Methodology
How driver scoring works
Roam's Drive tab turns a manually started session into a private, on-device coaching score. Nothing here is a safety system or a guarantee. It's feedback built entirely from measured GPS speed changes and phone motion.
What a drive session measures
GPS continuity, distance, speed, duration, and overlap with a planned practice route.
Measured changes in accepted GPS speed can identify rapid acceleration and hard braking events.
Course change and gravity-free Core Motion readings can identify sharper cornering when sensor quality is sufficient.
Qualifying miles can contribute to after-dark, faster-road, continuous-driving, and weekly progress totals.
How one drive is scored
Every drive starts at 100 and loses points for coaching events, normalized per 10 miles so a short trip and a long trip are judged fairly.
Deceleration ≥ 3.2 m/s² while moving 4+ m/s.
Acceleration ≥ 2.8 m/s² while moving 4+ m/s.
Course change ≥ 28°/s while moving 6+ m/s, with a reliable GPS heading.
A sustained acceleration-and-rotation pattern while the vehicle is moving. A single bump is ignored.
Penalties are capped at 55 points and the floor is 20, so one rough trip is never scored as a zero. GPS-derived events count on their own; motion data that corroborates a GPS event upgrades its provenance from “GPS speed” to “GPS + motion.”
Confidence tiers
A score is only as trustworthy as the data behind it, so every drive is labeled honestly.
At least 5 minutes of usable trace, 2+ miles, 24+ accepted GPS fixes, and 1,200+ motion samples.
At least 90 seconds of usable trace, 0.5+ miles, 8+ accepted GPS fixes, and 240+ motion samples.
Short or sparse drives, or drives where over half of delivered GPS fixes had to be rejected for accuracy.
The overall coaching score
Progress and Profile combine many drives into one evidence-weighted number, never a single trip.
Each qualifying drive's sensor score is nudged by that trip's independently analyzed route difficulty. A smooth drive on a demanding route counts a bit more than the same smoothness on an easy one.
Weight grows with the square root of measured miles, so one unusually long drive can't dominate a broad, consistent history.
Older drives fade on a 120-day half-life, so the score reflects how someone drives now, not only when they started.
Only medium- or high-confidence drives are included at all, and high-confidence drives count more than medium.
Fewer than 3 analyzed drives or 10 miles.
3+ drives and 10+ miles.
8+ drives and 30+ miles.
“Can I drive this?” route readiness
Before showing a verdict, Roam requires a real base of recorded experience, not just a good-looking score.
At least 3 qualifying drives on 3 different days, 15+ miles of validated trace, and 45+ minutes of continuous driving before any route gets compared at all.
Each of the route's 8 demand categories (after-dark, fast roads, merges, intersections, weather, sustained drive, traffic, road conditions) is checked against matching recorded exposure. After-dark miles are compared with after-dark miles, not a single overall grade.
On-device GPS overlap with saved, directionally aligned traces contributes to, but never replaces, the demand comparison. Parallel roads and long gaps are excluded.
Privacy
Manual driving scores stay on the device. Only after a completed drive with a continuous trace does Roam send the measured start and end coordinates, never an address or the recorded route geometry, to the route-analysis service, and it stores only the resulting compact difficulty snapshot locally.
A driver's name and licensing stage are self-declared and never change a route score or a driving score.
Important limits
Roam is a planning and coaching tool, not a safety guarantee.
A drive score is coaching feedback, not proof a person or route is safe.
The phone-motion detector never identifies handheld use. It only flags a sustained, abrupt movement pattern.
Follow local laws, license restrictions, supervision rules, and your own judgment.