How accurate is the location check?
What phone location can and cannot do, why no web app can reach ±0.1 m, and how Presence combines several checks to place a student in the right room.
4 min read
The short answer
No phone browser, and no ordinary phone app, can locate a person to within ±0.1 m (10 cm), or even ±1 m, indoors. Presence does not rely on one precise position. It combines several independent checks, each of which a student would have to beat separately:
- Rotating join code shown on the board (changes every 10-120 seconds, 30 s by default).
- Classroom and campus location checks: the phone checks the radius around the teacher’s phone on every GPS reading and reports every minute (every two once settled in), and at once when it leaves; the server re-checks every report.
- Room sound check (optional): a near-silent signal from the teacher’s device that walls and floors block. It tells apart rooms stacked on different floors. See Classrooms on different floors.
- Campus network check (optional): the phone must be on the campus Wi-Fi.
- One registered phone per student: attendance is tied to the phone the student first marked attendance from.
- Signed location updates and surprise presence checks during the lecture.
- Face verification (optional).
What phones can actually do
| Technology | Typical accuracy | Notes |
|---|---|---|
| Phone GPS, outdoors | about 3-5 m | Median errors under 5 m in open sky [1] |
| Phone GPS, indoors | about 10-50 m | Walls and ceilings reflect the signal; 28-48 m indoor errors are reported [1] |
| GPS altitude | too unreliable to tell floors apart | Error is larger than one storey |
| Wi-Fi round-trip time (RTT) | 1-2 m | Needs a native Android app and compatible access points; not available to web apps [2][3] |
| Dual-frequency GNSS with correction services | under 1 m | Needs open sky, minutes of data and post-processing [4] |
| Survey-grade or UWB equipment | centimetres | Needs dedicated hardware in every room |
What the browser’s “accuracy” means. The accuracy a browser reports is the radius of a circle that contains the true position with 95% confidence [5]. A reading of “20 m” therefore means “within 20 m, 19 times out of 20”, not “exactly 20 m off”. Presence uses this definition throughout: its position filter converts the radius into the matching statistical error before weighting each reading.
Why a sound signal works for rooms
Ultrasound and near-ultrasound do not pass well through walls and floors, so “can this phone hear the signal?” is a reliable room-level test. Research systems have used this to place devices in the right room [6][7]. Presence’s room sound check sends a short code that changes every 30 seconds, so a recording played later does not work.
Why the campus network check helps
A fake-location app can change what the phone reports as its position, but it cannot make the phone connect from the campus internet connection. When the network check is set to Required, attendance is accepted only from the institution’s own address ranges. To fill them in, connect to the campus Wi-Fi and tap Add the network I am on now in Settings, or ask your IT team for the ranges. Start with Advisory for a week to confirm the ranges are right.
How Presence handles imprecise GPS
- Teacher capture: up to 15 fresh readings from the teacher’s own phone, each weighted by its accuracy, with outliers removed. Indoors this is typically 10-30 m (better near a window). By default students must be within the radius around this position.
- Student phones: old (cached) positions and coarse network guesses are ignored. Readings are smoothed with a filter that ignores one-off jumps but follows a real walk out of the room within a few readings. A student can mark present only once the position has settled (a few readings, or one sharp one).
- Three possible results: inside, uncertain (weak indoor GPS - the student’s time keeps counting) and outside (beyond the classroom area by more than the combined error of the student’s and the teacher’s GPS). A student is checked out only after about 45 seconds clearly outside.
- Stale positions never count: if a phone stops getting fresh GPS fixes (for example a locked screen), its last position is not counted as inside.
- Light on the servers: each phone checks the classroom area itself for every GPS reading and reports every minute (every two once steadily inside), straight away when it leaves.
- Honest accuracy: Presence never reports a position as more precise than the phone’s hardware allows.
Sources
- Study of smartphone GNSS accuracy outdoors and indoors, arXiv 2603.26706 - arxiv.org/pdf/2603.26706
- GPS World, How to achieve 1-meter accuracy in Android - gpsworld.com
- Crowd Connected, Testing Wi-Fi RTT on Android P for indoor positioning - crowdconnected.com
- Study of dual-frequency smartphone positioning with correction services (PubMed Central PMC9788430) - ncbi.nlm.nih.gov/pmc/articles/PMC9788430
- MDN Web Docs, GeolocationCoordinates: accuracy - developer.mozilla.org
- Borriello et al., WALRUS: wireless acoustic location with room-level resolution using ultrasound, MobiSys 2005 - usenix.org
- SoundLoc, acoustic room-level indoor localization, arXiv 1407.4409 - arxiv.org/pdf/1407.4409