Unpredictable Hits #020: Why Top Advertisers Made 3.4x More Creatives but Not More Playables + 4 Winners in 13,204 Ads
What's up friend?
Last issue I wrote about the slow first million: most of this year's puzzle winners needed three months to get there, and a 30-day verdict would have killed them. This issue is the same story one level down, at the ad.
We read five months of Meowdoku's ad machine from the outside: 13,204 videos, 210 playables and 58 million downloads. Four videos ever carried a real share of a month. And the two biggest ads of September were sitting at 1% in August, in the pile most of us would have killed.
That's the theme this time: winners are slow, and many of them look like losers first. I said it on stage at our event in Barcelona, and our own analysis of AppMagic data agrees: of 447 games that launched at least ten playables in January, 43% saw their top playable peak in month four or later.
So the question for a UA team is how many bets it can afford to keep alive until the late ones turn. That's a production question, and this week the industry answered it three different ways: the UK's ad regulator closed the cheap shortcut, Playtika chose margin over scale, and the biggest advertisers tripled their creative output while playables stayed flat.
This edition:
โข Four in 13,204: how Meowdoku finds the ads that carry 58 million downloads, and why month one tells you almost nothing.
โข Two new UATOPIA sessions: Supercell's Peter McLaughlin, Turborilla's John Wright and PocketHaven's Amit Klein on where the edge lives when everyone has the same AI, and Supersonic's Samantha Benjamin, Zynga's Zeynep Tasoz Cataner and Madbox's Thibault Quillard on running a UA team when agents do the tasks.
โข The UK's ad regulator bans ads from King and Scopely, and says the "not actual gameplay" label doesn't save you. We updated our fake-ads piece with the line it drew.
โข The 49 biggest advertisers now launch 3.4 times as many creatives as a year ago. Playables didn't move.
โข Playtika ends its strategic review and stays independent, with record profit on flat revenue.
โข Rotatelab's Cube Land: ten years, twenty games, then $5.5M a month from a city more than a thousand kilometres from Istanbul.
Yalla.
๐ฅ Sett stuff and analyses
Four in 13,204: how Meowdoku finds the ads that carry 58 million downloads
Meowdoku is the cat sudoku that went from nothing to 57.8 million downloads by the end of October, 41.9 million of them in the five months we studied. Behind those downloads, as far as Sensor Tower could see: 13,204 video ads and 210 playables. We read the whole library from the outside, month by month.
The numbers that matter:
- Out of 13,204 videos, four in five months ever took a tenth of a month's video impressions.
- Of 210 playables, eight ever reached 10% of a month's playable impressions.
- Six of the eight playable winners reached 10% after the month we first saw them. Two of the four video winners did the same.
- In September, nine out of ten video impressions went to ads that were already running. 4,398 new videos arrived that month and got almost none of the exposure.
- The typical video that reached 10% held it for one month.
September's two biggest videos were at 1.0% and 0.9% of impressions in August. As the piece puts it: "If you kill a batch on its first month, and most of us do, you kill both of them." Meowdoku's team didn't pick its September winners in May. They kept the pile alive long enough for two slow starters to turn.
Why it matters for you: at this volume, a creative's first month is a weak signal, and the ads that carry a game are usually found late. That changes what "testing" means. The team that can afford to keep 4,000 videos in the air while the data settles is not running the same experiment as the team that can afford 40.
UATOPIA session: Where the edge lives when everyone has the same AI
Every studio has the same AI tools, so where's the edge? Supercell's Peter McLaughlin, Turborilla's John Wright and PocketHaven's Amit Klein dug into it at our event in Barcelona.
Their first answer is creativity, because "we've got less money, less resources, less time." The second is experience: making a game is easier, and finding one that scales and holds players at day 90 is the hard part. As Amit put it, "today, it's super easy to create a new game, but creating a fun game is even harder, and creating the right game is even harder."
The third is knowing your players beyond your game, because, in Peter's words, "we're also competing with social networks for attention." So true, I always forget that part.
UATOPIA session: Running a UA team when agents do the tasks
Agents can run a lot of the work now, so what's left for the UA team? Supersonic's Samantha Benjamin, Zynga's Zeynep Tasoz Cataner and Madbox's Thibault Quillard took that on in Barcelona.
Supersonic lets agents optimize its smaller games every week, but at launch the call stays human, because, in Samantha's words, the AI "doesn't know how to do strategic thinking or take risks." Zeynep's version: "Disruption comes from the human edge," and AI is "a tool. And I am using it. It's not using me." Thibault hires for curiosity first, then builds systems around the team, what he calls an "Iron Man suit for the creative team."
๐ฅ Industry stuff you should know to be better
1. The UK bans King's and Scopely's ads, and the "not actual gameplay" label doesn't help
Source: the UK ad regulator's ruling of Nov 4, with Isa Muhammad's read on PocketGamer.biz
The UK's Advertising Standards Authority (ASA) banned mobile game ads from four publishers for showing gameplay their games don't have.
The cases:
- King, Farm Heroes Saga: a Facebook ad where clearing match-three icons dropped nails and sent a saw toward a cow. The matching was real; the nails, hatch, cow and saw were not in the game. The ad carried "Not actual gameplay". The ASA said the label was "likely to contradict, rather than clarify, how consumers would understand the game". King's second ad, a promotional sunrise and then the real grid, was cleared.
- Scopely, Pixel Flow (via Loom Games): puzzle elements the game doesn't contain.
- Omnilojo, Dark War Survival: three Google ads that presented cinematics as play, plus "2026 Game Creative" badges that read like awards.
- Vistrex, Idle Zombie Miner: a Facebook ad and a Play Store listing that sold a 60-second minigame as the game.
