A forecast should not call itself accurate simply because it uses reputable inputs. CheckFlightComfort.com has begun saving anonymous forecast snapshots so future claims can be tested against a documented comparison method.

FORECAST SNAPSHOTSCollecting

Saved without passenger identity

ROUTE PAIRSCollecting

Distinct airport combinations observed

PUBLIC ACCURACY RATENot published yet

Requires a documented minimum sample and comparison method

First coverage report: September 21, 2026

This first baseline describes what the collection system stored from August 7 through September 18, 2026. It is a coverage report, not an accuracy report. The rows include repeated checks and internal or owner testing. They are not unique passengers, unique flights, or a representative sample of airline traffic.

Baseline measureObserved countWhat it does and does not mean
Saved snapshots

64

Successful route forecasts stored during the period. Repeated checks are separate snapshots.

Directional route pairs

20

Distinct departure and arrival combinations. A route in the opposite direction counts separately.

Confidence labels

31 moderate and 33 limited

Coverage and timing labels at the moment of each check, not a grade for the final ride.

Estimated routes touching United States coverage

56

The estimated path overlapped part of the site's United States advisory coverage. It does not mean every route segment had equal detail.

Routes labeled limited outside the United States

8

The site disclosed that advisory coverage could be more limited on those international paths.

Snapshots with at least one route advisory

24

An advisory intersected the estimated route window. It is not proof that the aircraft entered the area.

Snapshots with at least one nearby pilot report

39

A recent report was found near the estimated route. Reports outside the near departure window were context only.

What the baseline reveals

The collection pipeline is working across domestic, oceanic, and international examples, and it preserves confidence and coverage limitations beside the score. It also reveals a serious sampling imbalance: LAX to JFK and LAX to Cancun account for 31 of the 64 snapshots. That concentration is useful for testing repeated changes on familiar routes, but it is not broad enough for a sitewide accuracy statistic.

The baseline also lacks a consistent independent observation of how each completed flight actually felt. A pilot report near a route is not automatically a match for the checked flight, and no pilot report is not proof of smooth air. The current dataset can measure whether the forecast system stored its inputs and limitations. It cannot yet establish how often the comfort label matched the passenger experience.

Current conclusionCollection is functioning, geographic coverage is uneven, and the sample is too concentrated and incomplete for a defensible accuracy percentage. The honest public status remains collecting baseline.

What a snapshot contains

  • Departure and arrival airport codes
  • Departure day and optional flight number
  • Comfort score and strongest route signal
  • Counts of relevant advisories and nearby positive pilot reports
  • Available signal count, confidence label, and coverage summary

No passenger name, email address, payment information, or precise device location is stored in the verification snapshot.

Why an accuracy percentage is not shown yet

A small or biased sample could create a reassuring looking number that means very little. Before publishing an accuracy rate, the service needs enough routes, a defined observation window, rules for missing reports, and separate treatment of turbulence advisories, airport weather, and actual ride reports.

Planned comparison principles

  1. Freeze the prediction: use the forecast that existed at a defined lead time.
  2. Define the observation window: compare only reports reasonably near the estimated route, altitude, and flight period.
  3. Separate absence from evidence: no pilot report does not automatically mean smooth air.
  4. Publish false alarms and misses: report both directions instead of selecting flattering examples.
  5. Segment the results: show lead time and geographic coverage so unlike forecasts are not mixed together.

What will be published

Once the dataset and rules are sufficient, this page is intended to report sample size, route coverage, lead time groups, signal availability, false alarm frequency, missed stronger signals, and clearly stated limitations. Until then, the honest result is “collecting baseline.”

Read the complete forecast methodology, learn why a forecast can change, or create a current flight forecast.