EDUKATE SINGAPORE · GOLD STANDARD PRACTICAL CASE STUDY 228
Five-metre positioning accuracy is an estimate with conditions, not a guarantee
A fictional school field-trip app shows a phone location with a five-metre horizontal accuracy radius. Its teacher interprets that as proof the student is always within five metres of the map dot. Under the specified Android Location API, the value instead represents an estimated 68%-confidence radius. Location and personal safety require more than a position marker.
All positions and students in this story are fictional. The accompanying walking illustration is not evidence of this field trip.
The accuracy claim at a glance
- The fictional app uses Android Location.getAccuracy, which is defined as a 68th-percentile horizontal confidence radius.
- Displayed radius: 5 metres; corresponding circle diameter: 10 metres.
- There is no guaranteed maximum error of five metres from that statistic.
- A circle overlapping a field-trip boundary cannot establish which side contains the true position.
- Signal conditions, device, mapping quality and age of the fix can affect usefulness.
- Real student supervision and privacy safeguards must not be replaced with an app’s confidence indicator.
Chapter navigation
- 1. A confidence radius is mistaken for a hard boundary
- 2. Start with the documented meaning of five metres
- 3. A probabilistic estimate is not a guarantee for the next observation
- 4. GPS signal accuracy and phone map accuracy are different
- 5. A circle around a point does not establish which side of a line a child occupies
- 6. Horizontal position is not floor level or elevation
- 7. Location age and movement can change the usefulness of the dot
- 8. Repeated measurements still need an appropriate validation method
- 9. Privacy and consent remain part of any field-trip tracking plan
- 10. Correct the public statement without dismissing the technology
- 11. A simple exercise makes the uncertainty visible
- 12. The bounded field-trip decision
1. A confidence radius is mistaken for a hard boundary
A fictional school geography group uses a mobile location application during an outdoor learning walk. One handset reports its position together with a horizontal accuracy estimate of five metres. The supervisor concludes that the student must be within exactly five metres of the dot on the digital map and suggests that a five-metre radius is therefore a dependable attendance or safety boundary.
The location dot is an estimate derived from positioning signals and device calculations. An accuracy estimate describes uncertainty under a provider’s defined method; it is not a physical fence around the device. A phone can fall outside its indicated uncertainty circle, and conditions may affect the reliability of the estimate.
In this case, the fictional application is explicitly assumed to report the Android platform’s Location.getAccuracy() value. Android defines this as an estimated horizontal accuracy radius at a 68th-percentile confidence level, not a guaranteed maximum positioning error. Different apps may use different definitions, so their labels must be checked individually.
The walk, pupils, devices, positions and decisions are invented. The featured illustration is contextual material about a self-guided walk, not a screen capture of the fictional app or evidence of a real student’s location.
2. Start with the documented meaning of five metres
The Android Location API reference states that getAccuracy() returns an estimated horizontal accuracy radius in metres at a 68% confidence level. It describes a circle around the reported position and a modelled chance that the true position lies within that circle.
A radius of five metres means five metres from the centre to its edge. The circle’s diameter is ten metres. That simple geometric distinction matters if a map draws the uncertainty as a coloured disc. Some readers see a large dot and assume the whole disc is only five metres wide.
More importantly, the 68% confidence interpretation is not a declaration that every reported position is within five metres. Under well-calibrated conditions, the provider’s uncertainty model allows an appreciable proportion of positional errors to exceed the radius. It is not a strict, worst-case safety bound.
The fictional school’s first evidence requirement is therefore to identify the exact API or app that produced the number. A phrase such as ‘five-metre accurate’ has insufficient precision unless the measurement and confidence convention are stated.
3. A probabilistic estimate is not a guarantee for the next observation
For a classroom probability exercise, imagine many independent, well-calibrated position estimates made under the same measurement-error model that assigns 68% containment to a reported radius. In a long series, approximately 68% of true positions would be expected to fall within their respective circles, while the rest may fall outside. It is not a promise that exactly 68 out of each consecutive 100 observations must fit.
An individual location reading cannot be validated from its own accuracy label alone. The estimate can be sensible even when the actual device lies farther away than the radius. An unusually poor fix, device settings or model mismatch could make uncertainty larger than the displayed value implies.
The teacher should not treat a student’s absence from an arbitrary five-metre geofence as proof of misconduct, nor a dot inside it as proof the student is physically safe at an exact spot. The numerical estimate informs interpretation but cannot replace ordinary field-trip supervision and direct communication.
The Gold Standard separates the meaning of a statistical confidence value from the certainty demanded by a consequential decision. Where the two do not match, the process must use more appropriate evidence.
