Method
How the Score Is Built
What This Page Is
This page describes how FlyNook turns a flight's data into a number and a set of findings, as the product does it on the date above. It is versioned: when the method changes, the version and the changelog at the foot of the page change with it. Where something is planned rather than built, it says so.
An Index, Not a Probability
The 0–100 number is a Relative Risk Index. Higher means more hazard than usual for flights like this one. It ranks; it does not measure likelihood. A 64 is not a 64% chance of anything, and FlyNook publishes no probability of its own.
Six factors — the flight's weather, its aircraft, terrain, the landing field, NOTAMs, and weight and balance — are each scored 0–100 from tables set by people, not fitted to outcomes. The index is the higher of the weighted mean of the factors that could be scored, using the published weights below, and 0.8 times the worst of them, so one severe factor is never averaged down: LIFR at one end and VFR at the other reads High, not Low. Outcome data does not yet establish the size of any weight.
Accident history near the route is shown beside the index as context, never weighted: a count is not a rate. Your pilot profile, personal minimums and fitness answers are checked as findings — a minimum exceeded sends the flight to review — and never weighted, so the index is the same for every pilot on the same flight. While a CFI reviews this change, no flight's band falls below the band the previous method gave it; the page and the record say when that holds a number.
A factor that could not be read is left out of the mean and its weight is shared among the rest — with one guard: a missing factor is never allowed to lower the index. If counting it at its not-assessed level — set above what a source that answered and found nothing scores — gives a higher index, the higher value stands.
- Low0–29
- No additional review required.
- Elevated30–59
- Review the flagged drivers before departure.
- High60–79
- Apply mitigations, or defer.
- Review Required80–100
- Requires authorisation before departure.
A band names what a school's policy asks for next. It is not a verdict, and the same cut points apply in every school and every account.
The Factors and Their Weights
Each share is that factor's weight in a fully informed index (engine v10). When a factor is missing, the others' shares grow in proportion.
- Weather38.5%
- METAR and TAF at each end over the departure window, and SIGMET and G-AIRMET advisories along the route.
- Aircraft readiness18.5%
- The aircraft you selected and its profile.
- Terrain / route15.4%
- The highest terrain along the route against your planned cruise altitude.
- Airport / airspace12.3%
- Runway length at the landing field (FAA NASR) and its elevation.
- NOTAMs7.7%
- NOTAMs at the departure and destination.
- Weight & balance7.7%
- The latest weight-and-balance calculation saved for the selected aircraft.
What Is Shown Ahead of the Number
Some findings matter more than any average, so they are shown before the index, at every screen width, and are never folded into it silently:
- Critical findings
- Below VFR weather minimums (14 CFR 91.155): a flight not filed IFR, with a ceiling below 1,000 ft or a visibility below 3 SM reported at either end when it leaves within three hours or no forecast covers its window, forecast for its window (by the field's TAF or, for a local flight's field with none, by NBM guidance at 50% or more or the nearest TAF), or read or forecast for its working area. A convective SIGMET in force during the flight within 20 NM of the route or the working area (AIM 7-1-27). Weight and balance outside limits (14 CFR 91.9) on the calculation on file.
- Your minimums
- Any personal minimum the flight exceeds.
- Weather
- IFR or LIFR in the current report at either end.
- NOTAMs and TFRs
- NOTAMs graded critical, and TFRs along the route or over the practice area.
- Fuel and runway
- Fuel below reserve, and a runway shorter than your minimum.
A critical finding is never averaged away: it requires review, names its rule, and holds the index at no lower than 60 (High). Until a source resolves the airspace, the stricter 1,000 ft and 3 SM apply even where Class G would allow 1 SM clear of clouds. A flight with no weight-and-balance calculation is not checked, never within limits: weight and balance is listed among the sources that could not be read.
Today these findings inform the pilot's decision; they do not lock it. A Go against a flight the engine flags for review is recorded with the pilot's reason. On a school flight, an exceeded minimum, a critical finding, a school threshold or an index in High sends the flight to an instructor's review, and an index in Review Required to the chief pilot's (a chief instructor or a school admin), whichever of the engine and the school's policy asks for more, and the school cannot release a flight whose review was rejected, or an aircraft with a grounding squawk or a maintenance hold, whoever asks. Planned, not yet built: a status that stops a flight whatever the index says, each cited to the regulation or the school rule it enforces.
