Supercell
ML personalization of in-game offers (which card or offer to serve)
Clash Royale has served card offers personalized by machine learning since 2018 based on the player's inventory, progression, and purchase frequency, with a small data science team automating recurring monetization; in 2025, Supercell described extending AI to acquisition and monetization.
Objective
Increase conversion and LTV by serving each player the most relevant card offer, with a small team that automates recurring monetization rather than growing it.
The deployment
Since 2018, Clash Royale has served personalized card offers in its store through machine learning. The system looks at the number of cards already owned, how close a player is to an upgrade tier, and their purchase frequency, among other factors, to decide which offers to show. The price and bundle size remain hardcoded; the ML chooses the cards the player is most likely to buy. The logic the team embraces is to stay small and automate everything that recurs on a regular basis. In 2025, Supercell described a broader AI plan presented by its AI Lead, with mass personalization for acquisition, prediction of player behavior, and adjustment of retention and monetization activities to maximize LTV, along with pilots of UA asset creation. The group also runs in production an AI support agent built with OpenAI in Brawl Stars.
Results Proof C
Recognized specialized press reporting by name the interventions of Supercell's data scientists and AI Lead on monetization and UA, plus a quantified vendor case study for the support agent. No financial figure on monetization itself, hence C.
How it works
Documented architectureThe stack in detail
- outil Modele de propension in-house ML that chooses which cards to propose based on the cards owned, the proximity of an upgrade tier, and purchase frequency; price and bundle size remain hardcoded
- infra Pipeline de features joueur player state (inventory, progression, purchases) kept up to date to serve the store offer
- llm OpenAI models used for the AI support agent in production in Brawl Stars, a system distinct from ML monetization
- integrateur Tomoro.ai partner in building the Brawl Stars AI support agent
How it runs, concretely
For ops teams-
1Reading the player's state AI (data science)
The system measures inventory, progression, and purchase history.
-
2Offer selection AI (data science)
The ML chooses which cards to propose; the price and bundle size are hardcoded.
-
3Display in the store game (Clash Royale)
The personalized offer appears in the game store.
-
4Measurement loop data team
Conversion and purchases fed back to refine the propensities.
The player's state: cards owned, proximity of an upgrade tier, purchase frequency. Without this up-to-date signal, the served offer loses its relevance and conversion drops.
How your customers perceive this type of use
Sourced studiesLe paradoxe est documente des deux cotes : 71% des consommateurs attendent des interactions personnalisees et 76% sont frustres quand elles manquent (McKinsey, 2021), mais 75% declarent ne pas acheter aupres d'organisations auxquelles ils ne confient pas leurs donnees (Cisco, 2024). La « creepy line » est localisee : messages recus quelques secondes apres une recherche et suivi de localisation sont les pratiques qui mettent le plus mal a l'aise (Periscope by McKinsey, 2019).
Acceptance conditions
- La confiance dans le traitement des donnees precede l'achat : 75% ne achetent pas sans elle (Cisco 2024)
- Un cadre legal protecteur rassure : 59% des consommateurs disent que des lois fortes sur la vie privee les rendent plus a l'aise pour partager des informations dans des applications IA (Cisco 2024)
- La personnalisation elle-meme est attendue quand elle est consentie : environ la moitie des consommateurs (US 55%, UK 52%) disent s'inscrire souvent ou parfois a des services personnalises (Periscope by McKinsey 2019)
Red lines
- Le message declenche quelques secondes apres une recherche ou un achat : deuxieme ou troisieme cause de malaise selon les pays (Periscope by McKinsey 2019)
- Le suivi de localisation percu comme de la surveillance : 40% de malaise en Allemagne et au Royaume-Uni (Periscope by McKinsey 2019)
- Le mesusage des donnees personnelles par l'IA, devenu la premiere inquietude des consommateurs, a 53% et en hausse (Qualtrics 2025)
Sources: McKinsey & Company 2021 · Periscope by McKinsey 2019 · Cisco 2024 · Qualtrics 2025
How to replicate
Inference, not sourcedData prerequisites
- inventory and progression per player
- purchase history and frequency
- offer catalog
Org prerequisites
- small autonomous data science team
- A/B testing process for offers
- governance over offer personalization
Possible stack
- custom propensity model
- player feature pipeline
- in-game offer system
The plan, step by step
- Step 1Map the player's state: inventory, progression, and purchase history consolidated per playerDeliverable: Player feature pipeline kept up to date continuously
- Step 2Define the offer space: what the ML chooses (content) and what stays fixed (price, bundle size), plus personalization governanceDeliverable: Offer catalog and written rules
- Step 3Train the propensity model: predict per player the content most likely to be boughtDeliverable: Model evaluated offline on purchase history
- Step 4A/B test: personalized offers served to a segment against the standard storeDeliverable: Conversion and revenue-per-player reading on the test
- Step 5Generalize to the whole base and automate everything that recurs on a regular basisDeliverable: ML-served offers as routine, permanent measurement loop
First step: Map the player's state (inventory, progression, purchase frequency) and test a propensity model on a segment before generalizing to the store.
The same pattern, proven in other industries
DraftKings
ML personalization and optimization of promotion and pricing across player cohorts
FanDuel
personalization of the betting offer driven by AI pricing, with a genAI discovery assistant
Sources
- S1 Supercell's Jarno Seppanen on how Clash Royale uses machine learning to automate monetization Established press archive pending
- S2 Supercell's AI masterplan explained Secondary archive pending
- S3 How Supercell is Turning Player Support into Connected and Seamless Gameplay Experience with AI Agents (Tomoro.ai) Interested party archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.