节里这个广告系列的 1.67,就 是这一格被那条什么也没产出的实验臂摊薄之后的结果。

2026-08-16T16:05:37.523350 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.0 0.5 1.0 1.5 2.0 2.5 payers / installs % Payer rate by day — every asset (UTC) T3a/utility_sexy#2
分素材的每日付费率
2026-08-16T16:05:37.572985 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 within-day USD per install ARPU by day — every asset (UTC) T3a/utility_sexy#2
分素材的每日 ARPU
2026-08-16T16:05:37.619630 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 0 2 4 6 8 10 12 14 16 within-day USD per payer ARPPU by day — every asset (UTC) T3a/utility_sexy#2
分素材的每日 ARPPU
2026-08-16T16:05:37.668676 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ 07-23 07-24 07-25 07-26 07-27 07-28 07-29 07-30 07-31 day (UTC) 1.700 1.725 1.750 1.775 1.800 1.825 1.850 1.875 revenue / spend Within-day ROAS by day — every asset (UTC) T3a/utility_sexy#2
分素材的每日当日回收 ROAS
2026-08-16T16:05:37.736203 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 period (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 period are not plotted. Composite — every asset across every measure, 2026-07-23..2026-07-31 T3a/utility_sexy#2
综合图:各素材在各口径上按本周期归一化后的表现

C. 分素材 — 国家结构

⚠ 本表按账户日切,而本页其余各表都在 UTC 时钟上。 Meta 不提供小时 × 国家的交叉表,因此分国家的投放数据只存 在于广告账户自己的 UTC-7 日上,无法重新拼接。它与上面的 A、B 两表相差七小时;本表只读结构,不与它 们比水平。

B 表里某一行读不读得出来,看这里。最后一列是该素材自身当日回收中来自美国的占比:当它远高于 该素材的美国花费占比时,这条素材的回报是它买到的那块地域的属性。

素材花费印度占比美国占比其他占比美国占其自身收入
T3a/utility_sexy#2$28.9079.4%0.0%20.6%0.0%
T3a/control_sexy#3$11.7857.8%0.0%42.2%—

两条实验臂买到的结构并不一样:印度占比 79.4% 对 57.8%,其余落在 T3 的其他国家。印度是这几个国家里 最便宜、转化也最好的一个,所以上面那处投放差距里,有一部分是竞价把两条实验臂放到了不同的地方。

2026-08-16T16:05:37.370445 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ T3a/utility_sexy#2 0 20 40 60 80 100 % of that asset's spend Where each asset bought — share of its own spend by region, 2026-07-23..2026-07-31 (META days) IN ID PH PK NG other
各素材自身花费的地域结构
2026-08-16T16:05:37.407664 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ T3a/utility_sexy#2 0.0 0.1 0.2 0.3 0.4 USD per install Cost per install by region — every asset, 2026-07-23..2026-07-31 (META days) IN ID PH PK
各素材分地域的单次安装成本
2026-08-16T16:05:37.441970 image/svg+xml Matplotlib v3.11.1, https://matplotlib.org/ T3a/utility_sexy#2 0 10 20 30 40 50 60 USD Within-day revenue by region — every asset, 2026-07-23..2026-07-31 (META days) IN PH
各素材分地域的当日回收

7口径与边界

⚠ 口径校验落在区间之外,这里不做四舍五入把它凑进来。 在账户时区上(校验只在这个时钟上有效), 我们自己的记录是 Meta 入账 $3,242.34 的 1.07 倍,有记录的区间是 0.91 到 1.06。在本报告自己的 UTC 时 钟上,同一个比值是 1.08,而它根本不构成一次有效的校验,因为 Meta 的转化金额盖在转化发生的小时上, 那是分小时导出里唯一一列无法跨两个账户日重新拼接的。两个数字都碰不到当日回收口径,后者从不读 Meta 那一列;它们限定的是任何把我们的数字与 Meta 放在一起说的句子。

