MIRANEST › Winning Ads
Top 10 videos by orders, Jul 31 – Sep 28, CPO under ¥27k. Together: 131 of the account’s 250 orders on 41% of spend.













11–13 added Nov 4: Sep 29 – Nov 3, same rule (videos by orders, CPO under ¥27k), compared against that window. The ten and the sections below stay on Jul 31 – Sep 28. Meta’s purchase tracking was broken Oct 13–20.
Group ANV1002-1 · V2004-3 · V2004-1 · N304-1 · NV1002-4 · V2006-1 · NV1034-1![]() ![]() ![]() ![]() ![]() ![]() ![]() | Group BF02 · F03 · V61![]() ![]() ![]() | |
|---|---|---|
| Orders | 97 | 34 |
| CPO | ¥16.6k | ¥20.1k |
| Hook — stop in 3s | 17.6% | 20.2% |
| Hold — watch through | 4.1% | 2.6% |
| Click rate | 0.87% | 0.90% |
| Page visitors who buy | 4.3% | 3.0% |
| Shopify sessions → bought | 2.1% | 1.5% |
| Orders from Reels | 52% | 32% |
Group B stops people a little better. Group A keeps them watching and its visitors buy more — so A wins in the middle and the end of the video, not the hook.
| 0–3s | Hook | A Pom on screen + one line of text from frame 1. Proven: “Does your dog look like this?” · “My Pomeranian used to…” · “They said his fur would NEVER grow back… They were wrong” (best hook, 30%) · “What is MIRANEST?” |
| 3–10s | What didn’t work | “It’s not fish oil. It’s not your traditional allergy chew.” · “stop wasting my money on solution after solution” |
| 10–20s | Proof | Ideally ONE dog, bald → full coat, labelled Before » After — V2004-3 does this and it’s the only big winner that didn’t fade |
| by ~0:20 | Product | The stick in a hand, then a Pom licking it |
| — | Why it works | “They work from the inside” · “from the inside out” |
| — | Daily habit | “Just one MIRANEST jelly stick a day” |
| last 5s | Close | “Try MIRANEST today with 20% off and a 90-day money-back guarantee” |
9:16 · captions the whole way · a new shot every ~2s · a woman’s voice, or music + caption pills · our Japan monitor Poms only. Length isn’t the pattern — winners run 24s to 65s.
What to make: the build plans now have their own tab — What to make →
| Ad | Group | Type | Length | Voice | Before / after | Product on screen | Offer | Hook | Conv. | CPO | Orders | Reels | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
![]() | NV1002-1 | A | Explainer | 1:04 | Female VO | No | 0:00 | 20% off + 90-day | 16.5% | 4.6% | ¥17.9k | 26 | 11/17 |
![]() | F02 | B | Reveal | 0:18 | Music only | Yes | 0:11 | None | 21.6% | 2.8% | ¥20.7k | 26 | 5/17 |
![]() | V2004-3 | A | Owner story | 0:39 | Female VO | Yes | 0:15 | 20% off | 22.2% | 4.3% | ¥16.4k | 24 | 8/14 |
![]() | V2004-1 | A | Owner story | 0:39 | Female VO | No labels | 0:15 | None | 26.0% | 4.6% | ¥14.5k | 12 | 5/9 |
![]() | N304-1 | A | Problem → proof | 0:27 | Female VO | Yes | 0:19 | 90-day | 21.1% | 3.7% | ¥17.7k | 11 | 2/7 |
![]() | NV1002-4 | A | Explainer | 1:05 | Female VO | No | 0:00 | 20% off + 90-day | 5.4% | 7.0% | ¥8.4k | 10 | 2/7 |
![]() | V2006-1 | A | Presenter | 0:24 | Female, on camera | No | 0:00 | Sub & save 20% | 14.9% | 3.0% | ¥24.2k | 9 | 3/8 |
![]() | NV1034-1 | A | They said / wrong | 0:38 | Music only | Yes ×3 | 0:24 | 20% off + 90-day | 30.3% | 4.2% | ¥17.4k | 5 | 3/4 |
![]() | F03 | B | Education | 1:03 | Female VO | Inset | 0:52 | None | 16.6% | 3.4% | ¥19.8k | 5 | 2/6 |
![]() | V61 | B | Offer | 0:15 | Music only | No | 0:00 | 20% + $10 + gift | 9.3% | 6.7% | ¥14.6k | 3 | 1/2 |
Green = above the account median (hook 21.3%, conversion 3.4%). Reels = traced orders from Facebook or Instagram Reels.
Across the ten, hook rate and CPO barely move together (rank correlation 0.02). The cheapest ad, NV1002-4, has the worst hook in the account (5.4%). The best hook, NV1034-1 (30.3%), sits mid-table.
Page conversion lines up with CPO almost perfectly (rank correlation -0.90). The cheapest ads are the ones whose visitors buy.
Show the change — a bald dog, then a coated one (F02, V2004-3, V2004-1, N304-1, NV1034-1): 78 orders. Explain the product — open on the stick (NV1002-1, NV1002-4, V2006-1): 45 orders.
9 of 10 feature Pomeranians (the tenth has no dog). 7 have a voiceover — all female. The other 3 are music plus on-screen text.
The nine 9:16 ads got 37 of 74 traced orders from Reels (50%; all ads 34%). The one square ad, F02, got 7 of 17 from Instagram Feed.
NV1002-4 is NV1002-1’s voiceover with new visuals: ¥8.4k CPO vs ¥17.9k. V2004-3 and -1 share audio too — Meta put ¥393k behind -3 and ¥174k behind -1.
Four had a strong run of about three weeks and then got pricier — NV1002-1, F02, V2004-1, V2006-1. V2004-3 and NV1002-4 kept going. Account CPO also rose in September (≈¥17k late August → ≈¥26k), but those four rose faster. Ten ads is a small sample — read these as leads, not laws.
Orders and CPO from the HK hub’s daily Meta rows, Jul 31 – Sep 28 · hook = 3-second views ÷ impressions · conversion = orders ÷ landing-page views · ties broken by lower CPO · videos only · left out for CPO over ¥27k: N164 (¥40.1k)