Methodology · Method note 01
A score should not be a black box.
This page documents how a FlightReady readiness assessment is produced: what the score represents, what it does not represent, how the factors combine, where the AI is allowed to work, and where the system’s limits are.
Score semantics
What the number means — and what it doesn't.
The composite is a 0–100 advisory risk figure: a weighted summary of how much attention the available factors deserve. It is not a probability of an accident, a certification of safety, or a decision.
Lower advisory risk based on available data. Verify official sources.
Elevated advisory risk. Review drivers and mitigations before departure.
High advisory risk. Instructor or safety-manager review is recommended.
Multiple or severe advisory risk factors detected. Qualified review recommended.
Factor taxonomy
Eleven factors, weighted in the open.
Base weights are published below exactly as the engine applies them. Weather carries the most weight; traffic the least. No hidden factors, no adjustable dark knobs.
- Weather
- Pilot readiness
- Personal minimums
- Aircraft readiness
- Terrain
- Airports
- Accident history
- NOTAMs
- Human factors
- Weight & balance
- Traffic
Missing information
When data is missing, the assessment says so.
Factors marked unavailable are excluded from the weighted total. Their weight is redistributed proportionally across the factors that are available — and the assessment's confidence figure drops by a published penalty for each missing factor.
Example · confidence accounting
A high score with low confidence is a different situation from a high score with full data — and the assessment shows you which one you have. Weather absence carries the largest penalty; it is the factor the assessment least tolerates missing.
The AI boundary
Deterministic where it counts. Generative only where it explains.
What is deterministic?
The score. Each of the eleven factors is scored 0–100 by fixed rules over retrieved data and your inputs, then combined under the published base weights. Running the same inputs against the same data produces the same score.
What is retrieved?
Weather products from the Aviation Weather Center at analyze time, FAA airport and aeronautical data, terrain elevation along the route, and historical NTSB accident records near your corridor. Each retrieval is stamped with its source and age.
What is generated?
Only the narrative briefing. A language model drafts plain-language explanation of what the scoring engine found, citing the retrieved NTSB records behind it. If no relevant source records are retrieved, it does not invent them.
What is the AI never allowed to do?
Set or adjust the score, add factors, or overrule the deterministic engine. The model explains; it does not decide.
Accident relevance
How historical records are selected.
NTSB records are retrieved by geographic proximity to your route corridor, then weighted by distance and recency: closer and more recent events matter more. The result feeds both the accident-history factor and the briefing's citations.
Historical similarity does not predict an outcome. The value of a surfaced record is the pattern it shows — what conditions, decisions, and terrain combined to produce the event — not a forecast that it will repeat.
Operational guidance
How a pilot should use the result.
Structure your review, surface factors you might not have weighed, and state a rationale
Skip an official briefing, justify a flight against your own minimums, or outsource the decision
Conditions change, departure slips, the route changes, or the assessment is more than a few hours old
A factor contradicts what official sources tell you — official sources win, every time
Known limitations
What this system cannot do.
An honest tool states its limits. These are ours.
The score is advisory. It is not an official briefing, a dispatch release, or a measure of legal compliance.
An assessment reflects conditions at analyze time. Weather changes; re-run the assessment when it does.
NOTAM coverage varies with data availability. The app labels reduced coverage — always check NOTAMs through official channels.
NTSB records are historical investigative information. Historical similarity does not predict an outcome.
Factor scoring uses documented heuristics, not a statistically validated accident-prediction model. We do not claim the score predicts accident probability.
Missing inputs (for example, no personal minimums configured) reduce what the assessment can evaluate — and its confidence figure shows that.
The assessment does not know what you know: local knowledge, the aircraft's actual squawks, or how you feel today. Bring those to the decision yourself.
Methodology questions
Question about how a factor is scored, or a case where the assessment got it wrong? We want to hear it — that feedback is how the methodology versions forward. Write to support@flightready.ai.