---
title: "How to Measure Tracking Accuracy Against Ground Truth: A Test Method for GPS, Driver App, Carrier API and Scan Data"
id: "27586"
type: "post"
slug: "tracking-accuracy-ground-truth-test-method"
published_at: "2026-10-22T15:00:00+00:00"
modified_at: "2026-10-22T14:16:07+00:00"
url: "https://locus.sh/blogs/tracking-accuracy-ground-truth-test-method/"
markdown_url: "https://locus.sh/blogs/tracking-accuracy-ground-truth-test-method.md"
excerpt: "How to measure tracking accuracy against ground truth: pick independent evidence, size the sample, compare GPS, driver app, carrier API and scan data on error, latency, coverage and ETA."
taxonomy_category:
  - "General"
---

#### [General](https://locus.sh/blogs/category/general/)

# How to Measure Tracking Accuracy Against Ground Truth: A Test Method for GPS, Driver App, Carrier API and Scan Data

[Aseem Sinha](/author/aseem_locus/)

Oct 22, 2026

10 mins read

Tracking accuracy testing means comparing what a tracking source reports, such as a position, a status time or an ETA, with an independent record of what actually happened, called ground truth, so that error, delay and coverage can be measured for each source. Vendors quote accuracy figures that rarely say where, when or on what data they were measured, and a source that is accurate in an open suburb can fail between tall buildings. A defensible test uses independent evidence, a sample large enough to bound the error, and metrics that separate how wrong a source is, how late it is and how often it reports at all.

## Key Takeaways

- GPS.gov says GPS-enabled smartphones are typically accurate to within 4.9 meters under open sky and that accuracy worsens near buildings, bridges and trees, so claims need testing on your own routes.
- Android’s geofencing guidance recommends a radius of at least 100 meters and says events usually arrive in under two minutes, so a geofence is not an instant.
- A US Postal Service Inspector General audit found that 33% of 25.5 million sampled delivery scans in 2016 had no corresponding location data, mostly because of GPS signal obstruction.
- Effective ETA accuracy is coverage multiplied by accuracy: a source that tracks 60% of shipments at 90% accuracy delivers 54%.
- Locus states a first-party GPS feed from its driver app, but its public pages do not state a ping interval or an ETA accuracy figure.

## Why Tracking Accuracy Matters: The Business Case

Every exception rule, ETA and customer message is only as good as the position and time behind it.

- GPS.gov states that [GPS-enabled smartphones are typically accurate](http://www.gps.gov/gps-accuracy) to within a 4.9 meter radius under open sky and that accuracy worsens near buildings, bridges and trees.
- The US Postal Service Inspector General’s 2017 audit found that [33% of 25.5 million delivery scans](https://www.uspsoig.gov/es/node/2188) in July to December 2016 had no corresponding location data, which it attributed mainly to GPS signal obstruction.
- FedEx’s 2026 survey of 700 logistics leaders found that [only 18% can always intervene](https://digital-blog.fedex.com/3-key-insights-from-the-future-of-logistics-intelligence-report) when a delivery is delayed, and 97% say visibility alone is not enough. It is a carrier-run survey.
- A 2026 Loadstar guest article, based on 25 interviews, described an operator whose [ETA accuracy reached 90%](https://theloadstar.com/what-supply-chain-visibility-gets-wrong-about-eta-numbers-and-why-it-matters/) on trackable shipments but whose effective accuracy was 54% because only 60% of volume was trackable. It is anecdotal.

A visibility number without a coverage number and a test method is a claim and not a measurement.

| Also Read: Real-Time Visibility: The Latency Math Nobody Publishes |
| --- |

## Why Locus Leads for Verifiable Tracking

Tracking can only be trusted when its source is known and testable. Three things set Locus apart.

**Recognized by analysts and selected after evaluation.** Locus was recognized by Gartner for seven consecutive years and appears in the 2026 Hype Cycle for Supply Chain Execution and Logistics Technologies, as detailed on its [analyst recognition page](https://locus.sh/analyst-recognition/)
. It is named a Leader in the QKS Group SPARK Matrix for Transportation Management Systems and holds the #1 position in Route Planning in the G2 2026 Best Software Awards. In October 2025, Ingka Investments, the investment arm of Ingka Group, the world’s largest IKEA retailer, acquired Locus. Ingka’s [press release](https://locus.sh/press-releases/ingka-group-acquires-locus/)
 says that after an extensive evaluation of logistics software worldwide, it selected Locus for its strength in combining advanced route optimization, real-time visibility and smart use of vehicles and resources. Locus continues to operate independently.

