All Work
Experimental Build · A-03 · v0.5
Arcade Run
Make the ad feel like the first ten seconds of the game. Card Fanv0.5 / LAB
HOOK 01FAIL9:16
HOOK 02CHOICE9:16
HOOK 03UPGRADE9:16
HOOK 04WIN9:16
THE PROBLEM WE ADDRESSHow many distinct hooks can one gameplay loop sustain?
01Modelled business scenarioCompany and figures are fictional
02Working mechanismCard Fan
03Modelled outcomeFor commercial orientation
MODELLED BUSINESS SCENARIO
Cheaper installs that still play on day seven.
BUDGET$180,000 media / monthPERIOD8-week model
Modelled across Meta, TikTok and AppLovin with 40 video variants.
BUSINESS MODEL
What is sold and where the problem starts.
Scale installs without lowering day-seven retention or predicted player value.
CPI looked acceptable, but creative winners attracted low-retention users and country budgets followed install cost only.
Keep blended CPI below $1.45, raise D7 ROAS above 24% and find ten scalable hooks.
THE EXPERIMENT STORY
The cheapest install became the most expensive player.
This narrative uses a modelled scenario. The company, budget and outcomes are fictional; the diagnostic and operating logic show how a real engagement would unfold.- 01
01 / THE BRIEF The first symptom was not inside the ad account.
One spectacular fail-video won the auction and drove CPI down. It also attracted people who left after the first session. The dashboard rewarded the ad for opening the store, while the game paid for players who never reached the core loop.
Large mobile game publisher was selling free-to-play runner game across United States, Japan, Germany and Brazil. The model was in-app purchases and ad revenue. On the surface, the brief sounded direct: Scale installs without lowering day-seven retention or predicted player value.
- MODEL
- In-app purchases and ad revenue
- 02
02 / WEEK ONE Clickbait installs
CPI looked acceptable, but creative winners attracted low-retention users and country budgets followed install cost only. The trap was that isolated metrics still looked acceptable: Video CTR was 0.74%, click cpc was $0.42, and install cpa / cpi was $1.84.
Some hooks lowered CPI but brought players who left in the first session. That changed the diagnosis. The issue was not one toggle inside an ad account; it was what the system counted as success and which signal reached the team. The work had to connect media to the outcome that actually created business value.
- Install CPA / CPI
- $1.84 → <$1.45
- 03
03 / THE TURN Judge every hook by the cohort it creates, not by the install it wins.
The first decision was “Value cohorts.” Join campaign, creative, retention, IAP and ad-revenue cohorts. It was followed by “Creative taxonomy”: Tag every video by opening, mechanic, fail, reward, pace and CTA.
Ranked creative by D7 value per thousand impressions, not CTR or CPI alone. Once the measurement logic was shared, the team could see what to stop, what needed a different message, and where the real constraint sat in the page, lead handling or product availability rather than traffic.
- FIRST DECISION
- Value cohorts
- 04
04 / THE BUILD Impossible escape: lead with one instantly readable failure, then reveal the mastery loop.
Impossible escape: lead with one instantly readable failure, then reveal the mastery loop. The production plan became 10 concepts, 40 master videos, 120 hook mutations, 4 store-page sets and localized captions. The tests were not about abstract taste; they covered fail versus win, real gameplay versus staged scene, hand cam, score pressure, character skin and reward reveal.
The journey was assembled as one chain: Hooked video → Store page → First session → D7 player value. Meta, TikTok, AppLovin, Unity Ads, Google App Campaigns no longer lived in separate reports, while AppsFlyer, Firebase, BigQuery, App Store Connect, Google Play Console connected spend, behaviour and the final outcome. Built a pLTV proxy from tutorial completion, second session and day-one ad revenue.
- SYSTEM
- 5 channels · 5 tools
- 05
05 / THE DECISION The ad should feel like the first ten seconds of the real game — and attract the person who wants the next ten minutes.
In the model, video ctr moved from 0.74% to 1.36%, install cpa / cpi from $1.84 to $1.29, and d7 roas reached 27%. These are not promises; they make the decision thresholds explicit before real data enters the system.
Weeks 7–8: Scale by predicted player value. The main conclusion reached beyond a single campaign. The cheapest install was not the best acquisition. The winning creative bought enough attention and correctly previewed the game. That is the logic a real engagement would retain, using the client's actual evidence, constraints and accountable decision makers.
- D7 ROAS
- 27%
KPI / CTR / CPC / CPA
Not one vanity percentage. The full metric chain.
Every figure in this section is modelled to demonstrate the scenario economics.01Video CTR
- BEFORE
- 0.74%
- TARGET
- 1.20%
- MODEL
- 1.36%
Hook strength
02Click CPC
- BEFORE
- $0.42
- TARGET
- <$0.30
- MODEL
- $0.26
Traffic cost
03Install CPA / CPI
- BEFORE
- $1.84
- TARGET
- <$1.45
- MODEL
- $1.29
Acquisition efficiency
04D7 retention
- BEFORE
- 12.8%
- TARGET
- >16%
- MODEL
- 17.4%
Player quality
05D7 ROAS
- BEFORE
- 18%
- TARGET
- >24%
- MODEL
- 27%
Early monetization
STRATEGY
Four decisions that change the economics.
- 01Value cohorts
Join campaign, creative, retention, IAP and ad-revenue cohorts.
- 02Creative taxonomy
Tag every video by opening, mechanic, fail, reward, pace and CTA.
- 03Country economics
Set CPI and pLTV limits per operating system and market.
- 04Automated rotation
Pause fatigued variants and replace them with mutations of proven hooks.
CREATIVE IDEA
Impossible escape: lead with one instantly readable failure, then reveal the mastery loop.
10 concepts, 40 master videos, 120 hook mutations, 4 store-page sets and localized captions.
Fail versus win, real gameplay versus staged scene, hand cam, score pressure, character skin and reward reveal.
PATH TO OUTCOME
- 01Hooked video
- 02Store page
- 03First session
- 04D7 player value
WHAT DID NOT GO TO PLAN
Problems are part of the work. So are the fixes.
Clickbait installs
PROBLEM
Some hooks lowered CPI but brought players who left in the first session.
FIX
Ranked creative by D7 value per thousand impressions, not CTR or CPI alone.
Signal delay
PROBLEM
D7 data arrived too late for daily budget control.
FIX
Built a pLTV proxy from tutorial completion, second session and day-one ad revenue.
OPERATING CADENCE
What happens while the project runs.
- 01Week 1
Validate attribution and value events.
- 02Weeks 2–3
Launch ten creative territories.
- 03Weeks 4–6
Mutate hooks and store pages.
- 04Weeks 7–8
Scale by predicted player value.
MAIN TAKEAWAYThe cheapest install was not the best acquisition. The winning creative bought enough attention and correctly previewed the game.
This demonstrates how we think — it does not fabricate a case study.
Discuss a Similar Solution This is not a client result; the numbers are sample data. For your project, we adapt the solution, tools and measurement to the task and budget.