4. GPS signal accuracy and phone map accuracy are different
The official GPS.gov accuracy explanation states that real positioning depends on satellite geometry, atmospheric conditions, receiver design, obstruction and reflections. It also distinguishes the accuracy of signals broadcast in space from actual user positioning accuracy.
A phone estimates a geographic coordinate from received signals and possibly other sources. Mapping software then draws that coordinate on a representation of streets, parks and buildings. The displayed dot can be wrong because the underlying coordinate has error, or because the map geometry or geocoding is imperfect, or both.
GPS.gov notes that buildings, bridges and trees can block or reflect signals and worsen the result. Open-sky conditions usually differ from sheltered courtyards or dense urban environments. Even the often cited approximate 4.9-metre smartphone accuracy under open sky is not a universal guarantee for all phones and places.
A school should avoid claiming the GPS satellites themselves promise that every device will lie within five metres of its dot. The system has several stages and the device-specific uncertainty belongs to the entire observation process.
5. A circle around a point does not establish which side of a line a child occupies
Imagine a fictional park map with a supervised activity boundary. The reported phone dot lies three metres inside the boundary, but the location uncertainty circle has a radius of five metres. That circle crosses the boundary. The reading does not establish that the device is definitely inside the permitted area.
One possible position within the circle is inside the line; another is outside. Because the confidence area itself can fail to contain the actual point, drawing a sharper line on the map cannot make the result certain. The appropriate conclusion is that the position estimate alone is insufficient to resolve the binary boundary decision.
Even if a reported dot lies farther inside the polygon than the stated radius, the estimate is still not a worst-case bound. Additional positioning, observation or supervised checking may be necessary where the decision has safety consequences. The geometry is informative but cannot turn a confidence radius into an absolute proof.
The Gold Standard treats a map boundary and an uncertain location fix as different inputs. Neither a coloured marker nor an app notification should be mistaken for direct knowledge of a student’s welfare.
6. Horizontal position is not floor level or elevation
Android’s documented getAccuracy() concerns horizontal positioning. It does not establish the accuracy of altitude. The platform exposes separate information for vertical accuracy when available, and some fixes may not contain every uncertainty field. The coordinate conventions and device documentation must be understood.
A student on a raised walkway or a lower path could share nearly the same horizontal coordinates with another person. A map that draws two positions together may conceal their vertical separation. A five-metre horizontal uncertainty estimate therefore cannot identify which floor or level the student occupies.
This also differs from the GNSS height-reference problem developed in an earlier Gold Standard case. A reported height may be referred to an ellipsoid or another vertical reference; vertical accuracy and datum interpretation are separate technical questions. Neither is resolved by the horizontal circle.
For the fictional field trip, staff should not use one horizontal coordinate to infer safe access to a platform or exit. The observation supports only what its stated variables measure.
7. Location age and movement can change the usefulness of the dot
A position obtained at a particular time may become less useful after the device moves. Some applications show a location fix from several seconds or minutes earlier, and its displayed marker can be drawn long after that fix. The accuracy figure applies to the position observation, not to all future motion.
If a pupil is walking, a five-metre radius attached to a stale coordinate is not the same as a real-time record of whereabouts. The supervisor must consider the location timestamp, update frequency and possible delays in app reporting. A good user interface makes the fix’s age visible and avoids presenting stale coordinates as immediate certainty.
Repeated fixes can also have correlated errors. Reflections from the same wall or an obstructed signal path may bias several successive readings in similar directions. Averaging many dots is not guaranteed to eliminate systematic error or make the uncertainty vanish.
The Gold Standard keeps time with the coordinate and its uncertainty. A snapshot cannot be transformed into continuous tracking merely because the map remains on the screen.
8. Repeated measurements still need an appropriate validation method
The class can collect repeated smartphone positions at a teacher-approved stationary reference point and compare them with a credible surveyed or otherwise suitably known location. The exercise should record conditions, fix times, device configuration and the provider’s stated uncertainty definition.
A cloud of points may reveal scatter, bias or occasional large errors. It does not automatically demonstrate that a reported radius is statistically calibrated unless the experimental design and reference quality are suitable. A cluster of readings near one location could even be systematically displaced from the true point.
The school should avoid claiming that the average error, a particular percentile and the maximum observed error are interchangeable. Those are distinct summaries and depend on sample size and environmental conditions.
Testing gives learners practical knowledge of uncertainty without turning an educational phone application into a surveying instrument or a safety-certified location service.
9. Privacy and consent remain part of any field-trip tracking plan
Location records can reveal personal movements, attendance patterns and sensitive information about students. The fictional school should therefore use only necessary data, follow applicable policies and approvals, control access, avoid unnecessary retention and explain clearly how any location information is used.
Technical accuracy is not sufficient justification to collect continuous location history. A field-trip organisation should select a proportionate supervision process, including staff headcounts, direct check-ins and authorised emergency procedures. A map may supplement rather than replace these measures.
