Turn real flight data
into better pilots.

The same mistake shows up on the controls in a handful of different ways, and “fly it again” does not say which one it was. Every measurement carries its evidence: which runway, which side of the centreline, the median beside the maximum, the sampling rate — and what these data cannot show.

A morning of circuit training. Each bar is the lateral offset from the runway centreline, in metres, circuit by circuit. Nobody typed these in — they are computed from the track.

The circuit tightened through the session — 2487 m down to 1600 m by the seventh. The student was improving, and it shows before anybody writes a word about it. The eighth, at 1060 m, is tight the other way: that is a conversation too, and the instructor is the one who has it.

A pre-flight risk assessment on your own sheet. Each row carries its origin, because after an occurrence the question is “on what basis”, not “what did it total”.

The number of assumptions travels with the result, because missing data always understates risk in the same direction — towards permission to fly. A green built on guesses does not mean what a green built on measurements means.

The measurement is always recorded. A verdict only where a threshold exists, and the head of training sets it. Without one you see the number and a plain statement that there is nothing to compare it against — because how wide a circuit should be at your airfield is your school’s call, not ours.

An instructor does not type metres. They click “that was a good circuit” next to a measurement, and the band comes from the spread of the marked examples — widened by the uncertainty of the measurement itself, not by a percentage pulled from the air.

Flights arrive from the ground receiver or from an onboard recorder — SkyBox, with map replay and a 3D view at log.skyradio.aero. Both write to the same tables: the source is an attribute of a flight, not an architectural assumption. Each flight also states what cannot be read from it — at 0.2 Hz from ADS-B, slip and flare are not observable, and it says so rather than showing the same confidence everywhere.