**A first-party position source.** Locus’s Driver Companion App provides a first-party position feed on owned and contracted movements, so position does not depend on a carrier relay. Its [Control Tower](https://locus.sh/control-tower-software/)
 raises geofence breach alerts for route deviations and halts, and consolidates owned, contracted and 1,000-plus carrier movements, with ShipFlex integrating 160+ active carriers.

**An ETA built on live execution.** Locus’s ETA model uses live GPS, stop completion, actual service times and traffic, and notifies when deviation exceeds a configured threshold. Locus has optimized 1.5B+ deliveries for 360+ enterprise customers across 30+ countries.

Locus says it is not a telematics provider and does not set the reporting interval on a third-party carrier’s hardware, and its public pages do not state a ping interval, an ETA accuracy figure or an independent accuracy test, so ask for the connected feeds, refresh interval and the method behind any accuracy claim.

| Also Read: Why Real-Time Visibility Fails: The Data Problem 2026 |
| --- |

## How to Measure Tracking Accuracy Against Ground Truth

The six steps below make a vendor-neutral test you can repeat.

### 1. Define what you are testing

Separate four things: position error, event time error, ETA error and coverage. A source can be precise and sparse, or dense and late, and one blended score hides which.

### 2. Choose ground truth that is independent of the source

Ground truth must not come from the system under test. Options include a calibrated GNSS logger carried in the vehicle, time-stamped proof-of-delivery photos, driver-confirmed arrival times recorded separately and fixed reference points such as depot gates. Record how each was captured.

### 3. Design the sample

Stratify by source, geography, such as dense urban and rural, time of day and weather. For a proportion such as the share of stops tracked within a window, a sample of about 385 gives a margin of plus or minus five percentage points at 95% confidence.

### 4. Collect paired records

For each test stop, record the source’s report and the ground truth side by side with timestamps, including cases where the source reported nothing. Missing reports are data, and dropping them inflates accuracy.

### 5. Compute the metrics

Report the median and 95th percentile of position error, because NIST defines the percentile as the value below which at most that share of measurements fall. Report latency as the gap between the true event and its appearance in your system. Report coverage as the share of stops with usable data. For ETA, report mean absolute error and the within-window rate, which you define yourself, such as the share of ETAs within 15 minutes of actual arrival.

### 6. Report by segment and set thresholds

Publish results by source and segment, set acceptance thresholds tied to decisions, such as a geofence radius or an alert lead time, and repeat the test on a schedule, since devices, carriers and networks change.

| Also Read: Delivery ETA Accuracy: How AI Predictions Outperform Estimates |
| --- |

## Driver App GPS, Telematics, Carrier API and Scan Data: Key Differences

| Source | What it provides | Typical limit | What to test |
| --- | --- | --- | --- |
| Driver app GPS | Position from the phone’s location services, continuous while the app runs | Accuracy and update rate depend on device, permissions, battery settings and surroundings | Error by environment, update gaps, behavior when the app is backgrounded |
| Telematics device | Position and vehicle data from fitted hardware | Reporting interval is set by the device or provider, and only fitted vehicles report | Coverage across the fleet, interval, latency to your system |
| Carrier API | Status and sometimes position relayed by the carrier | Depends on the carrier’s own feed and update frequency | Coverage by carrier, event delay, field consistency |
| Scan events | A status at a scan point, often with a timestamp | Location only when scanned, and may be missing or batched | Share of scans with location, scan-to-system delay |

## What to Look for in Tracking Software and Its Accuracy Claims

### The source of every position and status

Ask where each position and status comes from and how often it updates. A map pin from a carrier relay and a pin from a first-party app are different things.

### A coverage number next to the accuracy number

Look for the share of shipments or stops actually tracked. A 95% accurate source on 60% of volume is 57% effective.

### A documented test method

Check whether the vendor can describe how accuracy was measured, on what sample, in what environments and against what ground truth.

### Latency as well as error

Ask for the delay between an event and its appearance, and for geofence settings. Android advises a radius of at least 100 meters because location accuracy varies.