⚠ 付款关联的修复够不到 7 月。 这个口径会丢掉任何无法关联到人的付款,2026-08-16 那次修复靠的是 按账户 UUID 关联,而注册推送从 2026-08-13 18:52 UTC 才开始带这个字段。7 月这批人的注册远早于此,修 复伸不过去,早期数据流当时归不上的那些付款仍然归不上(asset_performance_report_2026-08-15.md)。 7 月缺掉的部分会把收入读低,而口径校验读出来高于 Meta,所以这两件事不会叠加。

德语区投放里 ch_ 与 de_ 两个前缀标的是投放地域的切分。 这两组用的是同样的镜头、同样的字 幕、同样的配音,只有背景音乐不同(ad_ops/dach_stage2_asset_proposal_2026-07-30.md)。 DE/ch_arab_gulf_6s 买了 498 次展示、一次安装,单次 $21.750。

⚠ 本页没有任何一处是同预算区间的。 预算区间是广告组日预算没有变动的一段时间,而在这个账户上预 算是一个地域旋钮(DECISION_LOG.md #47)。估计器手工记录的预算区间表只覆盖旧广告系列和德语区投 放,六个 T1/T2/T3 广告系列一张也没有。在这九天里,单是旧广告系列就从每日 $150 动到 $250、$500 与 $875,而 07-23 那一轮的广告组从 $5 提到 $20,时间点没有任何记录。因此上面每一个分广 告系列和分素材的数字,都跨越了预算变动。

⚠ 分国家一节与所有 C 表走的都是账户自己的日,比本页其余部分晚七小时。在旧广告系列上,这个窗口 伸进了 2026-08-01 UTC,给出 $408.99 与 $150.33,而 A 表给出的是 $312.93 与 $8.64。它们只读结构,不 读水平。

这里的同期群已经走满,本系列里没有哪一份日报能这样说。 5,069 次安装在付款数据截止时间上的年龄是 371.0 到 567.3 小时,H=24 覆盖率 100%,所以 $2,697.93 已经冻结。同期群那一列不会再往上走,也不需要挂 任何覆盖率说明。

这个月测的那套分组,在这个月里就已经停用了。 DECISION_LOG.md #36 在 2026-07-26 放弃了 T1/T2/T3 这套切分:这些分组是按假定成本凑出来的,按实测付费率读出来次序是反的,而且每一组内部把性 质完全不同的国家平均在了一起。这六个广告系列写在这里,是因为它们花了钱。

2026-07-25 那一轮回答不了它自己提的问题,本报告继承了这一点。 十八个格,每格一个广告组,任何一 处都没有重复,16 付费用户门槛就是从这里来的(DECISION_LOG.md #37)。九天的买量下来,八个广告系列 里有三个、整份素材清单里有五条,过得了这道门槛。

本页贯穿着两套安装数,它们从不交叉相除。 Meta 的 4,561 对我们自己的 5,069 次注册。投放漏斗的分子分 母都用 Meta 的;付费率与每一个收入口径的分母用我们自己的。

本页在账户层带 18 张图,八个标签页里各再带 18 张。 每一张在发布时都以 SVG 内联进页面,所以即使 一个标签页被折起来,它的图也仍然随页面一起发出去,页面因此很大。把它们从标签页里搬出去,等于把本格 式已经删掉的图表附录又请回来,所以它们留在原处。

8要把这些问题定下来,需要什么

印度的 1.00 到底算不算一门生意。 它是本月唯一一格背后有足够多付费用户、可以当水平来读的,而它 正好停在为它付出的钱上,还没有扣掉应用商店的分成。7 月内部没有任何东西能改变这一点;8 月的报告是第 一份能说出它往哪边走的。

7 月有没有哪条素材在收入上与另一条不同。 五条素材过了 16 付费用户门槛,其中四条在同一个广告系 列里,而且没有一条是同预算区间的。这个账户自己的可判定性算法要求每条实验臂上千次安装,远高于一轮九 天、十八个格的测试能产出的量,所以老实的答案是:7 月测的不是素材,是地域。

