The User Engagement Score is calculated using a formula that considers various aspects of user behaviour:
Scaling
Scores are scaled using log transformation. This transformation allows for a nuanced scaling approach, providing operators with a more granular understanding of user engagement levels. By logarithmically transforming the scores, the distribution becomes more balanced, enhancing the sensitivity of the metric to variations in user behaviour.
Sigmoid Function
Following the scaling process, the application of a sigmoid function further refines the User Engagement Scores. The sigmoid function introduces a non-linear transformation, mapping the scaled scores between 0 and 1. This transformation is instrumental in providing a smooth and gradual progression in scores, ensuring that even slight variations in player engagement are discernible. The sigmoid function is mathematically expressed as:
Here, 'smoother' is a parameter that controls the rate of the transformation. Adjusting this parameter allows operators to fine-tune the difficulty level of achieving a score of one for users. A lower value increases the difficulty, making higher scores more exclusive and indicative of exceptional player engagement.
Score Buckets
Once the sigmoid transformation is applied, the User Engagement Scores are categorised into 10 distinct buckets. These score buckets serve as a user-friendly segmentation, allowing operators to quickly and intuitively identify different cohorts of players based on their engagement levels.