### Segment reporting

Look for results by carrier, geography and time of day. An average hides the segments that fail.

| Also Read: Real-Time Location Tracking Architecture for Last-Mile |
| --- |

## Tracking Accuracy Testing in Action: Real-World Results

**A worked example on illustrative results.** The figures are assumptions for a test of 400 stops per source, so replace them with your own. Effective ETA accuracy is coverage multiplied by the within-window rate, where within-window means an ETA within 15 minutes of actual arrival.

| Source | Coverage | Median position error | 95th percentile error | Median event delay | ETA within window | Effective ETA accuracy |
| --- | --- | --- | --- | --- | --- | --- |
| Driver app GPS | 96% | 8 m | 45 m | 0.5 min | 88% | 84.5% |
| Telematics device | 70% | 6 m | 30 m | 1 min | 90% | 63.0% |
| Carrier API | 60% | Not applicable | Not applicable | 38 min | 90% | 54.0% |
| Scan events | 67% | Not applicable | Not applicable | 25 min | 85% | 57.0% |

The carrier API looks as good as the telematics device on ETA accuracy, 90% against 90%, until coverage is applied, and then it delivers 54% against 63%. With 400 stops per source, the margin on each proportion is about plus or minus 4.9 percentage points at 95% confidence, so a gap of two or three points between sources is not evidence. The test also shows where a hybrid design helps: use first-party position where it exists and fall back to carrier events only for the shipments that have nothing else.

**An ASEAN apparel retailer.** The retailer had carriers that each reported their own status codes and no trustworthy checkout date. With one harmonized status standard and a network-aware delivery date at checkout, WISMO and returns queries fell by more than 40% and carrier onboarding dropped from 3 months to 3 days. [Read the apparel case study](https://locus.sh/case-studies/apparel-multi-carrier-parcel-management/)
.

**A North American retailer.** The retailer consolidated six systems into one and cut manual dispatch by more than 80%, with $1M+ in savings, 99%+ on-time delivery and exceptions resolved in under two hours, while its ERP and WMS stayed the systems of record. [Read the retailer case study](https://locus.sh/case-studies/retailer-multimodal-logistics-automation/)
.

## Common Tracking Accuracy Mistakes to Avoid

**Testing against the system’s own data.** Ground truth must be independent, or the test only measures consistency.

**Reporting accuracy without coverage.** A precise source that sees 60% of volume leaves 40% blind, so report both.

**Testing only in easy conditions.** Open suburbs flatter GPS, so include dense urban areas and poor weather.

**Dropping missing reports.** A source that says nothing has failed that stop, so count it.

## How Locus Approaches Tracking You Can Verify

Locus, the world’s first Decision-Intelligent, Agentic TMS, gets first-party position from its Driver Companion App on owned and contracted movements, raises geofence and route-deviation alerts in its Control Tower and consolidates carrier movements from its integrated carriers, with its ETA model recalculating from live GPS, completed stops, actual service times and traffic. Its [tracking page](https://locus.sh/shipment-tracking-system/)
 shares the live status with customers, and autonomy levels, explainability and an execution sandbox govern what the agents do with it. Locus’s public pages do not state a ping interval, an ETA accuracy figure or an independent accuracy test, so ask for the connected feeds, refresh intervals and a joint test on your own routes. To run a ground-truth test with Locus data, [schedule a demo](https://locus.sh/schedule-demo/)
.

For logistics leaders asking how to measure tracking accuracy against ground truth, Locus offers a first-party position feed and live execution data that can be tested on your routes, so accuracy is a result you measured and not a claim you were given.

## Frequently Asked Questions

**How do you measure tracking accuracy?** Compare each source’s reported position, event time and ETA with independent ground truth on a stratified sample, and report median and 95th percentile error, latency and coverage.

**What is ground truth in logistics tracking?** Independent evidence of what actually happened, such as a calibrated GNSS logger, time-stamped proof-of-delivery photos or fixed reference points, not data from the system being tested.

**How accurate is smartphone GPS?** GPS.gov says GPS-enabled smartphones are typically accurate to within 4.9 meters under open sky, and that accuracy worsens near buildings, bridges and trees.

**How many stops do you need to test?** For a proportion, about 385 give a margin of plus or minus five percentage points at 95% confidence. Use more where you need to compare sources closely.

**What is effective ETA accuracy?** Coverage multiplied by the within-window rate. A source that tracks 60% of shipments with 90% accurate ETAs delivers 54% effective accuracy.

**Does Locus publish tracking accuracy figures?** Locus states a first-party driver-app position feed and an ETA model on live execution. Its public pages do not state a ping interval or an ETA accuracy figure, so ask for a joint test.