那一轮分组测试到底花了多少。 第 5 节把六个分组广告系列逐个与旧广告系列、德语区投放放在一起标价 。那张表就是完整的账,读它不需要第二个周期;它说不了的是同样这笔钱如果放进一个不设限的全球买量里会 有什么结果,而账户在 2026-07-30 转向的正是那个形态。

9交给下一次读数的东西

8 月是本系列里第一个身后有东西的周期,而这份报告就是那个东西。它交出去的是一个水平:账户 CPI $0.538,当日回收回报 0.95,$2,344.02 背后是 233 个付费用户的付费率,以及一个已经定住的同期群数字。 任何 8 月的对比都对着这四个数来做。

在做那次对比时,有三件事要放在心里。这九天并不是同一个账户的九天,因为八个广告系列里有六个在这九天 之内被关掉或改过结构。旧广告系列只跑了其中两天,而且在这两天里把预算提到了原来的几倍,所以拿 7 月 的它去比 8 月的它,是拿两天的爬坡去比另一段爬坡。还有,分国家的切分两头都走账户自己的时区,所以到 了 8 月它仍然只是一个结构读数。

10复现

报表时区为 UTC。B1 的导出盖的是它自己账户时区(UTC-7)的时间戳,因此上面每一行投放数据都由两个 Meta 日重新拼接。分国家的表与所有 C 表不做拼接,并在出现的地方标注。

运行目录_runs/2026-08-16_periods/,由 generate.sh 生成
周期2026-07-23 → 2026-07-31 UTC,九天,之前没有任何周期
配置_runs/2026-08-16_receipt_series/run_config_full.json;口径校验走 run_config_full_meta.json
付款数据截止时间报告为 2026-08-16 10:59 UTC,口径校验为账户时区的 2026-08-16 03:59
Meta 导出分小时按账户日取自 run config 的存档,已验证带满 24 个小时桶;分日与国家×日取自 _runs/2026-08-16_2026-08-15_daily_auto/exports/
注册与付款_runs/2026-08-16_2026-08-15_daily_auto/lum
汇率_runs/2026-08-02_today_read/fx_rates.json
预算区间没有任何预算区间表覆盖 T1/T2/T3 广告系列;旧广告系列的表是 ad_ops/asset_estimator/cohorts.py 里的 REGIMES_260730

7 月账户花掉、而本报告不统计的部分。 2026-07-15 至 2026-07-21,账户上跑的是 youguqi_ 黑哈测 试,$477.52,投的是另一个 App,走的是借来的归因栈。2026-07-30 与 2026-07-31,账户上还跑了 0730_starfall_,$71.62,是替第三方的 App 代投。两笔都不是这个 App 的钱,在 run config 里都 标为跳过;账户自己的投放导出里 7 月花费比本页的 $2,455.00 多,就是因为它们。

CFG=_runs/2026-08-16_receipt_series/run_config_full.json
python -m ad_ops.measures      --config $CFG --from 2026-07-23 --to 2026-07-31
python -m ad_ops.report_tables  --config $CFG --from 2026-07-23 --to 2026-07-31
python -m ad_ops.report_tables  --config $CFG --from 2026-07-23 --to 2026-07-31 --campaign-table
python -m ad_ops.report_tables  --config $CFG --from 2026-07-23 --to 2026-07-31 --campaign t3_0725
python -m ad_ops.report_charts  --config $CFG --from 2026-07-23 --to 2026-07-31 \
    --out ad_ops/figures/month_2026-07 --md _runs/2026-08-16_periods/chartblock_month_2026-07.md \
    --relpath figures/month_2026-07
python -m ad_ops.measures --config _runs/2026-08-16_receipt_series/run_config_full_meta.json \
    --from 2026-07-23 --to 2026-07-31

上面最后一条命令是唯一一条其口径行构成有效校验的,因为它跑在账户自己的时区上。generate.sh 一次跑 完六个周期和每一个标签页;每个广告系列写进自己的 --out 目录,因为图表文件名里不带广告系列,两个 广告系列共用一个目录会互相覆盖。

