Every month this account has ever spent money — 2023-07 to 2026-07. What is actually winning, and where the trend is going.
Conversions are Meta omni_purchase throughout — the same metric the V2 report uses.
Coverage. The account opened 2022-09-05, but Meta refuses insights older than 37 months, so this starts 2023-07. That costs almost nothing — only ¥39,093 of the ¥67,605,523 lifetime total falls outside the window. In practice this is the whole account. One caveat: 2026-07 is a partial month (through 2026-07-20), so it is not comparable to full months.
Bar height is monthly spend, bar colour is that month's cost per order, and the white line is the CPO itself. When the bars grow and the line climbs with them, the account is buying worse orders for more money.
The single most important number here. In the 16 months where spend stayed under ¥2,500,000, the account bought orders at ¥19,503 (ROAS 0.87). In the 6 months it pushed past ¥4,000,000, the same account paid ¥31,733 (ROAS 0.52) — 63% more per order. Scale moves CPO here more than creative, audience or placement do.
| Month | Spend | Orders | CPO | vs prev | ROAS | AOV | CVR | CTR | Ads live | New ads |
|---|---|---|---|---|---|---|---|---|---|---|
| 2023-07 | ¥40,020 | 0 | no orders | — | — | — | 3.48% | 1 | 1 | |
| 2023-08 | ¥10,214 | 0 | no orders | — | — | — | 3.43% | 1 | 0 | |
| 2023-12 | ¥9,039 | 0 | no orders | — | — | — | 1.47% | 3 | 3 | |
| 2024-02 | ¥10,741 | 0 | no orders | — | — | — | 0.56% | 1 | 1 | |
| 2024-05 | ¥5,136 | 0 | no orders | — | — | — | 1.53% | 6 | 6 | |
| 2024-06 | ¥174,339 | 0 | no orders | — | — | — | 2.73% | 23 | 17 | |
| 2024-07 | ¥314,298 | 0 | no orders | — | — | — | 3.23% | 12 | 7 | |
| 2024-08 | ¥329,207 | 17 | ¥19,365 | 1.53 | ¥29,555 | 1.61% | 3.00% | 40 | 34 | |
| 2024-09 | ¥546,834 | 25 | ¥21,873 | ▲ 13% | 0.85 | ¥18,686 | 1.82% | 2.34% | 48 | 34 |
| 2024-10 | ¥857,315 | 32 | ¥26,791 | ▲ 22% | 0.70 | ¥18,778 | 3.84% | 1.60% | 35 | 27 |
| 2024-11 | ¥1,137,847 | 30 | ¥37,928 | ▲ 42% | 0.48 | ¥18,126 | 4.05% | 0.99% | 38 | 32 |
| 2024-12 | ¥1,085,022 | 24 | ¥45,209 | ▲ 19% | 0.44 | ¥19,693 | 2.14% | 2.18% | 28 | 25 |
| 2025-01 | ¥1,055,493 | 31 | ¥34,048 | ▼ 25% | 0.56 | ¥19,236 | 3.31% | 1.37% | 26 | 24 |
| 2025-02 | ¥1,053,837 | 52 | ¥20,266 | ▼ 40% | 0.96 | ¥19,369 | 3.08% | 1.93% | 28 | 19 |
| 2025-03 | ¥1,020,878 | 23 | ¥44,386 | ▲ 119% | 0.38 | ¥16,758 | 1.45% | 2.70% | 56 | 44 |
| 2025-04 | ¥1,027,637 | 40 | ¥25,691 | ▼ 42% | 0.74 | ¥18,964 | 1.91% | 2.25% | 76 | 71 |
| 2025-05 | ¥1,422,256 | 128 | ¥11,111 | ▼ 57% | 1.48 | ¥16,481 | 3.22% | 1.78% | 179 | 143 |
| 2025-06 | ¥1,513,671 | 97 | ¥15,605 | ▲ 40% | 1.01 | ¥15,838 | 2.99% | 1.28% | 371 | 256 |
| 2025-07 | ¥1,734,192 | 120 | ¥14,452 | ▼ 7% | 1.13 | ¥16,301 | 2.97% | 1.48% | 217 | 135 |
| 2025-08 | ¥2,042,284 | 118 | ¥17,307 | ▲ 20% | 0.91 | ¥15,728 | 2.59% | 1.65% | 188 | 140 |
| 2025-09 | ¥2,056,199 | 132 | ¥15,577 | ▼ 10% | 0.96 | ¥15,027 | 2.67% | 1.58% | 278 | 209 |
| 2025-10 | ¥2,197,921 | 114 | ¥19,280 | ▲ 24% | 0.88 | ¥16,963 | 2.47% | 1.56% | 263 | 226 |
