MGS 2026 Recap: How Actionfit Scaled a 10B Won Mobile Game
Has a new game's test come back lukewarm, and you've spent weeks unable to decide whether to push it or kill it? Every week of indecision eats into the time and budget for your next attempt, while another game grabs the opportunity.
In game marketing, CPI (cost per install) is the ad spend needed to win one app install. This post recaps "The Story Behind a 10 Billion Won Game," a session by Hyegang Seo and Hyerin Choi (Actionfit and AB180) at Modern Growth Stack (MGS) Week 2026, held July 21 to 24, 2026 at the Westin Seoul Parnas. The daglo marketing team attended to hear how Actionfit built and scaled its puzzle game Ppajinyang.
We transcribed the session with daglo on site, then reread the transcript to pull out the numbers and decision rules.
How Ppajinyang hit 10 billion won in four months
Ppajinyang is a hybrid-casual puzzle game Actionfit built in its second year of in-house development. It passed 10 billion won in cumulative revenue just four months after launch, and it started with a marketer spotting a benchmark game.
- 6 million downloads in its first year, 70% of revenue from in-app purchases, and over 90% from outside Korea
- A prototype built in one to two weeks for a market test, then three to four more weeks of development to scale
- Just three people at the start: a marketer, an artist, and a developer
Actionfit, founded in 2020, is a mobile casual game studio of about 50 people whose revenue has grown 50% a year on average. Most studios start a game with a game designer. At Actionfit, marketers, who sit closer to the market, propose games and act as PMs, because they're more likely to decide based on metrics and revenue.
One rule, a 1 dollar CPI, to build fast and kill fast
Actionfit tests on Meta, Android, and the US market. If a prototype can't reach a 1 dollar CPI, it's dropped. If it does, the team builds more and moves on to a ROAS test.
In casual games, being one or two weeks late can decide who owns a market, so Actionfit aims to follow trends quickly and take second place. Today one developer builds two game cores a week, and the goal this year is 100 to 200 new titles. Choosing what to build comes down to four questions.
- Do we have the skills to build it?
- Can we build it fast?
- Is the benchmark game still trending up?
- Is there no competing game yet?
Ppajinyang checked every box. Its benchmark game was crossing 10,000 dollars in daily revenue, the puzzle format was easy to build, and no competitor had entered.
Running creatives: impressions come first
For Actionfit, the most important marketing metric is impressions. Strong ROAS and retention mean nothing if an ad never gets served, so creatives with high impressions are kept running even when their ROAS is low.
| Item | How Actionfit runs it |
|---|---|
| Creative team | 4 people on video, 1.5 on playables |
| New creatives per week | About 30 video concepts, 2 playables |
| Video templates | About 250, turning one concept into 250 variations |
| Refresh cycle | Every 2 to 3 weeks, weekly on high-spend channels |
To find concepts the algorithm favors, the team remixes its own top-impression concept or checks competitors' impression rankings in analytics tools. To lift click-through rates, it emphasizes the call-to-action button, for example by showing gameplay for only the first 20 seconds of a 60 or 120 second video and an end card for the rest. Country differences showed up in format rather than concept: on DSPs, native images worked best in Japan and playables in the US.
When to scale: can you earn the money back?
Scaling depends on evidence that spend can be recovered. When Ppajinyang scaled, its LTV was on average 160% higher than CPI, and its ROAS kept climbing at a steady rate.
- LTV vs. CPI: With no retention data early on, the team uses ad platforms' data on competing titles and finds similar retention curves among its own past games.
- ROAS growth: Ppajinyang held roughly 35% growth from D1 to D7 and 25% from D7 to D14.
- Higher KPIs: After a title with similar early metrics later failed, the team raised the bar to absorb outside variables, targeting cohort ROAS of 50% by D7 and 100% by D30.
Early in scaling, countries and creatives run together so platforms can learn fast. Later, campaigns split by country: in-app purchase campaigns for the US and Japan, and ad-revenue ROAS campaigns for tier-3 markets such as Brazil.
Key takeaways: success is mostly luck, so take more shots
Actionfit tried to reproduce Ppajinyang's success factors in new games, but most attempts failed. The team concluded that a game's success is close to luck, so it chose clear rules and more attempts over searching for a formula.
Over the past year it built 46 MVPs, and fewer than 0.5% hit early KPIs, yet the market still produces a hit every two to three months. If you work in game or app marketing, these questions are worth asking.
- Is your rule for killing a new attempt a single, clear number?
- Where will you get benchmark data before you have your own?
- Is your creative bottleneck ideas, or the production system?
- Are you testing creative formats by country or segment?
- Are you switching off high-impression creatives just because ROAS looks low?
When a session has two speakers or a Q&A, knowing who said what matters when you review the recording. daglo's Speaker ID remembers voices you've named once and fills in real names instead of "Speaker 1" from the next recording on. daglo is the first service in Korea to offer this to every user. Learn more about Speaker ID
For the other sessions we covered, see our full MGS 2026 recap.
#MGS2026 #ModernGrowthStack #GameMarketing #MobileGames #UserAcquisition #CPI #Actionfit #daglo
Where talk becomes work.
Get started with 1,000 credits a month — no credit card required.