11附录 — 定义

报表周期是一串 UTC 日。 Adjust 按 UTC 报数,按 UTC 分桶的报告与归因面板天然吻合,不需要任何人 在脑子里换算。B1 固定在 UTC-7,因此这个账户上的一个 UTC 日等于账户日 D−1 的 17:00 到 23:59加上账户 日 D 的 00:00 到 16:59。

分小时一节把一天中的小时跨周期合并。 九天有 216 个小时,而表格只有 24 行,所以每一行是九个小时 的买量叠在一起,每天各出一个。它读的是一天之内的形状,不是时间轴,表里没有哪一行是一次事件。

总量在长度不同的周期之间不可比,比率可以。 本报告只印一个周期、完全不做对比,所以这里不存在这 个风险;但这条规则正是 8 月的报告要比 CPM、CTR、点击→安装、CPI、ROAS、付费率、ARPU 与 ARPPU 的原 因。

三种收入口径,从不只用一种,从不混用,永远标注:

口径计什么本期结束后还会动吗
入账收入在本周期之内到账的钱,不论付费用户哪天安装午夜后约四小时结清,此后固定。Meta 自己的数字就是这个口径
当日回收当天安装并且在当天结束前付费永远不动。 午夜封存
同期群 H 小时当天的安装,各自从自己安装起算 H 小时内的付费。默认视野 H=24在所有安装都活满 H 小时前持续上升,之后冻结。必须写出覆盖率

跨日对比用当日回收口径,因为它不可变且不需要视野,代价是它会按一天中的时点做截尾:01:00 安装的人在 午夜前有 23 小时可以付费,23:00 安装的人只有一小时。固定视野的同期群口径对时点是中性的。入账是 Meta 会拿来跟你说的数字,写出来是为了解释差额,不是为了排次序。三种口径都读在我们自己的付款记录上 ;Meta 只能表达入账,因此它跟在那一列旁边的括号里,永远不单独成行。

放到一个周期上,入账收入按付费用户已安装的天数拆分。 在一天之内,已安装天数与安装日是同一件事 ;到了九天,两者就分开了,这个拆分中安装当天那一行,恰好等于当日回收的合计。

存在两套安装数,它们从不交叉相除。 Meta 的 Leads 与我们自己的注册数相差 5% 到 10%。投放漏斗用 Meta 的,因此 CPI 与点击→安装的分子分母都在同一把尺子上。付费率与所有收入口径用我们自己的,因此付费用 户与他们来自的安装是同一批人。

预算区间是广告组日预算没有变动的一段时间。它之所以重要,是因为预算在这个账户上是一个地域旋钮: 调高预算会买到更宽、更便宜的国家,跨过一次预算变动去比较两条素材,比的是两批不同的人。

16 付费用户门槛(DECISION_LOG.md #37):付费用户数低于约 16 时,一个收入数字只能看方向,不能 看幅度。这个账户上每次安装的收入是一个几乎处处为零、偶有大额的向量,要把两条素材分出五分之一的差距 ,每条实验臂需要上千次安装。

这些广告系列的名字是什么意思。 T1、T2、T3 是账户成组买入的国家分组,由 DECISION_LOG.md #24 固定下来:T1 是美国、加拿大、英国、澳大利亚与新西兰;T2 是阿联酋、沙特、马来西亚、南非与墨西哥;T3 是印度、印尼、尼日利亚、巴基斯坦与菲律宾。它们一共跑过两轮:2026-07-23 那一轮,素材标签前缀里带 a;2026-07-25 那一轮,带 b。第一轮测的是两个文案方向,utility 讲把一张照片变成视频, control 讲一个按你的意思动的陪伴对象,各用一条十五秒的片子。第二轮把这两个方向乘上三种时长,在三 个分组里各跑一遍,十八个格、每格一个广告组。sasian_ 与 latina_es_ 是 2026-07-26 加进第二轮印度 与墨西哥广告组的南亚素材包与西语素材包;紧跟其后的词是片中的服装,数字是片长。旧广告系列是 2026-07-30 取代这套分组结构的全球买量,德语区广告系列是同一天开投的 DACH 测试。

Internal — noindex. Not for distribution outside the team.