| 2025-11 | ¥4,206,742 | 217 | ¥19,386 | ▲ 1% | 0.83 | ¥16,087 | 3.20% | 1.43% | 194 | 141 |
| 2025-12 | ¥5,866,329 | 239 | ¥24,545 | ▲ 27% | 0.66 | ¥16,290 | 2.91% | 1.10% | 239 | 149 |
| 2026-01 | ¥7,440,965 | 243 | ¥30,621 | ▲ 25% | 0.56 | ¥17,105 | 1.98% | 1.18% | 360 | 242 |
| 2026-02 | ¥9,245,099 | 241 | ¥38,361 | ▲ 25% | 0.43 | ¥16,353 | 1.72% | 0.92% | 233 | 115 |
| 2026-03 | ¥8,365,473 | 194 | ¥43,121 | ▲ 12% | 0.39 | ¥16,717 | 1.76% | 0.85% | 302 | 195 |
| 2026-04 | ¥4,573,399 | 117 | ¥39,089 | ▼ 9% | 0.42 | ¥16,467 | 1.63% | 0.86% | 339 | 225 |
| 2026-05 | ¥3,762,656 | 103 | ¥36,531 | ▼ 7% | 0.46 | ¥16,654 | 2.34% | 0.73% | 214 | 136 |
| 2026-06 | ¥2,596,006 | 106 | ¥24,491 | ▼ 33% | 0.67 | ¥16,489 | 3.11% | 0.69% | 160 | 62 |
| 2026-07 * | ¥1,865,381 | 91 | ¥20,499 | ▼ 16% | 0.78 | ¥16,046 | 3.23% | 0.71% | 188 | 45 |
The same 37 months grouped into the eras the data falls into. Bar length is average monthly spend; colour is that phase's cost per order.
What scaling cost. Going from the efficient run (¥1.8M/mo) to the scale-up (¥7.0M/mo) added ¥24,158,085 of spend and bought 425 extra orders — a marginal cost of ¥56,843 per additional order, 3.7× what the account paid before it scaled.
And what the retreat proves. Cutting ¥22,327,166 back out gave up 717 orders that were costing about ¥31,140 each. Efficiency has improved every month since: ¥39,089 → ¥36,531 → ¥24,491 → ¥20,499. The cut was the right call.
Every proven creative family placed by how many orders it produced (across) against what each order cost (up). Bubble size is spend. Grouped by family, not ad row — the same creative runs as many ads across ad sets and two naming eras, so 2,764 ad rows collapse to 1,763 families.
Read the bottom-left corner. Low on the chart means cheap orders; far left means small volume. Families like F03 (¥7,978), 1002 (¥9,396) and F303 (¥9,492) sit down there — proven cheap, and tiny. The big bubbles high up are where the money actually went.
| Creative family | Ad rows | Spend | Orders | CPO | ROAS | Share of orders |
|---|---|---|---|---|---|---|
| 1. 5 Image | 59 | ¥4,607,172 | 176 | ¥26,177 | 0.76 | 6.9% |
| 2. V1012 Video | 22 | ¥2,426,863 | 130 | ¥18,668 | 0.85 | 5.1% |
| 3. V36-1 Video | 3 | ¥2,305,489 | 120 | ¥19,212 | 0.83 | 4.7% |
| 4. 51 Image | 33 | ¥1,330,940 | 105 | ¥12,676 | 1.28 | 4.1% |
| 5. lara_n_lulu Video UGC General-LP1 | 7 | ¥1,943,926 | 79 | ¥24,607 | 0.63 | 3.1% |
| 6. 96 Image Review Alopecia-LP3 | 5 | ¥1,845,988 | 75 | ¥24,613 | 0.64 | 3.0% |
| 7. Mochi 1 Video | 2 | ¥1,680,602 | 70 | ¥24,009 | 0.63 | 2.8% |
| 8. V46 Video | 7 | ¥1,295,381 | 63 | ¥20,562 | 0.80 | 2.5% |
| 9. V35-3 Video | 2 | ¥893,054 | 50 | ¥17,861 | 0.95 | 2.0% |
| 10. V94-3 Video | 2 | ¥1,308,297 | 47 | ¥27,836 | 0.59 | 1.9% |
| 11. 71 Image | 26 | ¥583,481 | 46 | ¥12,684 | 1.37 | 1.8% |
| 12. N165-1 Video Story Alopecia-LP1 | 2 | ¥1,020,600 | 37 | ¥27,584 | 0.64 | 1.5% |
| 13. 177 Image | 5 | ¥1,275,213 | 36 | ¥35,423 | 0.47 | 1.4% |