When Data Is Missing
Every factor is reported as read, estimated or unavailable, and the result carries a data-completeness figure: how much of what the index depends on was actually read. A source that could not be read is shown as missing. It is never shown as clear.
A source that was not read, or whose read failed, is held at its not-assessed level and never scored as clear; it is named on the verdict, on the saved record and in the PDF.
When the weather at either end, the forecast the flight needs, the NOTAMs, the TFR check or the SIGMET and G-AIRMET advisories could not be confirmed, or the inputs have aged since the check ran, the result does not state its most favourable finding. It reads “Incomplete — verify manually” and names what was not confirmed.
Known gaps, being closed: when terrain cannot be resolved, or when no pilot profile or aircraft is set, those factors are scored at a stated default rather than marked unavailable; the weight-and-balance factor and its finding read the latest calculation saved for the aircraft, which may belong to an earlier flight, and check its weight and its forward and aft CG limits only, not the envelope's sloped edges or the landing CG; no flight rules are recorded yet, so an IFR flight is held to the VFR minimums too.
Not read yet: thunderstorms reported or forecast in a METAR or TAF (TS, CB) or seen by lightning detection, where no convective SIGMET covers them; and the weather between the two fields of a cross-country, beyond the SIGMETs and G-AIRMETs along the route; and, for a cross-country field with no TAF, its NBM guidance or the nearest TAF (its current report decides instead). None of these is a finding today: check them in an official briefing.
Where the Data Comes From
Each saved briefing lists its sources. As of the date above, they are:
- NOAA Aviation Weather Center
- METAR, TAF, PIREP, SIGMET and G-AIRMET, read when the check runs. US Government work, public domain.
- NOTAMs
- SkyLink, a commercial NOTAM service used under its own terms, and the FAA's NOTAM service where it answers. Not a sole source: read NOTAMs in an official briefing.
- FAA TFR list
- Temporary flight restrictions and their details from tfr.faa.gov. Public domain; the FAA describes it as informational and not a sole source.
- Terrain
- USGS elevation, from AWS Terrain Tiles: 3DEP (formerly NED), GMTED2010 and SRTM terrain data courtesy of the U.S. Geological Survey. Public domain in the US.
- Runways and airports
- FAA NASR runway data on the 28-day cycle, and an airport index built from FAA and OurAirports records. Public domain.
- NTSB accident data
- The NTSB aviation accident database, ingested and searched near the route. Public domain. FlyNook's summaries are not NTSB findings.
- NOAA National Blend of Models
- On local flights, when a recent model run is on hand, the chance of a low ceiling or visibility over a practice area, shown as issued with its model run and forecast hour. NOAA open data.
- OpenSky Network
- Live ADS-B traffic near the route, read only through FlyNook's licensed access and shown as context with its read time. It is not part of the index.
- What you enter
- Your pilot profile, aircraft, personal minimums, weight and balance, and preflight answers.
Maps are drawn with Mapbox; no map data feeds the index. FlyNook is not affiliated with, or endorsed by, the FAA, the NTSB, NOAA or the National Weather Service.
How Often It Looks Again
A check reads its sources when you run it; it is not a live feed. Route Watch re-checks watched routes once a day, at 11:10 UTC, for departures in the next 72 hours, and emails you when the result crosses your threshold.
What FlyNook Will Never Claim
These are not claimable at FlyNook's scale, or not true of the product by design. The same list drives the check that every page and email in the product is held to.
NeverNot claimable at FlyNook's scale, or not true of the product by design.
- A per-flight probability of an accident or a fatality.About 1.5 million training flights would be needed to detect a 50% miscalibration of the instructional accident rate.
- That FlyNook reduces or prevents accidents by any amount.No causal design at FlyNook's scale can establish it.
- That a flight is safe, cleared or approved, or that safety is guaranteed.FlyNook states findings; the pilot in command decides (14 CFR 91.3). Unknown is never safe.
- That the index or any FlyNook number is a probability, a percentage risk or a prediction of an accident.The index is ordinal: it ranks hazard, it does not measure likelihood.
- That the model, the score or the method is validated.Say exactly what was verified, on how many forecast-observation pairs, in which published report.
- Confidence, as a word or a percentage, for how complete the data is.Completeness is a count of what was and was not read; call it data completeness.