MEET THE AUTHOR

Aseem Sinha

Vice President - Marketing

Aseem, leads Marketing at Locus. He has more than two decades of experience in executing global brand, product, and growth marketing strategies across the US, Europe, SEA, MEA, and India.

### Related Tags:

[https://locus.sh/blogs/prove-tms-value-90-days-baselines-control-groups/](https://locus.sh/blogs/prove-tms-value-90-days-baselines-control-groups/)
#### [General](https://locus.sh/blogs/category/general/)

## [How to Prove TMS Value in 90 Days: Baselines, Control Groups and the Metrics Finance Will Accept](https://locus.sh/blogs/prove-tms-value-90-days-baselines-control-groups/)

[Ishan Bhattacharya](https://locus.sh/blogs/author/ishan_locus/)

Oct 22, 2026

How to prove a TMS is working in 90 days: freeze a baseline, use a matched control group, set attribution rules and classify savings so finance accepts them. Includes a worked example.

[Read more](https://locus.sh/blogs/prove-tms-value-90-days-baselines-control-groups/)

[https://locus.sh/blogs/shipment-visibility-maturity-model-five-levels/](https://locus.sh/blogs/shipment-visibility-maturity-model-five-levels/)
#### [General](https://locus.sh/blogs/category/general/)

## [The Shipment Visibility Maturity Model: Five Levels From Status Updates to Autonomous Response](https://locus.sh/blogs/shipment-visibility-maturity-model-five-levels/)

[Anas T](https://locus.sh/blogs/author/anas_locus/)

Oct 22, 2026

A five-level shipment visibility maturity model, from reported status to responsive action, with measurable criteria, a self-assessment example and what each level needs from software.

[Read more](https://locus.sh/blogs/shipment-visibility-maturity-model-five-levels/)

## How to Measure Tracking Accuracy Against Ground Truth: A Test Method for GPS, Driver App, Carrier API and Scan Data

- Share
- [Print](javascript:window.print())
- [Download](#)
- [Schedule a Demo](https://locus.sh/schedule-demo/)

### Is your team spending more time on fixing logistics plan than running the operation?

- Agentic transportation management from order intake to freight settlement
- Route optimization built on 250+ real-world constraints
- AI-driven dispatch with automatic execution handling

20%Cost Reduction

66%Faster Planning Cycles

[Schedule a demo](/schedule-demo/)

Insights Worth Your Time

#### [General](https://locus.sh/blogs/category/general/)

## [Locus 2026 UK Consumer Survey: Why Returns Visibility is Now the Conversion Engine for AI-Driven Shopping in UK Retail](https://locus.sh/blogs/returns-visibility-conversion-engine-ai-shopping-uk-retail-locus-q2-2026-consumer-survey/)

[Aseem Sinha](https://locus.sh/blogs/author/aseem_locus/)

May 29, 2026

#### [General](https://locus.sh/blogs/category/general/)

## [Locus 2026 US Consumer Survey: Generative AI isn’t Just Changing How Consumers Shop, it’s Breaking the Demand Patterns US Retail Was Built On](https://locus.sh/blogs/generative-ai-shopping-effect-retail-fulfillment-operations-locus-q2-2026-consumer-survey/)

[Ishan Bhattacharya](https://locus.sh/blogs/author/ishan_locus/)

May 29, 2026

#### [General](https://locus.sh/blogs/category/general/)

## [Embedded vs Bolted-On AI: The Architecture Question European Logistics Buyers Are Asking](https://locus.sh/blogs/embedded-vs-bolted-on-ai-european-logistics-platform-architecture-business-benefits/)

[Aseem Sinha](https://locus.sh/blogs/author/aseem_locus/)

May 21, 2026

#### [General](https://locus.sh/blogs/category/general/)

## [Hybrid Fleet Management: How Owned, 3PL, Gig, ICE, and EV Capacity Actually Operate at Most Enterprises](https://locus.sh/blogs/three-workforce-fleet-reality-owned-3pl-gig-drivers/)

[Aseem Sinha](https://locus.sh/blogs/author/aseem_locus/)

May 7, 2026

#### [General](https://locus.sh/blogs/category/general/)

## [US Returns Hit $850 Billion in 2025: Why US Retailers Are Restructuring Reverse Logistics in 2026](https://locus.sh/blogs/850-billion-us-returns-ai-routing-reverse-logistics-2026/)

[Ishan Bhattacharya](https://locus.sh/blogs/author/ishan_locus/)

May 7, 2026