Development of an End-to-end Machine Learning System with Application to In-app Purchases

Aug 5, 2026

Upload by Ophelia

Machine learning systems are essential in mobile gaming to optimize player experience, particularly in-app purchases. This paper presents an ML system designed to predict when a player is likely to make their next in-app purchase, enabling personalized offer timing. It outlines the problem definition, the modeling approach, and key results, then provides a detailed end-to-end description of the full ML pipeline. The work concludes with lessons learned, challenges faced, and plans for future research.

LanguageEnglish

CategoryTechnology

FormatPDF

Pages8

Reading Time20 min

AccessFree to download

Introduction

Offer personalization

Data-driven timing for personalized offers

arXiv :2412 . 12390v1 [ cs .LG] 16 Dec 2024 Development of an End-to-end Machine Learning System with Application to In-app Purchases Dionysios Varelas Activision Blizzard King London, UK [dionysis.varelas@king.com](dionysis.varelas@king.com) Anders Englesson Activision Blizzard King Stockholm, Sweden [anders.englesson@king.com](anders.englesson@king.com) Elena Bonan Activision Blizzard King Barcelona, Spain [elena.bonan@king.com](elena.bonan@king.com) Christoffer Åhrling Activision Blizzard King Stockholm, Sweden [christoffer.ahrling@king.com](christoffer.ahrling@king.com) Lewis Anderson Activision Blizzard King London, UK [lewis.anderson@king.com](lewis.anderson@king.com) Adrian Chmielewski-Anders Activision Blizzard King Barcelona, Spain [adrian.chmielewskianders@king.com](adrian.chmielewskianders@king.com) ABSTRACT Machine learning (ML) systems have become vital in the mobile gaming industry. Companies like King have been using them in production to optimize various parts of the gaming experience. One important area is in-app purchases: purchases made in the game by players in order to enhance and customize their gameplay experience. In this work we describe how we developed an ML system in order to predict when a player is expected to make their next in-app purchase. These predictions are used to present offers to players. We briefly describe the problem definition, modeling approach and results and then, in considerable detail, outline the end-to-end ML system. We conclude with a reflection on challenges encountered and plans for future work. KEYWORDS Machine Learning, Machine Learning Platform, Mobile Gaming, Personalization, In-app Purchases ACM Reference Format: DionysiosVarelas, Elena Bonan, Lewis Anderson, AndersEnglesson, Christoffer Åhrling, and Adrian Chmielewski-Anders. 2023. Development of an End-to-end Machine Learning System with Application to In-app Purchases. In Proceedings of 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD’23) . ACM, New York, NY, USA, 8 pages. 1 INTRODUCTION Players of King games (e.g. Candy Crush Saga, Candy Crush Soda Saga), progress through a map of levels of increasing difficulty by solving match-3 puzzles. The speed of progression depends on the player’s skill level which is a combination of deciding on the right move to make, the right boosters to use and the right time to use them. For example, a skilled player would know exactly when to use a chocolate bomb to overcome a really challenging gameboard. Not all players need extra boosters and those who need extra Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions [from permissions@acm.org](from permissions@acm.org). KDD’23, August 06–10, 2023, Long Beach, CA © 2023 Association for Computing Machinery. do not purchase them with the same frequency. Indeed, purchase behavior varies a lot from player to player and depends on multiple characteristics related to both long-term and short-term player behavior. Machine learning (ML) allows us to combine a large number of behavioral and game-related features to predict when a player will make their next in-app purchase. This insight can inform our personalization engine to offer unique gameplay experiences to every player. In section 2 we will expand on the use case for personalized offers. 2 OFFER PERSONALIZATION Historically, personalized experiences at King were achieved using rule based approaches that were based on large

FAQs

What problem does the end-to-end ML system address?

It predicts when a player is expected to make their next in-app purchase so that the game can present offers at the right time.

How are personalized offers determined in the system?

The offer timing is driven by an ML model that estimates the most likely next purchase moment, while offer location and contents are handled through tested configurations and prior learnings.

Why is data-driven timing important for in-app purchase offers?

Player engagement and purchase intervals vary significantly, so fixed timing fails to match individual purchasing patterns and personalized timing improves relevance.

Aug 5, 2026

Upload by Ophelia