How the Trust Score Works
Most platforms show you a star average. HYFY goes deeper. Every Trust Score passes through a multi-stage pipeline designed to produce a number that is fair, resistant to manipulation, and reflective of genuine community perception.
The Scoring Pipeline
When someone rates you on HYFY, the raw input goes through several transformation stages before it becomes part of your visible Trust Score.
1. Raw Rating
A user selects a rating for a specific trait, a named quality that matters for the type of profile being rated. Traits are organized by persona (social or professional) and each carries a weight reflecting its importance. The available scale depends on the type of account being rated (see For People and For Brands).
2. Calibration
Not everyone rates on the same curve. Some people are generous raters, others are harsh. HYFY adjusts for this by comparing each rater's pattern against the community average for each trait. This removes individual bias so that a "4" from a tough critic carries appropriate weight.
3. Normalization
After calibration, the score is mapped to a unified internal scale. This ensures that all ratings - regardless of which scale they originated from - are directly comparable within the scoring engine.
4. Bayesian Smoothing
A single rating shouldn't define you. HYFY uses a Bayesian prior that blends your actual ratings with a community baseline. This means:
- New users aren't penalized for having few ratings yet
- Scores stabilize quickly once enough feedback arrives
- Outlier ratings (extremely high or low) don't cause wild swings
5. Trait Aggregation
Your profile isn't a single score - it's a weighted composite across multiple traits. Each trait carries a configurable weight that reflects its importance. Your final Trust Score aggregates all trait-level scores proportionally.
6. Display Score
The internal score is mapped to the display range for your account type. Additional guardrails apply at this stage - for example, rated people never display below a minimum floor (see For People).
Key Design Principles
| Principle | What It Means |
|---|---|
| Multi-signal | Your score is built from multiple traits, not a single number |
| Bias-corrected | Rater calibration removes individual bias patterns |
| Sybil-resistant | Reviewer credibility weighting reduces the impact of fake or low-quality raters |
| Recency-aware | Older ratings naturally carry less weight over time |
| Abuse-proof | Rate limiting, cooldown periods, and audit trails prevent manipulation |
| Convergent | Scores stabilize with more ratings - they don't swing wildly |
Reviewer Credibility
Not all ratings carry the same weight. HYFY assigns each reviewer a credibility score based on multiple factors:
- Experience - How many ratings have they submitted over time?
- Accuracy - How close are their ratings to the eventual community consensus?
- Reputation - What is their own standing in the community?
These factors are combined into a reviewer weight that scales each rating's influence. New users start at a baseline weight and earn influence as they demonstrate quality judgment.
Note: The exact formula and parameter values for reviewer credibility are proprietary. The system is designed so that no single rater - regardless of how many accounts they control - can meaningfully distort another user's score.