Location tracking is the collection, storage and analysis of data that reveals where a person or device is, either at a single moment or as a trajectory over time. In privacy, it is treated as one of the most sensitive forms of tracking because a sequence of positions exposes home and work addresses, routines, relationships, health visits and religious practice without the person ever stating any of them.
Location data reaches companies and authorities through device sensors, network infrastructure, apps, advertising software and connected vehicles. Privacy law, platform permission systems and privacy engineering techniques all aim to limit who can collect it, how precisely, for how long and for which purpose.
How location tracking determines position
Location tracking works by combining positioning signals from several sources, each with its own accuracy and its own party that sees the data. A smartphone typically fuses satellite positioning, Wi-Fi scans, cell tower identifiers and Bluetooth beacons into one estimate, and the operating system then passes that estimate to apps that hold the relevant permission.
The table below compares the main positioning methods by typical accuracy and by who can observe the signal, which determines where privacy controls must apply.
| Method | Typical accuracy | Who observes it | Works without user action |
|---|---|---|---|
| GNSS (GPS, Galileo) | 3 to 10 meters outdoors | The device, then any app granted access | No, requires an app or OS service |
| Wi-Fi positioning | 10 to 50 meters | The OS vendor and location database operators | Partly, through background scans |
| Cell tower data | 100 meters to several kilometers | The mobile network operator | Yes, always while connected |
| Bluetooth beacons | 1 to 5 meters indoors | Venue operators and SDKs in apps | Partly, if scanning is enabled |
| IP geolocation | City or region level | Every website and online service | Yes, on every connection |
Accuracy alone does not define the risk. Network-side data such as cell tower records is generated whether or not the user opens an app, which is why it is often the basis of lawful access requests and why device-level permission settings cannot fully prevent it.
Why location data is so revealing
Location data is highly identifying because human movement is regular and unique. Research on mobility traces has repeatedly shown that a handful of time-stamped points, often four or fewer, is enough to single out most individuals in a large dataset, even when names and device identifiers have been removed.
From repeated positions an analyst can infer sensitive attributes that are often protected as special categories of data under the GDPR and similar laws. Visits to a clinic suggest health conditions, nightly stays reveal the home, and co-location with another device over weeks reveals relationships. This is why regulators increasingly treat precise location as sensitive data in its own right, as the CPRA in California explicitly does for precise geolocation within a radius of about 1,850 feet.
- Home and workplace inference from overnight and weekday dwell times
- Health and religion inference from visits to clinics, pharmacies or places of worship
- Social graph inference from repeated co-location with other devices
- Behavioral profiling for advertising from shop visits and travel patterns
Location tracking compared with related tracking techniques
Location tracking differs from other techniques in the Tracking family because it follows the person through physical space rather than through websites or apps. It often works together with them, since an advertising ID or a device fingerprint is what links location points to a single profile.
| Technique | What it follows | Typical identifier | Relation to location |
|---|---|---|---|
| Location tracking | Physical movement | Device or advertising ID | Core subject |
| Web tracking | Browsing activity | Cookies and scripts | Adds IP-based location |
| Device fingerprinting | Device configuration | Derived fingerprint | Links location across apps |
| Advertising ID | App usage | Resettable mobile ID | Main key in location data markets |
| Behavioral advertising | Interests and habits | Profile segments | Uses visit history for targeting |
A particular concern is the location data broker market. Software development kits embedded in ordinary apps collect positions, attach the advertising ID and sell the stream onward, so a user who granted location access to a weather app may unknowingly feed dozens of downstream buyers.
Legal and platform controls on location data
Location tracking is governed by consent requirements, purpose limitation and platform permission models that operate in parallel. In the European Union, the ePrivacy Directive requires consent for processing location data from electronic communications beyond what is needed to deliver the service, and the GDPR requires a lawful basis and data minimization for all other location processing.
Mobile operating systems add technical controls that narrow what apps can obtain. The following table summarizes the main levers and what each protects against.
| Control | Mechanism | What it limits |
|---|---|---|
| Foreground-only permission | Access only while the app is open | Continuous background trails |
| Approximate location | OS returns a coarse area, about 1 to 3 km | Address-level inference |
| One-time permission | Access ends when the app closes | Long-term silent collection |
| Wi-Fi and Bluetooth scan controls | Disables background radio scanning | Indoor and beacon tracking |
| Consent under ePrivacy and GDPR | Legal requirement before collection | Unlawful secondary use |
Enforcement has targeted brokers that sold visits to sensitive locations, and several authorities now regard the sale of precise location without clear, specific consent as an unfair practice. A consent screen buried in terms of service is generally not considered valid consent for this purpose.
Privacy engineering for location data
Privacy engineering reduces location risk by collecting less, coarsening what is collected and deleting it early. Simply removing names does not work, because trajectories re-identify people; effective protection changes the data itself.
| Technique | How it works | Main trade-off |
|---|---|---|
| Spatial generalization | Rounds positions to grid cells or postal areas | Loses street-level utility |
| Temporal aggregation | Reports counts per hour or day instead of points | Hides individual journeys |
| Differential privacy | Adds calibrated noise to aggregate statistics | Reduces accuracy for small areas |
| On-device processing | Keeps raw positions on the phone | Limits server-side analytics |
| Short retention | Deletes raw traces after days, not years | Restricts historical analysis |
The most common mistake is to treat hashed device IDs or truncated coordinates as anonymization. Hashing keeps records linkable, and three decimal places of latitude still locate a person within roughly 100 meters. A sound design starts from the purpose, asks what minimum precision and retention serve it, and documents that decision in a data protection impact assessment.
Stripped to the essentials
Location tracking means keeping a record of where you go. Your phone knows its position from satellites, nearby Wi-Fi networks and phone masts, and apps that you allow can see that position too. Some of them pass it on to other companies.
One location point says little, but a trail of points tells a story: where you sleep, where you work, which doctor you visit and who you spend time with. That is why location history is treated as very sensitive. The best protection is to give apps location access only while you use them, choose approximate location when exact position is not needed, and remove access from apps that have no clear reason to know where you are.
Where it is applied
A city transport authority wants to know how commuters move between districts so it can redesign bus routes. A vendor offers raw, per-device location trails bought from app SDKs. The authority’s data protection officer rejects the offer because the trails would allow staff to identify individual residents, and the source consent is doubtful.
Instead, the authority asks the mobile network operator for aggregated origin-destination counts: trips between 1-kilometer cells, grouped by hour, with any cell pair under 20 trips suppressed and differential privacy noise added. No device identifiers leave the operator. The planners still see that morning demand between two suburbs is twice the current capacity, and they add an express line. The result is useful mobility insight without building a database of individual movements, documented in an impact assessment that the authority publishes with the route plan.