| 14. 40 Image | 19 | ¥606,828 | 34 | ¥17,848 | 0.98 | 1.3% |
| 15. V39-2 Video Offer General-LP1 | 3 | ¥643,798 | 34 | ¥18,935 | 0.88 | 1.3% |
| 16. V54 Video | 5 | ¥985,643 | 34 | ¥28,990 | 0.60 | 1.3% |
| 17. 78 Image Review Alopecia-LP3 | 19 | ¥395,357 | 33 | ¥11,981 | 1.34 | 1.3% |
| 18. 1010-3 Image | 17 | ¥401,856 | 27 | ¥14,884 | 1.02 | 1.1% |
| 19. 188-8 Image | 3 | ¥372,412 | 26 | ¥14,324 | 1.17 | 1.0% |
| 20. V39-4 Video Offer General-LP1 | 3 | ¥444,812 | 26 | ¥17,108 | 0.99 | 1.0% |
| 21. V1-7 Video | 4 | ¥540,111 | 26 | ¥20,774 | 0.76 | 1.0% |
| 22. V49-2 Image | 11 | ¥616,324 | 24 | ¥25,680 | 0.65 | 0.9% |
| 23. V96-3 Video | 4 | ¥764,602 | 24 | ¥31,858 | 0.56 | 0.9% |
| 24. N173-1 Video Education Alopecia-LP1 | 1 | ¥477,183 | 24 | ¥19,883 | 0.84 | 0.9% |
| 25. C1 Video | 16 | ¥624,975 | 22 | ¥28,408 | 0.64 | 0.9% |
The top bar is where the budget went. The bottom bar is where the orders came from. Both are split by cost per order, so any mismatch between them is money in the wrong place.
This is the headline finding. The account already owns creatives that produce orders at a profit. The 9 families under ¥15,000 CPO ran at a blended ¥12,403 with ROAS 1.31 — profitable on the first order alone, before any repeat purchase. They received 5.3% of the money while delivering 11.4% of the orders. The problem is not that MIRANEST USA lacks winning creative; it is that winning creative never got the budget.
And look at the grey. Nearly half the account — 48.6% of all spend — went to creatives that never reached 10 orders, so they were never gradeable either way. That is ¥32.9M spent without buying a usable answer.
| Creative family | Ad rows | Spend | Orders | CPO | ROAS | % of account spend |
|---|---|---|---|---|---|---|
| F03 Video Education | 1 | ¥151,576 | 19 | ¥7,978 | 1.96 | 0.2% |
| 1002 Image | 2 | ¥103,353 | 11 | ¥9,396 | 1.71 | 0.2% |
| F303 Image Before/After | 2 | ¥123,390 | 13 | ¥9,492 | 1.82 | 0.2% |
| 78 Image Review | 19 | ¥395,357 | 33 | ¥11,981 | 1.34 | 0.6% |
| 51 Image | 33 | ¥1,330,940 | 105 | ¥12,676 | 1.28 | 2.0% |
| 71 Image | 26 | ¥583,481 | 46 | ¥12,684 | 1.37 | 0.9% |
| FV1001 Video | 5 | ¥134,602 | 10 | ¥13,460 | 1.12 | 0.2% |
| 188-8 Image | 3 | ¥372,412 | 26 | ¥14,324 | 1.17 | 0.6% |
| 1010-3 Image | 17 | ¥401,856 | 27 | ¥14,884 | 1.02 | 0.6% |
| V78-4 - Alopecia LP Video | 5 | ¥150,659 | 10 | ¥15,066 | 1.09 | 0.2% |
| V39-4 Video Offer | 3 | ¥444,812 | 26 | ¥17,108 | 0.99 | 0.7% |
| 40 Image | 19 | ¥606,828 | 34 | ¥17,848 | 0.98 | 0.9% |
| V35-3 Video | 2 | ¥893,054 | 50 | ¥17,861 | 0.95 | 1.3% |
| V1012 Video | 22 | ¥2,426,863 | 130 | ¥18,668 | 0.85 | 3.6% |
| V39-2 Video Offer | 3 | ¥643,798 | 34 | ¥18,935 | 0.88 | 1.0% |
| V36-1 Video | 3 | ¥2,305,489 | 120 | ¥19,212 | 0.83 | 3.4% |
| N173-1 Video Education | 1 | ¥477,183 | 24 | ¥19,883 | 0.84 | 0.7% |
| V46 Video | 7 | ¥1,295,381 | 63 | ¥20,562 | 0.80 | 1.9% |
Format comes from Meta's creative object, never from the ad name — 2,888 of 2,764 ad rows are tagged authoritatively. Bar length is cost per order, so shorter is better.
