Scoring Methodologies
Overview
At the heart of Provider Intelligence is a sophisticated scoring engine that converts raw performance data (latency in milliseconds, packet loss percentages, etc.) into normalized, comparable scores (0–100). These scores enable apples-to-apples comparisons of providers and determine the overall scores.
Understanding how scoring works is essential for:
Customizing queries: Choosing the right results calculation and metric priorities.
Interpreting results: Knowing why one provider scores higher than another.
Validating decisions: Explaining the methodology to stakeholders who want to understand the logic behind the recommendations.
Additionally, we calculate metric trends, identifying whether raw metric performance is improving, worsening, or staying the same over time.
This section provides a detailed breakdown of the algorithms used.
Key Principles
Relative vs. absolute scores: A provider's performance score calculation, which transforms raw values into normalized scores, is calculated relative to other providers in the same city during the same time period. A score of 100 means "best in class for this location," not "perfect performance." Other score types, such as stability and overall score, are either calculated in an absolute manner or derived from already normalized calculations.
Higher is always better: All scores are normalized such that 100 is optimal and 0 is worst. This applies even to metrics like latency and packet loss, where the raw value should be low.
Hourly granularity: All metric scores are first calculated per hour, then averaged over your selected time frame (1, 3, or 6 months).
Related Articles
Last updated