Thursday 17 July 2025
The quest for a more accurate way to predict customer lifetime value has been an ongoing challenge in the digital gaming industry. With billions of dollars at stake, game developers and publishers are constantly looking for ways to better understand their users’ behavior and optimize their monetization strategies.
One approach that’s gained popularity is machine learning, which involves training algorithms on large datasets to make predictions about user behavior. However, traditional machine learning models often struggle with the complexities of customer lifetime value prediction, such as delayed payment behavior, sparse early user data, and the presence of high-value outliers.
Researchers have now proposed a new framework called SHORE, which aims to address these challenges by integrating short-horizon predictions with long-term targets. The idea is that by combining information from different time scales, SHORE can better capture the nuances of customer behavior and make more accurate predictions about lifetime value.
The key innovation behind SHORE is its use of auxiliary tasks, which involve training separate models to predict user behavior at shorter horizons (such as one week or one month) in addition to the traditional long-term target. This allows the model to learn patterns and relationships that might not be apparent from a single, long-term prediction.
SHORE also employs a novel loss function that combines order-preserving multi-class classification with a dynamic Huber loss. The former helps the model preserve the monotonic structure of lifetime value accumulation, while the latter is designed to mitigate the influence of extreme-value users and outliers.
To test SHORE’s efficacy, the researchers conducted extensive offline and online experiments on real-world datasets from digital games. The results were impressive: SHORE outperformed existing baselines by a significant margin, achieving a 47.91% relative reduction in prediction error.
The implications of SHORE are far-reaching. By providing more accurate predictions about customer lifetime value, game developers can optimize their monetization strategies and make more informed decisions about user acquisition and retention. This could lead to increased revenue and profitability for the gaming industry as a whole.
While SHORE is specifically designed for digital games, its principles and techniques could be applied to other domains where customer lifetime value prediction is critical, such as e-commerce or digital finance. As machine learning continues to evolve and improve, we can expect to see even more sophisticated approaches to predicting customer behavior and optimizing business strategies.
Cite this article: “Predicting Customer Lifetime Value with SHORE: A New Framework for Digital Gaming”, The Science Archive, 2025.
Machine Learning, Customer Lifetime Value, Digital Gaming, Shore, Prediction, User Behavior, Monetization Strategies, Game Developers, Revenue, Profitability
Reference: Congde Yuan, “SHORE: A Long-term User Lifetime Value Prediction Model in Digital Games” (2025).
