Apple iOS 14: Mobile Gaming Advertising in the post-IDFA world A study of predicting customer lifetime value in an anonymized data landscape

Keywords
Loading...
Thumbnail Image
Issue Date
2021-02-14
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
In September 2020, Apple Inc. announced that the next update for iOS devices (iOS14) will include an updated privacy policy regarding user tracking and targeting. Speci cally, the privacy changes entailed a deprecation of the current Identi ers for Advertisers (IDFA), which in the future requires mobile advertisers to ask users for tracking permission when a newly installed app is opened for the rst time. Currently, advertisers rely on real time tracking, user-level analytics and retargeting of users to measure the success of advertising campaigns. However, after iOS14 all their incoming data will be anonymized and summarized through a conversion value if users opt-out of tracking. These conversion values will then be send through Apple's proposed SKAdNetwork which will send the data to advertisers after a randomly initiated timer between 24 and 48 hours has passed. Current estimations state that the opt-out rate will be roughly 90% of users (Koetsier et al., 2021). Therefore, advertisers and mobile developers need to nd alternative methods to predict important metrics such as customer lifetime value (LTV) with anonymous data. This thesis introduces a LTV model that uses forecasted user retention (opt-ins) through a long-short term memory neural network in combination with simulated average revenue per daily active user (ARPDAU) data without using user-level IDFA values. The resulting best performing model has shown to accurately predict LTV values of players in the iOS game Road Crash (RMSE = 0.052, mean overall prediction error of $0.042) when compared to an existing LTV model using pre-iOS14 data. The results demonstrate that even in a post-IDFA marketing world opt-out user data can still be used to predict the lifetime value of users.
Description
Citation
Faculty
Faculteit der Sociale Wetenschappen