01📱 Platform choice: work back from your audience's age
Start platform choice not from which social network is trending, but from where the people you want to reach actually are. In MIC's FY2025 survey (ages 13–79), overall usage was YouTube 81.5%, Instagram 54.8%, TikTok 36.7% and Facebook 27.0%. TikTok reports about 49.5 million monthly users in Japan (end of May 2026, TikTok's own figure excluding overlap with TikTok Lite).
More important than the overall figures are the differences by age.
| Age group | YouTube | TikTok | ||
|---|---|---|---|---|
| Teens | 95.0% | 69.3% | 67.9% | 14.3% |
| 20s | 96.3% | 82.6% | 61.5% | 23.4% |
| 30s | 94.0% | 76.6% | 44.7% | 37.4% |
| 40s | 91.6% | 67.1% | 42.6% | 38.9% |
| 50s | 84.9% | 54.4% | 32.2% | 33.1% |
| 60s | 72.6% | 36.1% | 21.2% | 22.6% |
| 70s | 48.5% | 15.8% | 11.1% | 12.5% |
Source: MIC Institute for Information and Communications Policy, FY2025 Survey on Usage Time of Information and Communications Media and Information Behavior (June 2026). Usage rates, ages 13–79.
Decide with both statistics and practice
Usage rates show how many people use a platform, which is a different question from ad performance. Here is how the du-o team reads it in practice; treat these as working guidelines, not statistics.
- Most audiences you want to reach with ads can be covered by three platforms: TikTok, Meta (Instagram and Facebook) and YouTube
- Instagram and TikTok work well for people up to their 40s, while YouTube tends to perform for products aimed at older audiences, such as people aged 55 and over
- TikTok's ad submission and review move smoothly, which also makes it a good place to run creative tests quickly
One thing to watch: ad review standards differ by platform. The same video may pass on one platform and be stopped on another. Google Ads (including YouTube) in particular has stricter policies and leaves less room in how you express things. Google explains that it uses a combination of AI and human evaluation to check that ads comply with its policies. Build a plan to adapt your creative to each platform's standards from the start. Detailed data by age and gender is in Japan Major SNS Adoption Trends 2026.
02💰 Concentrate or spread your budget?
A common approach when entering Japan is to put a little budget on each of TikTok, Instagram, YouTube and X to "see what works." It is easy to understand, but it tends to work against delivery optimization.
Exiting the learning phase takes volume
Each platform's delivery system keeps exploring how to deliver your ads until it has collected enough results data. This period is called the learning phase.
- Meta: ad sets exit the learning phase once they can deliver stably. This usually happens after about 50 results in the week after the ad set's last significant edit (Meta Business Help Center).
- TikTok: volatility starts to decline after about 25 campaign results or 7 days from when the campaign enters the learning phase. During learning, TikTok advises against pausing campaigns or ad groups, making edits that retrigger learning, and setting unreasonable budgets or creative volumes (TikTok Ads Manager Help).
In other words, the more finely you split your budget, the fewer results each unit gets, and the harder it is to finish learning. These guidelines and conditions can change, so check each platform's help pages for the latest before you launch.
How to run it in practice
| Stage | How to place budget | Goal |
|---|---|---|
| Launch | Concentrate on one or two platforms that fit your target | Finish learning and build up comparable data |
| Validation | Add variants of winning creative on the same platform | Confirm the winning pattern is not a fluke |
| Expansion | Add a third platform or another age group | Avoid plateauing on a single platform |
Changing budgets or placements frequently during learning tends to restart learning each time. Batch your changes and make them on a fixed day of the week so the data stays comparable.
03📐 Set up measurement before you spend
If you start delivery before measurement is in place, you cannot tell which video drove results, and the delivery system does not get the data it needs to learn. Complete the following before you start spending.
When results happen on the web
- Combine the browser pixel with a server-side connection: for optimal ad performance, Meta recommends implementing the Conversions API alongside the Meta Pixel. TikTok likewise recommends using its Pixel and Events API together with event deduplication for web.
- Set up deduplication: because the same result arrives from both the pixel and the server, you need a way to avoid counting it twice. For Meta, events are treated as identical when the pixel's event ID matches the Conversions API
event_idand the event names match. - Define one conversion: decide up front which event you optimize for (sign-up, purchase, lead form and so on), and align its name and conditions across platforms.
When results happen in an app
- Use an MMP (mobile measurement partner): a measurement tool for comparing installs and in-app actions across platforms on the same basis. Simply adding up the numbers in each platform's dashboard leads to discrepancies from duplication and differing definitions.