Image wins, narrowly. Image buys orders at ¥25,954 against ¥27,284 — a 5% gap on a near 50/50 budget split. Format is not the lever: the spread within each format is far bigger than the spread between them. Worth noting the cheapest proven families are mostly images, so "video always wins" does not hold on this account.
107 campaigns have run here. Bar length is lifetime spend, colour is that campaign's cost per order.
The pattern repeats at campaign level. Campaigns above ¥2.0M lifetime spend bought orders at ¥27,699; campaigns between ¥300k and ¥2.0M bought them at ¥22,451. The best campaign with real volume is ASC #8 at ¥11,662 (ROAS 1.41) on just 2.1% of spend — the Advantage+ Shopping campaigns punch far above their budget share.
All three re-pulled in six-month slices, because Meta silently drops purchase actions from breakdown queries spanning three years. Each now ties the account total exactly.
Three things worth acting on. Instagram feed is the biggest single line of spend (31.4%) and costs 32% more per order than Facebook feed (¥30,075 vs ¥22,836) — the larger allocation is going to the weaker placement. Desktop is the cheapest device at ¥18,226 but gets only 2.1% of spend — small volume, so widen it as a test rather than swinging budget in. Android costs 25% more per order than iPhone. And audience is not a lever: every age × gender cell with real volume sits between ¥24,859 and ¥31,249.
2,764 ad rows have spent money here. Most never spent enough to prove anything either way.
Three quarters of the ads produced nothing. 2,077 of 2,764 ad rows generated zero orders while consuming ¥15.4M. Some of that is the unavoidable cost of testing, but the median ad only ever received ¥4,541 — less than a single order costs at this account’s CPO. Most “tests” were therefore unreadable: too small to win, too numerous to manage.
Under ¥2,500,000/month it buys orders at ¥19,503 (ROAS 0.87). Above ¥4,000,000/month it pays ¥31,733 (ROAS 0.52). Pushing from the efficient run to the scale-up cost ¥56,843 per incremental order. There is a ceiling somewhere between ¥2.0M and ¥4.0M/month, and crossing it destroys the economics rather than bending them.
9 creative families have produced 10+ orders at under ¥15,000 CPO, blended ¥12,403 at ROAS 1.31. Together they received 5.3% of lifetime spend. The best — F03 (¥7,978), 1002 (¥9,396), F303 (¥9,492) — are proven and under-funded. This is the largest and most fixable gap in the account.
2,077 ads (75.1%) produced zero orders on ¥15.4M. The median ad received ¥4,541 — far below the budget needed to read a result. More creative is not the constraint; giving each creative enough budget to return a verdict is.
Video and Image differ by only 5%. Every meaningful age × gender cell sits within a few thousand yen of the others. The one allocation worth changing is placement: Instagram feed is the biggest spend line and costs 32% more per order than Facebook feed.
Since the retreat began the trend has improved every month: 2026-04 ¥39,089 → 2026-05 ¥36,531 → 2026-06 ¥24,491 → 2026-07 ¥20,499. The last three months blend to ¥27,413 — still the middle band, but the 2 most recent months are back under ¥25,000, which had not happened since 2025-12. It is not yet back to the ¥15,468 of the efficient run, so the task is to hold this while adding budget slowly rather than stepping back toward ¥9.2M/month. The final month is partial, so read it as directional.
One honest caveat on ROAS. Lifetime ROAS of 0.63 is first-order revenue only, as Meta reports it. MIRANEST is a repeat-purchase product, so the real question is whether lifetime value clears the ¥26,664 acquisition cost — and that answer lives in Shopify cohort data, not in this account. Read ROAS here for relative comparison between months, creatives and placements, which is what it is reliable for; not as the profit line.
state/alltime.json · renderer render_review.py