Bonus Recommendations

Overview

The Bonus Recommendations Engine, a key module within Amplifier AI, empowers sports operators to drive engagement and optimise bonus expenditure by delivering personalised bonus values for every player. By leveraging advanced AI, operators can retain players more effectively while reducing overall retention costs.

Key Features and Components

• Bonus Value Calculation provides tailored recommendations using Head-Weighted or Tail-Weighted formulas, offering either responsive or balanced bonus amounts based on recent player stakes and median patterns.
  These formulas can be customised to align with operator goals and recalculated every three hours for real-time accuracy. Learn more.
• The User Engagement Score complements bonus strategies by analysing player behaviours to highlight engagement trends. It enables customisable scoring, early intervention for disengaging players, strategic decision-making, and segmentation into ten engagement buckets for precise targeting. Learn more.
• Next Bet & Player State Detection predicts when players will next engage and categorises their activity status (e.g., Active or Churned) using machine learning models, allowing CRM teams to personalise retention strategies effectively. Learn more.

Use Cases

These features come together to empower operators with versatile use cases, enabling CRM teams to orchestrate campaigns such as re-engaging disengaged players, boosting loyalty through targeted incentives, and tailoring communication strategies for distinct player segments.

Use case 1: The CRM team aim to identify and address users with significant drop in engagement 

Once the CRM team receives a daily list of users who have experienced a significant drop in their User Engagement Scores, they can do the following to help re-engage with those players:

• Data Integration: Coupling early intervention with personalised incentives begins by integrating data from the Bet Recommendations Engine. This includes the player's favourite sports, competitions, teams, and athletes. 
• Tailored Communication: Utilising this integrated data, the CRM team sends personalized messages highlighting upcoming events, matches, or competitions related to the player's favourite sports or teams. 
• Exclusive Offers: Special promotions or bonuses are crafted based on the player's historical preferences, ensuring that the incentives resonate with the individual's interests.

Use case 2: The CRM team wants to launch a new campaign to boost user loyalty. 

The User Engagement Score categorises users for diverse A/B testing. Pre and post-campaign, the engagement score of the targeted user group is measured. Insights from these scores guide the CRM team in refining campaign strategies for enhanced user resonance.

Use case 3: The CRM team aims to tailor their communication for different player segments. 

Using the segmentation provided by the User Engagement Score buckets, the CRM team customises communication strategies for players in each segment. High-engagement players may receive VIP offers, while lower-engagement players may receive reactivation incentives. This ensures communication is relevant and resonates with each player group. 

---

How Bonus Values Are Calculated

The Bonus Recommendations Engine uses two core approaches to calculate bonus values based on betting behaviour and operational objectives. Operators have the flexibility to choose between two approaches for bonus calculation, selecting the method that best aligns with their objectives and player behaviour patterns.

• Head-Weighted Bonus Value:
  Focuses on recent player stakes, providing reactive and seasonally adaptive recommendations.
  • Pros: Captures immediate changes in player behaviour.
  • Cons: Prone to fluctuations due to outliers (e.g., a single large bet).

• Tail-Weighted Bonus Value:
  Delivers a balanced recommendation by considering median player stakes over time.
  • Pros: Avoids outliers and provides conservative bonus values.
  • Cons: Less responsive to sudden behavioural changes.

Key Considerations

• Customisable formulas tailored to operator goals.
• Existing bonuses in players' wallets can be factored in to avoid inefficiencies.
• For real-time data ingestion - bonus can be recalculated every three hours for accuracy.

---

User Engagement Score

The User Engagement Score complements the Bonus Recommendations Engine by providing a holistic view of player activity.

1. Desired Behaviours
   Higher tenancy and frequency are recognised as desired behaviours, indicating sustained and frequent user interactions.
   The User Engagement Score rewards users exhibiting these positive behaviours.
   
   
2. Undesired Behaviours
   Undesired behaviours, such as extended periods of inactivity, unpredictable engagement patterns, are identified and noted by the User Engagement Score.
   This ensures a balanced approach to encouraging consistent user engagement. 

Core Benefits

1. Customisation  
   The key strength of the User Engagement Score lies in its customisability. Operators can adapt the formula based on specific requirements, whether it is modifying the weightage of frequency and tenancy or incorporating additional parameters, the score can be fine-tuned to align with diverse business goals.
   
   
2. Early Intervention  
   The User Engagement Score acts as an early warning system for operators. By promptly identifying players with declining engagement scores, operators can intervene before the disengagement becomes significant.
   This proactive approach enables targeted strategies to retain users and enhance their overall experience.  
   
   
3. Strategic Decision-Making
   Operators can make informed, data-driven decisions to enhance overall performance. Whether it is optimising bonus recommendations, tailoring marketing strategies, or refining player experiences.
   
   
4. User Segmentation  
   The User Engagement Scores is distributed into 10 buckets to allow for easy player segmentation. Operators can categorise and target players into cohorts, each representing a different level of engagement.

---

Next Bet & Player State Detection

This feature predicts players' next bets and assigns actionable states (e.g., Active, Churned) using AI-driven analysis of betting behaviours.

Player State Detection

Player State Detection utilises the RFM (Recency, Frequency, Monetary) model to assign specific actionable states to each player based on their betting history and engagement.

Each player's state is determined by their "alive" score (e.g., Churned, Active), which is based on predefined thresholds derived from RFM metrics.

The RFM model is calculated as follows:

• Recency (R): This measures the time since the player's last bet. Any recent activity indicates higher engagement.
• Frequency (F): This quantifies the number of betting days a player has had within a defined period. Players that bet consistently are suggested as active users.
• Monetary (M): Reflects the player's lifetime value based on previous betting activity. Higher monetary value indicates valuable players.
• Algorithmic Analysis: Employs machine learning techniques to compute alive scores and predict player behaviour

Key Outputs

• Player State: Identifies engagement status.
• Next Bet Prediction: Forecasts the timing of the next betting session.

These insights enable CRM teams to craft precise retention strategies and personalised bonus offers.

---

Data Requirements & Integration

Data Ingestion 

During data onboarding, Anthem requests a sample dataset from operators to ensure that all necessary fields for effective modelling and analytics are present. Essential data elements typically include timestamp details, player and bet slip identifiers, player registration information, and comprehensive bet details—such as bet amount, bonus amount, payout, bet slip status, bet type (e.g., prematch or live, single or acca), and sport type. Optional data elements may include league, event, and market information, along with deposit/withdrawal records, bonus data, and customer demographics.

Data Delivery

Operators receive a unified CSV output that consolidates all calculated bonus recommendations alongside variables such as the User Engagement Score. This output is tailored for seamless integration into existing analytical workflows, with customisation options and real-time recalculations ensuring insights remain accurate and up-to-date.