How the Reliability Score is calculated
Published in full so that any reader, vendor or organizer can reproduce a score from the underlying data.
The formula
Score = Uptime (max 40)
+ Incident Severity (max 25)
+ MTTR Performance (max 15)
+ Transparency & Disclosure (max 10)
+ Verified Event Volume (max 10)
− Penalties
1. Uptime — 40 points
Calculated across all completed events in the rolling window as
(1 − total disruption minutes ÷ total scheduled minutes) × 100.
- 100% uptime → 40 points
- 99.0% → 30 points
- 95.0% → 15 points
- 90.0% or below → 0 points
- Linear between each pair of anchors
The curve is deliberately smooth rather than stepped. A sharp threshold would let a difference of a few minutes swing a score dramatically, which is neither statistically sound nor defensible if a vendor challenges the figure.
2. Incident severity — 25 points
Based on the rate of vendor-attributable incidents per event. Incidents whose root cause code assigns responsibility to the venue, ISP, publisher or organizer are excluded, as are incidents explicitly flagged not attributable to the network vendor.
A rate of 0 incidents per event scores the full 25. A rate of 0.5 or higher scores 0.
3. MTTR performance — 15 points
Mean time to resolve, averaged across vendor-attributable incidents.
- 5 minutes or under → 15 points
- 45 minutes or over → 0 points
- Linear between those anchors
- No vendor-attributable incidents recorded → full 15 points
4. Transparency and disclosure — 10 points
- 6 points — proportion of right-of-reply notices answered within the 72-hour window
- 4 points — proportion of events for which live network telemetry was published
5. Verified event volume — 10 points
A logarithmic curve rewarding a sustained track record: 3 events scores roughly 3 points, 12 or more scores the full 10. This prevents a vendor with a single flawless event from outranking one with a long clean record.
Penalties
- −2.0 per incident (capped at 5.0) where no reply was received within the right-of-reply window
- −1.0 per completed event (capped at 4.0) delivered with no backup uplink declared
What is excluded
- Unconfirmed community reports never affect a score. They are displayed for transparency and nothing more.
- Incidents in draft, withdrawn or notice sent status are not counted; only published and amended records are.
- Events outside the rolling 24-month window are excluded. The window currently begins 08 Sep 2024.
- Events not yet marked completed are excluded from uptime.
Insufficient Data
A vendor with fewer than 3 qualifying events in the window is shown as Insufficient Data — never as a low score. A new entrant should not be penalised for being new, and a single bad event should not permanently define a vendor.
Recalculation and disputes
Scores are computed on request from live data, so any correction to an underlying record is reflected immediately. Every change to an underlying record is written to the hash-chained audit log, and material corrections are published in the public corrections log.
A vendor who believes a score is wrong should identify the specific underlying record they dispute. Scores are not adjusted directly; only the underlying facts are corrected, and the score follows.
Operator entries
The registry operator is also a listed vendor. Its entries are scored by exactly this formula with no exemption or adjustment. Operator staff recuse themselves from classification decisions affecting their own records; those are reviewed by the independent Advisory Board. See the governance charter.