- Plan around ATT: if your app shares people's data with other companies to track them across apps and websites, it must request permission through App Tracking Transparency (Apple). Design on the assumption that a share of users will decline.
- Understand SKAdNetwork's conversion windows: SKAdNetwork supports up to three postbacks (notifications of results), with three conversion windows from first launch: days 0–2, days 3–7 and days 8–35. After a window closes, the postback is sent after a random delay (Apple). In other words, results do not arrive in real time. Factor this delay into when you judge whether a creative is working.
Because iOS has a high share in Japan, ATT and SKAdNetwork (including its successor AdAttributionKit) weigh heavily on measurement design. See Why Only Japan Has ~65% iOS Share for details.
Naming rules to set before launch
Just as important as tool setup is a naming convention for campaigns, ad sets and creatives. Putting the elements into the name, such as "platform_angle_hook_talent_length_version," makes it easy to tally later which hook worked.
04📉 Reading the signs of creative fatigue
When the same video keeps running, it is shown to the same people again and again, and response gradually drops. This is called creative fatigue. Fatigue is hard to judge from a single metric. Look at whether several signals are moving in the same direction.
| Signal | What it indicates | First move |
|---|---|---|
| Falling CTR (click-through rate) | People who see the ad are responding less | Add variants with a new opening hook |
| Rising CPM (cost per 1,000 impressions) | The same placements are costing more | Add new creative to give delivery more options |
| Rising frequency (impressions per person) | The same people are seeing the ad repeatedly | Broaden targeting or add a video with a different angle |
| Falling view rate in the first few seconds | People are swiping away in the opening seconds | Rebuild only the hook |
| Rising CPA or CPI | Efficiency on final results is declining | Isolate the cause together with the four signals above |
Things easily mistaken for fatigue
- Restarted learning: numbers fluctuate right after a significant edit. Check recent edit history before concluding it is fatigue
- Seasonality and competitors: CPM can rise during year-end sales, long holidays or a competitor's large campaign. If several creatives worsen at the same time, suspect external factors
- Measurement delay: for apps, SKAdNetwork postbacks arrive late. Do not judge from the last few days alone
Some platforms show an indicator in the dashboard suggesting possible fatigue, but only under limited conditions. Set your own criteria in advance rather than waiting for an indicator. Because the level at which to swap varies by product and budget, this article does not give specific numeric thresholds.
05🧩 Refreshing cheaply and quickly with modular production
If you go back to planning and reshooting every time fatigue appears, you will not keep up. This is where it helps to build videos as combinations of parts (modules).
| Module | Contents | How often to swap |
|---|---|---|
| Hook (opening) | The first few seconds: a question, a demo, a comparison and so on | Most often. Swap this first when fatigue appears |
| Body | The problem, the solution and the proof | Moderately. Swap it when you change the message angle |
| CTA (close) | The prompt to install or sign up | Least often. Swap it when testing wording |
For example, if you shoot five hook patterns, you get five variants while keeping the body and CTA the same. Because you can swap only the opening while keeping a winning body, you can refresh without adding shoots.
Preparing to run modular production
- During a shoot, capture alternate hooks and extra takes from the talent in one go
- Template the subtitles, music and CTA screens so a swap is all it takes to export
- Use the naming convention from the previous section so the module combination is clear from the name
- Agree on talent usage periods and permitted platforms on the assumption that variants will be made
Hook types, adapting message angles, structures by length and how to set test variables are covered in detail in Anatomy of Winning Vertical Video Ads in Japan.
06📊 The metric hierarchy: from attention to business outcome
Video ads come with many metrics. What matters is to separate which metric represents which stage. Strong numbers at an upper stage mean nothing if they do not lead to business results.
| Stage | Main metrics | What it tells you |
|---|---|---|
| ① Attention | View rate in the first few seconds, average watch time | Whether people stopped in the feed |
| ② Response | CTR, visits to the profile or landing page | Whether the content caught their interest |
| ③ Conversion | CVR, CPA, CPI | Whether it led to sign-ups or installs |
| ④ Business outcome | ROAS, retention, LTV (lifetime value) | Whether people kept using it and it drove revenue |
Reading by stage reveals the cause
- ① is low: a hook problem. Rebuild the opening
- ① is high but ② is low: the body's message does not fit. Change the angle
- ② is high but ③ is low: the ad and the landing page or store page do not match. Review the destination
- ③ is high but ④ is low: you are attracting the wrong people. Review the conversion definition or targeting
Make final decisions on the business-side metrics in ③ and ④. Use ① and ② as diagnostics to understand why you got a result. For typical acquisition costs in Japan, see Japan App CPI Benchmarks 2026.
