Rider handbook

What the numbers mean

The Summary cover shows five figures, and three of them carry a small verdict marker. This page says what each one measures, which way is better, where the bands sit, and whose expectation they are. It reads in about five minutes — or thirty seconds, if the legend below is all you came for.

Targets, not measurementsOwner-declaredWithheld, never guessed

What the markers mean

Five states, and only three of them are a colour. This is the app's whole verdict vocabulary.

Good
The figure sits inside the target set for it.
Watch
Between the two edges. Worth a look, not a problem.
Off target
Past the far edge of the target.
No verdict
This session cannot support a judgement here, and the marker says why. Withheld is not the same as good, and it is deliberately not a fourth colour.
No marker at all
The app deliberately sets no target for that figure. A lap time is neither good nor bad on its own.

The bands behind those three words are targets the app's author declared, not optima measured across riders. That distinction is the whole of the next section.

These are targets, not measured optima

Every band on this page is a declared target, not something discovered by measuring lots of riders and finding a "good" range. Nobody has established a scientific optimum for, say, lap-time consistency across trackday riders on consumer GPS loggers — that measurement does not exist. What exists is the owner's own judgement call about what a rider on this app should aim for, written down as a number so a colour on the screen means something instead of nothing.

That is a deliberate, named distinction (see ADR 0031 in the repository if you want the engineering reasoning): a band edge is a criterion applied to a value, not a correction to the value itself. Nothing about your recorded lap times, speeds, or lean angles changes because of a band. The band only decides which word — Good, Watch, or Off target — gets attached to a number that was already there.

Because they are targets and not facts, they are revisable. If a band looks wrong for the way you ride, that is worth raising — it is one edit in one file, not a re-measurement of physics.

The five cover figures

Every number the Summary cover shows, in the order it appears. The three rated figures below quote their bands exactly as declared in src/gpx/kpiJudgement.ts; the two unrated figures are listed too, with the reason no verdict is offered.

Best lap KPI-002 · unrated

Your fastest complete, valid lap in the session.

No verdict marker. A lap time is neither good nor bad in the abstract — it is the outcome the other four figures try to explain. Colouring it would make colour mean nothing.

Gap to ideal KPI-015 · lower is better

How much slower your best lap is than the lap you would get by joining together the best sector splits you already demonstrated in this same session. It never compares you to a theoretical or another rider's line — only to yourself.

Good
under 0.5 s
Watch
0.5 s to 1.5 s
Off target
1.5 s or more

Owner target set 2026-07-25 against the bundled fixture range (0.12 s–2.21 s of demonstrated gap). Under half a second means the sectors were basically assembled into one lap; past a second and a half, the pace exists but has not come together yet. No data-capability floor — this works on any session that produces laps.

Consistency KPI-023 · lower is better

Lap-time coefficient of variation: how much your lap times vary relative to their own average, as a percentage. Normalizing by the average means the same bands apply at a short club circuit and a long fast one.

Good
under 1.5 %
Watch
1.5 % to 3 %
Off target
3 % or more

Owner target set 2026-07-25 against the bundled fixture range (0.07 %–4.64 % across sessions). No data-capability floor.

Neutral motion (inferred) KPI-107 · lower is better

The share of cornering time where the derived longitudinal acceleration sits near zero while you are leaned over — not a measurement of the throttle or the brake lever. Neither Diablo nor RaceBox records those inputs, so this is inferred from GPS speed and lean, and is named "inferred" everywhere it appears for that reason.

Good
under 5 %
Watch
5 % to 10 %
Off target
10 % or more

Owner's own edges, set 2026-07-25. Capability floor: full. A roughly 1 Hz export can still produce a share, but it is too coarse a measurement to hang a verdict on — below that rate you get the ⚠ "no verdict" marker with an explanation, never a colour standing in for a guess.

Valid laps KPI-029 · unrated

How many complete laps in the session were valid for timing and comparison.

No verdict marker. A lap count is context for reading the other figures, not a performance outcome by itself.

Figures without a verdict

Open a coaching card in Coaching or on the Summary cover and choose "Show evidence", and you will see a few more numbers alongside the five above. These live inside the supporting detail for one piece of advice, not on the cover itself. The app deliberately sets no target for them — doing so would mean declaring a band nobody has actually decided on — so none of them carries a colour, a tick, or a `!`. No marker at all is the honest state here, the same way it is for best lap and valid laps.

What the app deliberately does not claim

Neither supported device records brake pressure or throttle position, so neutral motion is always inferred from speed and lean, never measured directly. A verdict on it says something about coasting time, nothing about what your hands or feet were doing.

On a low-frequency session — below the 5 Hz floor the neutral-motion figure needs — the app withholds that verdict instead of colouring a guess. You will see a marker with the reason on hover, which is a different thing from a green tick: withheld is not the same as good.

For the full engineering explanation of how each derived number is computed and where it stops being trustworthy, see How it works.