Problem
Our client, an online sports betting company, faced the challenge of effectively segmenting their user base. They needed to understand which users belonged to specific categories and be alerted to any changes that could impact their business. Identifying users who were becoming less valuable, such as those who rarely played or made low payments, was crucial for the risk and marketing teams.
Specific pain points and obstacles:
- Limited visibility into user categories and changes over time
- Difficulty identifying valuable players versus less active or low-spending users
- Inability to predict shifts in user behavior or preferences
Solution
To address the challenge, the SmartCat team analyzed the extensive dataset consisting of nearly 1,000,000 players and over half a billion transactions. We extracted key features, including profit, margin, payments, active months, and duration of inactivity. These features were used to represent each user, and we applied the K-means clustering algorithm to segment users into three categories.
By examining the features of users within each category, distinct patterns emerged. One category consisted of highly valuable players who played strategically and wagered large amounts, while another category comprised frequent players who bet smaller amounts for recreational purposes. We presented these findings through interactive dashboards, allowing end-users to customize reports based on their preferences.
Specific strategies and tactics used:
- Data analysis of player transactions and behavior
- Feature extraction and representation for each user
- Application of the K-means clustering algorithm for segmentation
- Development of separate dashboards for different game types
- Integration of customizable filters to enable personalized reporting
Unique value brought by our company:
Our expertise in data analysis, machine learning algorithms, and dashboard development enabled us to provide an effective solution for user segmentation. By creating intuitive dashboards using Apache Superset, we empowered the client to gain actionable insights and monitor user behaviors in real-time. The customization options allowed them to make informed decisions and tailor marketing strategies based on the evolving needs of their user base.
Technologies Used
- Python for data analysis and algorithm implementation
- Apache Airflow for workflow automation
- Apache Superset for dashboard creation and customization
Results
The clustering approach we implemented was evaluated by analyzing the characteristics of users within each cluster. The clusters were successfully formed and exhibited clear distinctions, enabling effective segmentation of the user base.
Key metrics:
- Profit tracking for specific games
- Margin analysis
- Number of players per category
- User behavior changes based on cluster affiliation
Quantifiable results:
- Improved understanding of user categories and behavior patterns
- Enhanced ability to identify valuable players versus low-activity or low-spending users
- Real-time monitoring of changes in cluster affiliation for targeted marketing strategies
Smart Tip
When segmenting users based on financial transactions, it is essential to consider multiple features and apply clustering algorithms to identify distinct groups. The ability to customize dashboards and filters empowers stakeholders to extract actionable insights and make data-driven decisions for their business.
Smart Fact
Through user segmentation, our client discovered that valuable players are not solely defined by frequency of play but also by strategic wagering and larger bet amounts. This insight allowed them to adjust their marketing approach and tailor promotions to attract and retain high-value users.
About the Client
Our client is an online sports betting company operating in the gambling industry. With a significant user base and a focus on risk management and targeted marketing, they sought to gain a deeper understanding of their users and improve their decision-making processes.
