A top-10 mobile RPG studio came to us six weeks before their planned global launch. They had a ~$1.2M influencer budget earmarked for a coordinated push across US, JP, KR, DE, FR, BR and SEA. The strategic question wasn't "can creators move installs" — they already knew that from their previous title. The question was: does this specific game retain in regions where the meta-genre is saturated?

Their internal UA team had a D7 retention benchmark of 18% from soft-launch CPI campaigns on Meta and Google. Creator-sourced installs historically retained 1.4-1.7x better, but only when the creator's audience actually overlapped with the core fantasy-RPG demographic. The risk: spending $1M+ on a global creator push only to discover the game's combat loop doesn't click with a Korean audience the way it does with a Japanese one.

We pitched Tier-2 validation. Hong Kong, Taiwan, Korea — three markets with mature RPG audiences, lower creator rates than JP/US, and enough behavioral signal to extrapolate. Total ask: $35K, 6 weeks, full attribution back to D60.

What we forecast

Our forecast engine pulled comparable placements from 47 prior mobile RPG campaigns across APAC (2023-2025 vintage). For this game's combat-style, art direction, and monetization model, the closest reference clusters were two soft-launches in TW and one HK campaign from late 2024.

Predicted CPA range: $3.10 – $3.65 at 80% confidence, with $3.40 as the central estimate. Predicted view volume: 820K – 1.05M across 6 weeks at the proposed creator mix.

Why YouTube-heavy (roughly 70% of spend): the game's session length skews to 12-18 minutes, and the monetization hinges on players understanding the gear-progression loop within their first three sessions. That's a "show me how it actually plays" pitch. Short-form can drive curiosity, but the install-to-D7 retention math falls apart if the creative doesn't communicate the loop. We've seen this pattern hold across 200+ RPG placements: long-form YT delivers CPI 20-30% higher than TikTok, but D7 retention 2.1x stronger. Net D7-weighted CPA is lower.

TikTok complement (the remaining 30%): not for installs, for top-of-funnel awareness in HK specifically, where TikTok penetration in the 18-34 male gaming demo is highest of the three markets. We modeled TikTok as a discovery layer feeding YT search lift — and budgeted accordingly.

Creator selection logic: filter for channels with >40% audience overlap on RPG/gacha content in the last 90 days, drop anyone whose last 5 sponsored placements underperformed their organic median by more than 35%, weight toward channels with prior soft-launch coverage (they handle NDA flows without us hand-holding).

What we did

11 creators across HK/KR/TW, ranging from 180K to 1.6M subscribers. Breakdown:

• 4 KR creators (2x YT mid-tier 400K-800K, 1x YT macro 1.6M, 1x TikTok 350K)
• 4 TW creators (3x YT mid-tier 220K-650K, 1x TikTok 290K)
• 3 HK creators (1x YT macro 1.1M, 1x YT mid 180K, 1x TikTok 410K)

Formats: dedicated 8-14 minute gameplay reviews on YT, with a hard requirement that the creator hit three beats — first combat encounter, first gear upgrade decision, first PvP queue. We wrote the brief, the creators wrote the script. No reading from a sheet, no forced punchlines. TikTok was looser: 30-60s, creator's choice of format, only constraint was a visible install CTA in the last 5 seconds.

Timeline: 2 weeks of creator outreach, contracting, NDA flow with the studio. 3 weeks of staggered publishing (we didn't drop all 11 in the same week — we cohorted by market so weekly cadence wouldn't cannibalize search). Final week was tail-tracking and D30/D60 attribution close.

Attribution: AppsFlyer with creator-specific OneLink slugs, reconciled against the studio's internal MMP at D30 and D60.

What happened

Final numbers at D60 close:

• Spend: $35,000
• Views: 907,400 (forecast: 820K–1.05M — landed inside the band)
• Conversions (installs attributed): 1,861
• Blended CPA: $3.12 (forecast central: $3.40 — beat forecast by 8.2%, beat upper band by 14.6%)

Where it got interesting was the by-market breakdown:

• TW: $2.78 CPA — best of the three. The 3x mid-tier YT creators (220K-650K) carried this market. Their CPAs were $2.10, $2.40, and $3.05. The TikTok creator delivered $4.80 CPA but drove a measurable 31% lift in branded YT search the week after posting.
• KR: $3.04 CPA — the 1.6M macro underperformed at $4.20 CPA, while the two mid-tier YT channels hit $2.60 and $2.85. Classic pattern: macro audience was too broad, mid-tier audiences were RPG-natives.
• HK: $3.71 CPA — the weakest market on CPA, but D14 retention from HK installs was 23.4%, vs 19.1% TW and 17.8% KR. Higher CPA, higher quality user.

D7 retention across all creator-sourced installs: 22.6%, vs the studio's paid-social D7 benchmark of 18%. That's the number that mattered.

The studio used this data to do two things: (1) green-light the global launch but reallocate ~$340K from JP to TW/HK based on retention-adjusted CPA modeling, and (2) cut the KR macro tier entirely from the global push and double down on the 400K-800K KR YT band.

The lesson

Tier-2 markets are underused as validation sandboxes because most agencies sell them as standalone campaigns rather than as pre-flight diagnostics. The wrong question is "what CPA can I get in TW?" The right question is "what does TW performance tell me about JP and KR creative-market fit before I commit a million dollars?"

For RPG and other retention-driven verticals, the second question saves you roughly half the global launch budget — not because Tier-2 is cheaper (it is, but that's a side effect), but because it tells you which creator tier, which format length, and which creative beats actually translate into D7+ retention before you scale. In this case, killing the KR macro tier alone saved ~$180K of waste that would have been invisible until D30 post-launch.

What it means for your campaign

If you're staring at a global rollout budget and your forecast confidence interval is wider than ±25%, you're not ready to deploy. A $30-50K Tier-2 validation buys you a confidence interval of ±10-12% on the markets that actually carry the launch — and tells you which creators to drop before they cost you $20K each at macro rates.

When we forecast your CPA, we're not pulling from a generic benchmark. We're pulling from the placements that look most like yours — same vertical, comparable monetization, similar creative format. If the forecast is wrong, we want to know why before you spend, not after.