Why it matters: the ASA banned Playrix's pin-puzzle ads for the same reason in 2020. What's new is the caption: this ruling says "Not actual gameplay" doesn't fix it. It's UK-only and carries no fine, but King and Scopely by name in a published ruling is a memo to every studio's legal team. The ad that kept the real mechanic was cleared. That's the line.
If you want more on why fake ads work and where this ruling puts the line, we updated our piece on them this week.
๐ Read our updated fake-ads piece
2. The 49 biggest advertisers tripled their creative output. Playables didn't move.
Author: Jeff Cohen, on Gamigion, with Sensor Tower data
Sensor Tower tracked the 49 mobile games with the biggest ad spend. The median advertiser launched 135 new creatives a month in Q1 2025 and 465 in Q1 2026, a 3.4x jump, and has held near 400 a month since.
The numbers that matter:
- Video is 70 to 75% of new creatives every month. Images went from 17 to 60 a month. Playables stayed flat.
- Spend across the cohort rose about 1.7x over the same period, by Cohen's count, so the dollars behind each new creative roughly halved.
- Moloco's data, cited by Cohen: breakout launches had 20 videos carrying meaningful spend against 5 for everyone else, but only 6 playables against 3.
Cohen's explanation: AI made video nearly free to produce and hasn't done the same for playables, which need an HTML5 build, network-specific file limits and testing across hundreds of devices. The platforms demanding volume, Meta's Andromeda and TikTok's automated campaigns, run on video and static. His conclusion, and ours: the first teams to produce playables at close to video speed will have something competitors can't copy with a prompt.
3. Playtika stays independent, with record profit on flat revenue
Author: Vikki Blake, MobileGamer.biz, with Playtika's Q3 release
Seven months after it opened a strategic review with Morgan Stanley as adviser, Playtika's board ended it without a deal and decided to stay independent. Its line: staying independent "best positions the company to create long-term value for shareholders".
The numbers that matter:
- Q3 revenue $673.8M, flat year on year. Adjusted EBITDA a record $252.1M, up 15.9%, at a 37.4% margin.
- Sales and marketing came down from $206.3M to $167.8M year on year, a step-down the company had planned.
- Direct-to-consumer revenue rose 31.4% to $275.1M, about 41% of the total. Disney Solitaire made $111.2M, up 88.4% year on year. Revenue per daily user rose from $0.89 to $1.03.
The read: Playtika is earning more from each player and keeping more of each dollar, by moving revenue to its own stores and spending less on marketing. For UA teams, it's a live case of a big portfolio choosing margin over scale, and the next few quarters will show where that leads.
4. Cube Land: ten years, twenty games, then $5.5M a month
Author: Aylin Yazฤฑcฤฑ, Deconstructor of Fun
Rotatelab is a gaming studio from Gaziantep, more than a thousand kilometres from Istanbul, where most of Tรผrkiye's game industry sits. The founders have made games together since 2016, spent about two and a half years making hyper-casual titles with Rollic, launched around twenty games with 17 million installs between them, then spent a year and a half learning to self-publish before putting the whole studio behind Cube Land once its retention test came back.
It took four months to go from idea to a validated game, and the global launch came on July 2.
The numbers that matter:
- US top 50 free and top 80 grossing within weeks of launch. In the last 30 days: 1.3M downloads and $5.2M in in-app revenue.
- October: $5.5M from purchases and about $980K from ads, per Sensor Tower. 308K daily players, 46.7 minutes a day each.
- About 200 to 250 creatives in a 30-day window while scaling.
- A seed round in August, co-led by Laton Ventures and Arcadia Gaming Partners, amount undisclosed, ten years after the founders started.
The mechanic: Cube Land takes the This-is-Blast family and turns the picture into a 3D voxel object the player rotates, so every move is two decisions: which shooter to fire and which side of the object to face. Aylin's point, and the one that matches this issue's theme: "I see hit-making in this genre as a numbers game", and the studios that get there kept taking shots long after most would have stopped.
๐ Read on Deconstructor of Fun
Quick hits
- Silver Lake and Saudi Arabia's PIF have approached Unity about a takeover at around $25B, per a Reuters report. The same pair, with Affinity Partners, took EA private for $55B in August. The approach came the same week Unity posted its first GAAP profit since its IPO, on $622M of revenue, with its Vector ad platform up 126% year on year. If it goes private, expect Unity Ads to be run for cash. PocketGamer.biz
- AppLovin's Q3: revenue $2.18B, up 55%, at an 84% adjusted EBITDA margin, and a Q4 guide of up to about $2.4B. Esra Senekci's question for studios on Gamigion is the right one: is your ROAS improving as fast as AppLovin's revenue? Gamigion
- Hutch co-founder Shaun Rutland raised a $5M pre-seed for his new studio, Anorak, co-led by Play Ventures and Backed VC, with Supercell and MTG in. It builds generative video that responds to the player's inputs in real time, with its first playable formats due in early 2027. MobileGamer.biz
- Century Games now runs AI creatives across its portfolio, from Whiteout to King Shot, Lord Rush and Top Force, and Lord Rush alone runs roughly 5,000 ads a month, per Matej Lancaric. His read: the only thing that matters is the transition into gameplay. Brutally Honest
Wrap up
One thing you can try this week: pull your creative report, take your five best ads right now, and look at their first month. If any of them would have failed your own kill rule, that rule is killing late winners, and it's time to change it.
Till next time,
Jonathan (more of a dog-person) Fishman
About the Author
Fishi is the Head of Marketing at Sett. His brain is a chaotic jukebox of ideas with more cultural references than any feed can handle. He collects sneakers and plays chess while youโre still counting sheep.
