How Protocols & Packets measures, scores, and reports public network signal, and where the boundaries are.
The Global Network Health Index (GNHI) is a composite indicator representing overall network stability. It is not a measurement — it is a system indicator that aggregates multiple independent signals into a single interpretable score.
GNHI is calculated globally and by region (North America, Europe, Asia Pacific). It is updated continuously using rolling time windows.
GNHI is derived from four categories of public signal:
Each input signal is normalized to a 0–1 scale against its own regional baseline. The composite GNHI score is a weighted combination of these normalized signals, scaled to a 0–100 range:
GNHI = Σ (wᵢ × normalized_signalᵢ) × 100
Where wᵢ are the metric weights and the sum of all weights equals 1. A high score (closer to 100) indicates stable, predictable networks. A declining score indicates stress, congestion, or instability.
Weights reflect the relative importance of each signal category. They may be adjusted as signal coverage improves, but conceptual integrity is preserved. Current approximate weights:
| Signal | Weight |
|---|---|
| Latency deviation | 0.30 |
| Packet loss clustering | 0.25 |
| Jitter volatility | 0.20 |
| BGP routing instability | 0.25 |
| Score Range | Interpretation |
|---|---|
| 80–100 | Stable — networks behaving predictably |
| 60–79 | Elevated — noticeable deviations from baseline |
| 40–59 | Stressed — significant instability detected |
| 0–39 | Degraded — widespread disruption signals |
Short drops may indicate isolated events. Sustained declines indicate systemic issues.
Every GNHI score is accompanied by a confidence value. Confidence reflects how much independent signal the observatory has to support its score — not how "correct" the score is.
Confidence scales with the number of independent data contributors. With few contributors, confidence is low because the index may reflect a narrow view. As contributors increase, confidence rises along a saturation curve: each additional contributor adds progressively less confidence.
This means early contributors have a large effect on confidence, while additional contributors beyond a certain threshold provide diminishing returns. The curve is designed so that confidence approaches but never reaches 100%.
Protocols & Packets is built on public, legal, defensible data and opt-in contributions only. Trust is the product.
An observation is a directly measured signal from public or opt-in data. An inference is a conclusion drawn by correlating multiple observations. Inferences are always expressed probabilistically. The observatory reports correlations and observed patterns — it avoids asserting causation without evidence.
Aggregation is preferred over raw precision. Trends and deltas matter more than absolutes. Indices are used to express system health; they are not exact measurements.
Protocols & Packets does not assign blame, does not editorialize, and does not advocate for vendors, carriers, or policies. The platform observes and reports.
All published language is technically accurate, defensible under scrutiny, and free of speculation or sensationalism. Phrases implying certainty without proof are avoided.
Protocols & Packets is not a network management system, a monitoring tool, a vendor comparison platform, or a news site. It is a neutral observatory.