- Monte Carlo, or any simulation, before a simulation of forecast uncertainty ships.Earlier material implied a Monte Carlo that does not exist.
- A monitoring cadence the scheduler does not run, such as 'every 15 minutes' or 're-scored until departure'.Production schedules are daily; state the cadence actually scheduled.
- FAA approval or certification, an official briefing, dispatch or operational control, 'AC 120-78B-conformant', or 'policy-as-code satisfies 141.93(a)'.None is true of the product; counsel signs off any wording that mentions the FAA.
- Any claim about a named competitor on a public page.Public pages state FlyNook's own commitments. Internal competitive claims are dated and expire 90 days after their last verification.
Verification Before Any Probability
What may be said is unlocked in stages, and each stage needs published evidence before its claim is made. The only probabilities shown today are National Weather Service guidance, as issued and labelled.
Claimed nowBacked by evidence today.
- A transparent, rules-based Relative Risk Index: a heuristic index from expert-set factor scores and weights, never a probability.
- Every factor and its weight is published, and the weights are expert-set rather than fitted to outcomes.
- Data completeness is shown, and a source that could not be read is shown as missing, never as clear.
- The method is published, dated and versioned, with a changelog.
- FlyNook records information and decisions; it does not make them. The pilot in command is the final authority (14 CFR 91.3).
- Inputs come from named public sources, listed with their licences on the method page and on each saved briefing.
Next, in this order: a public verification of NWS guidance against archived observations — how often LAMP and NBM ceiling and visibility forecasts were right, by lead time and station. Only after that, a FlyNook probability that a named weather condition at your arrival time exceeds your limit, for the stations and lead times the verification covers, and nowhere else.
NextNot claimed until its evidence is published.
- A calibrated probability that a named weather variable at your ETA exceeds your limit, for verified strata only.Not claimed until: Verification targets met, and the report published, for that variable, lead time, station class, guidance product and region
- A public verification of NWS guidance (LAMP and NBM) against archived observations, published before any FlyNook probability.Not claimed until: The verification of NWS guidance published
- Simulation of forecast uncertainty behind an input-uncertainty range.Not claimed until: The simulation ships, versioned and reproducible
- Named customers, quotes and case studies.Not claimed until: A real customer's written permission, and counsel review under the FTC rule on reviews and testimonials (16 CFR Part 465)
Later, and only when the data exists — not on a calendar:
LaterGated on data volume or a peer-reviewable study, not on the calendar.
- Population-level accident-rate ratios from NTSB data for a named condition.Not claimed until: A peer-reviewable technical note
- Flights in higher index bands had higher rates of a named surrogate event.Not claimed until: Enough recorded events (at least 77-135) and validation on later flights
- Pilot-specific proficiency indicators associated with a named surrogate event.Not claimed until: Evidence from enough pilots, each with enough linked flights
Who Decides
FlyNook records information and decisions; it does not make them. The pilot in command is directly responsible for, and is the final authority as to, the operation of the aircraft (14 CFR 91.3). On a school flight, the school decides whether its aircraft is released under its own policy; once released, the pilot in command is the final authority.
FlyNook is not an official weather briefing, a dispatch or operational control system, or a substitute for flight instruction. Verify weather, NOTAMs, airworthiness and your minimums with official sources before every flight: read the terms
Changelog
- v1.0October 6, 2026
- First publication. Engine v7 (traffic is context, not weighted), bands risk-taxonomy-v1.
- v1.1October 13, 2026
- Critical findings (engine 2026.10.2): below VFR weather minimums, a convective SIGMET on the route and weight and balance outside limits each require review and hold the index at no lower than 60 (High). Weight and balance with no calculation is stated as not checked; the page no longer says it is shown ahead of the number.
- v1.2October 13, 2026
- Engine v10 (2026.10.5): a source that was not read, or whose read failed, is held at its not-assessed level, never scored as clear, and named on the verdict, the saved record and the PDF. The missing-factor guard counts a missing factor at that level, no longer at its most benign reading.
- v1.3October 14, 2026
- Relative Risk Index v1 (engine 2026.10.10): the index weighs six factors, the flight's conditions and aircraft; accident history is context, and the pilot's profile, minimums and fitness act through findings. One severe factor holds the index (0.8 times the worst factor). High needs an instructor's review and Review Required the chief pilot's, from the band alone; the 75-point rule is gone.