07🗓️ A weekly operating rhythm (example)
To keep operations from depending on one person, it helps to fix the flow of the week. Below is one example of how the du-o team works; adjust it to your team and budget.
| Day | What to do |
|---|---|
| Mon | Review last week's numbers through the metric hierarchy. List videos showing signs of fatigue |
| Tue | Decide which module to swap (hook, body or CTA). Prepare the script and on-screen text |
| Wed | Edit and export the variants. Check each platform's review standards |
| Thu | Upload the new creative. Batch any budget or placement changes on this day too |
| Fri | Check review status and early performance of the uploaded videos. Do not draw conclusions yet |
A monthly review
- Summarize what winning patterns have in common (hook type, talent, message angle) and feed it into the next shoot's plan
- Revisit budget allocation by platform and decide which to scale up and which to cut back
- For apps, include late-arriving SKAdNetwork results and retention in the review
Keeping this rhythm assumes you have new creative ready to swap in every week. If producing it in-house is hard to sustain, working with a production partner is an option. Also, include a disclosure such as "広告" (Ad) or "PR" in the videos you run so viewers can tell they are ads. For details, check the Consumer Affairs Agency's materials or consult a professional.
08✅ Operations checklist for video ads in Japan
- Defined target age and gender first, then narrowed to one or two platforms
- Checked each platform's review standards and planned to adapt creative accordingly
- Each ad set or ad group has enough budget to gather the results needed to exit learning
- For web, combined the pixel with a server-side connection and set up deduplication
- For apps, installed an MMP and factored ATT and SKAdNetwork reporting delays into decision timing
- Set a naming convention for campaigns and creatives
- Defined fatigue signals (CTR, CPM, frequency and so on) and the review procedure
- Built hook, body and CTA as modules and have swap-ready footage in stock
- Agreed as a team that final decisions are made on business-side metrics such as CPA, CPI and ROAS
🚀 Plan your video ad operations for Japan side by side with our hands-on Duo team
2,000+ creatives produced per month for TikTok, Instagram and YouTube, from scripts, talent, shooting and subtitles to running the ads. Rough, early-stage questions are welcome!
Book a free consult →❓ FAQ
Which platform should we start with for video ads in Japan?
Is it wrong to split budget across several platforms from the start?
When should we swap out creative?
Are there Japan-specific points to watch in app ad measurement?
Isn't it hard to prepare new creative every week?
📚 Related articles
📖 Sources
- CyberAgent, "2025 Domestic Video Advertising Market Survey" (March 9, 2026) — vertical video ad market ¥204.9 billion (155.9% of the prior year)
www.cyberagent.co.jp/news/detail/id=33050 - MIC Institute for Information and Communications Policy, "FY2025 Survey on Usage Time of Information and Communications Media and Information Behavior" (June 26, 2026) — overall and by-age usage of major social platforms (ages 13–79)
www.soumu.go.jp/menu_news/s-news/01iicp01_02000130.html - TikTok Japan, "Socio-Economic Impact Report" (June 2026) — about 49.5 million monthly active users in Japan (end of May 2026, TikTok's own figure excluding overlap with TikTok Lite)
newsroom.tiktok.com/tiktok-socio-economic-impact-report-june-2026?lang=ja-JP - Meta Business Help Center, "About the learning phase" — ad sets usually exit learning after about 50 results in the week after the last significant edit
www.facebook.com/business/help/112167992830700 - TikTok Ads Manager Help, "About Learning Phase" — volatility starts to decline after about 25 results or 7 days; operations to avoid during learning
ads.tiktok.com/help/article/learning-phase - Meta for Developers, "Handling Duplicate Pixel and Conversions API Events" — recommends the Conversions API alongside the Pixel; deduplication by event_id and event name
developers.facebook.com/docs/marketing-api/conversions-api/deduplicate-pixel-and-server-events - TikTok For Business Help, "About Events API" — recommends using the Pixel and Events API together with event deduplication for web
ads.tiktok.com/help/article/events-api - Apple Developer, "App Tracking Transparency" — framework for requesting permission when sharing data with other companies to track people
developer.apple.com/documentation/apptrackingtransparency - Apple Developer, "Receiving postbacks in multiple conversion windows" — SKAdNetwork's three conversion windows (days 0–2, 3–7, 8–35) and random delay
developer.apple.com/documentation/storekit/receiving-postbacks-in-multiple-conversion-windows - Google Advertising Policies Help, "Google Ads policies" — ads are reviewed for compliance using a combination of AI and human evaluation
support.google.com/adspolicy/answer/6008942