Joe Eddy, Yikai Kuo, Sophie Long
Frosty Metrics
Chill Factor Under the Hood: The Technical Gears Behind Our Personalized On-Chain Activity Scoring System
Overview
We’ve recently launched an address-level analytics and scoring feature as an enhancement to our long-standing Frosty Metrics platform, a one-stop-shop for Avalanche data and analytics. Chill Factor is a first-in-class system for the Avalanche ecosystem akin to the Ethereum DegenScore. Our system is designed to be intuitive and transparent to users, leveraging metrics that are easy to understand and rewarding a holistic range of accessible on-chain activity patterns. Chill Factor empowers users to better understand their own historical activity and place it against the context of the entire C-Chain user base.
In this post, our Data Scientists dive into the technical methodology that drives Chill Factor and share some of the most interesting insights yielded by this work.
Methodology and Evaluation
The Chill Factor Score is derived from a sum of the Activity, Diversity, and Consistency component scores. Each component score is calculated as a weighted sum of corresponding metrics (normalized by min-max scaling). To prevent any single metric or component from dominating the score, a decay factor is applied to each component score by adding a small value then taking a logarithm.
Insights and Takeaways
Chill Factor Score Distributions Across Tiers Highlight Diverse User Behavior and Progressive Engagement Levels Within Tiers
The distributions of Chill Factor scores across tier buckets demonstrate the wide variety of Avalanche user profiles. Intra-tier variance is greatest for the Warm and Subzero tiers, which have exponential distributions and a few users exhibiting either very high or very low scores. The linear distributions in the middle tiers emphasize the consistent progression experience that users can expect when they engage with this system.
Component Score Distribution Across Tiers Reveal a Healthy Progression Trend:
The distributions of component scores show a step-like pattern, presenting a healthy progression trend. Scores for Warm-tier users do show a wider spread, which is unsurprising as score contributions for newer and exploring users will typically be more concentrated around specific activities than those of deeply engaged, high-tier users who exhibit more range.
Balanced Component Contribution Promotes Multifaceted User Behavior:
The Chill Factor score is designed to promote multifaceted user behavior. Each component contributes between 20% and 40% to the overall score for all user tiers (except the warm tier). As users move up the tiers, activity becomes more important, while diversity's influence decreases slightly, and consistency stays steady. This setup encourages users to get more involved in a range of platform activities. The warm tier users show significant behavioral variance, with diversity being the primary contributing factor.
Highlight User’s Chill Factor Journey: a Year From Warm to Subzero
Highlight User’s Component Scores Over Time:
This real wallet highlight (address redacted for privacy) demonstrates how a user can go from Warm to Subzero in just a year! This wallet belongs to an active retail C-Chain user who regularly trades low market capitalization assets, NFTs, and holds several blue chip tokens (AVAX, BTC.b). This user’s Chill Factor score evolution follows a fascinating trend: diversity trumps consistency, which trumps activity. But as the user climbed the tiers, activity's share grew significantly. 2024 marked a turning point as the user’s activity skyrocketed. In just two months at the start of the year, the user went from chilly to subzero!
Taken together, the bird’s-eye view and address spotlight reveal Chill Factor’s ability to measure accessible and steady progress across a variety of interactions with the C-Chain. We’re excited to see users further engage with the Avalanche ecosystem through the lens of Chill Factor. Our roadmap for this feature also includes an expansion of metrics/activity patterns and heightened levels of personalization, so stay tuned!
Sign up to receive an email when we release a new post