A school should not publicly identify a child by sharing a live location link or publishing a set of coordinates in the name of proving that an app worked. Even an inaccurate record can create privacy risks if interpreted as an actual movement history.
The Gold Standard joins correct measurement with responsible governance. The right question is not merely whether a system can collect a coordinate but whether using it serves a legitimate need with suitable safeguards.
10. Correct the public statement without dismissing the technology
A defensible school note would read: ‘The mobile app returned a position estimate and a five-metre horizontal accuracy radius under its documented confidence convention. That radius is not a guaranteed maximum error, and the school’s supervision does not rely on the dot alone. Staff will verify attendance and whereabouts through established procedures.’
That wording recognises that satellite-based location is often useful while making its limits clear. It avoids the opposite exaggeration that GPS information is worthless. A coordinate with a documented uncertainty can support navigation, data collection and learning when used at an appropriate level of confidence.
The school should preserve the device’s model, provider, reading time and uncertainty definition in its teaching notes. If it corrects an earlier claim about guaranteed five-metre precision, the update should identify the evidence and avoid implying that the school had verified each student’s exact position.
Honest technology communication helps families understand what the school can and cannot know from a device.
11. A simple exercise makes the uncertainty visible
Give a group a paper map and a fictional location dot. Draw a circle with a five-centimetre radius on a scaled classroom diagram where one centimetre represents one metre. Ask students to describe the region associated with the estimate and to calculate the circle’s ten-metre diameter in the represented ground units.
Now place an activity boundary three metres from the centre of the circle. Students should recognise that the circle crosses the boundary, so the dot cannot definitively decide which side of the line contains the true position. An advanced group can explore the meaning of a 68% confidence level without pretending that every set of 100 readings must contain exactly 68 successes.
A second task compares a fresh fix in open sky with a stale fix near a wall. Learners identify possible causes of greater error, the age of the data and the limits of extrapolating a current position. The exercise can be completed with paper diagrams, not real child tracking.
The central learning outcome is that statistical uncertainty remains attached to the estimate even when software draws a crisp point on a map.
12. The bounded field-trip decision
The school’s application may report a five-metre horizontal accuracy estimate, but the number must be interpreted under its documented confidence convention. For the stipulated Android API, that is a 68th-percentile radius, not a guaranteed physical limit. Real accuracy also depends on the measurement environment and device.
Field-trip supervision should use appropriate direct methods, documented accountability, privacy safeguards and authorised incident procedures rather than trusting a dot or geofence as conclusive proof. A location fix can assist awareness while the exact whereabouts and welfare of a student require independent evidence.
The school can use the issue to teach coordinate systems, measurement error and the difference between a confidence statement and an absolute boundary. It should never convert modelled uncertainty into unqualified safety promises.
The Gold Standard conclusion is that ‘five-metre accuracy’ is not a universal guarantee. A useful position report includes its reference, probability meaning, time, environment and decision limits.
Practical location-evidence checklist
- Identify the exact location provider and confidence convention.
- Record the fix time, age, uncertainty radius and positioning environment.
- Do not confuse a horizontal radius with a guaranteed error or vertical precision.
- Test boundary decisions against the entire uncertain area, not just the dot.
- Use proportionate student supervision and minimise location-data collection.
- Preserve mapping and positioning errors as different possible causes.
Frequently asked questions
Does a five-metre accuracy label guarantee the true position is within five metres?
No. Under the stipulated Android API it represents an estimated 68%-confidence horizontal radius, not a maximum possible error.
Is the five metres a radius or diameter?
For Android’s getAccuracy value, it is a radius; the corresponding circle is ten metres across.
Does the value show which floor someone is on?
No. It is horizontal accuracy. Vertical position and its uncertainty are separate.
Why can accuracy become worse near tall buildings?
Obstruction and reflected satellite signals can affect the receiver’s location solution.
Can an app alone confirm a student is safely inside a field-trip boundary?
No. A probabilistic, possibly delayed position estimate is not a substitute for suitable direct supervision and authorised procedures.
Primary technical references and connected Gold Standard cases
- Android Developers · Location.getAccuracy Reference
- GPS.gov · GPS Accuracy
- NIST · How Do You Measure Your Location Using GPS?
- The Gold Standard of Geographic Information Science
- Gold Standard Case 206 · Ellipsoid height versus vertical datum
- Gold Standard Case 214 · Paper map scale and printed distance
- Gold Standard · Core Requirements and Evidence
- Gold Standard · Casebook
Android’s reference describes the particular API assumed in this fictional scenario. Not every mapping app uses the same accuracy convention. GPS.gov and NIST explain positioning limitations generally; no real student was tracked or located for this example.