RTK Service Landscape: How Do You Choose When Everyone Promises Centimeters
On this page
TL;DR
Choose an RTK correction service by how reliably your receiver meets the job's requirements. Use compatibility and coverage to find candidates, then test repeated starts and accuracy across locations and VRS handovers. Count failed attempts alongside successful starts. CAIRN supports discovery; field evidence shows which service-and-receiver combination you can rely on.
In our previous article on Network RTK, we followed the move from setting up a base in the field to using a continuously operated reference network. Sharing that infrastructure can reduce setup work and spread its cost across more users. We expect commercial RTK correction services to become a common choice for that reason.
The U.S. market already has many public networks and commercial RTK service providers. Their coverage maps look extensive, and nearly all promise centimeter-level positioning. With so many similar claims, how do you tell which service will work well for you?
The U.S. RTK Service Landscape: How Are the Networks Built?
Behind the services are public reference networks, commercially operated station networks and stations contributed by independent participants. How those stations are built and maintained explains much of the difference between service strategies.
Public networks are often maintained by transportation or surveying agencies, supporting local surveying and coordinate control. Oregon's ORGN currently issues rover accounts at no direct charge, while Florida's FPRN states that it has no current plans to charge for its services or products. Both require the appropriate account registration. They are useful places to begin looking for local resources.
Some commercial operators build stations themselves or deploy them with professional partners. They coordinate station installation, reference coordinates, monitoring and repairs. Direct control gives them a clear route to correcting problems, along with the continuing cost of sites, equipment and maintenance.
Other platforms expand through independent station owners. Data contributions, quality scores and financial incentives encourage participants to connect existing stations or install new ones. This can spread the cost of construction and accelerate coverage. Maintaining that coverage then requires effective participant support, automated quality control and alternative stations when a contributor becomes unavailable.
Partnerships and aggregation can sit above any of these sources. A service may combine networks and give customers one access arrangement across them. Station ownership shows who can intervene when a site fails; processing and long-term maintenance determine how that failure reaches the user. A network can combine several sources, so the practical question is how the operator keeps their data and availability consistent.
What Do RTK Correction Services Actually Deliver?
Station supply and product delivery are separate decisions. A network of independently owned stations can supply several downstream products, just as an operator-managed network can offer several ways to use its data. The main choices are single-base RTK, Network RTK or VRS, PPP or PPP-RTK, and post-processing products.
Single-base services send observations from a real reference station to the rover, which solves a baseline using the familiar RTK measurement relationship. A nearest-base service adds automatic station selection. This delivery path needs little regional modeling on the platform: wider coverage primarily comes from placing usable stations closer to more working locations.
Network RTK processes observations from several stations to model how errors change across an area. A virtual reference station, or VRS, represents those conditions through observations generated near the rover. The rover processes that stream much like observations from a nearby real base. More of the service's useful reach comes from estimating local corrections between physical stations.
PPP and PPP-RTK offer further choices for wide-area positioning, with different demands on the receiver, convergence time and regional corrections. The OSR and SSR article explains those architectures. For selection, consider the application: if a particular product promises 5–7 cm horizontal accuracy, would that satisfy the job? A wider service area or lower total cost may justify that accuracy for some jobs; others need tighter tolerances.
Tasks that permit a later result can use post-processed kinematic positioning, or PPK. The purchased product might be reference observations, virtual RINEX data or a complete trajectory-processing service. Each requires a suitable rover logging workflow. Real-time VRS can provide field feedback while PPK produces the final trajectory; working from recorded GNSS logs makes that combination practical. The choice begins with when the result must be ready.
How to Choose an RTK Correction Service: Set the Requirements First
The position you see comes through a long chain: reference observations, network processing, communications, receiver estimation and coordinate handling. A problem anywhere along that chain can look like a weak correction service. Set the requirements first so each trial tests the same job.
First establish which services you can use with your equipment and in your operating area. Record the receiver and firmware so a later configuration change does not become an apparent service difference. Then define what usable means for the job: how close the coordinates must be to the reference, how long they must stay there, and how long the crew or machine can wait. A survey checkpoint and a moving vehicle may need very different interruption limits.
Check RTK receiver compatibility against the actual correction stream before starting a trial. Does the stream contain the constellations, signals and RTCM observation messages the receiver supports? Does the selected mountpoint require GGA position feedback? Confirm the coordinate frame and epoch against the project, including any transformation between NAD83(2011) and an ITRF realization. A receiver can report Fixed while the delivered coordinates remain offset from the project's reference.
The subscription price is only part of the deployment cost. A fleet plan must allow the number of rovers that will connect at the same time, and its usage terms must cover the intended work. Ask who responds when a mountpoint or local station goes offline. An SDK or API is valuable when it reduces the effort of provisioning devices and finding faults; count that work alongside communications and subscription fees.
RTK Coverage: What Does the Map Actually Promise?
A coverage map helps build a shortlist. Its meaning depends on how it was drawn. One service may display circles around individual stations; another may publish a modeled network service area. Increasing a circle's radius from 20 km to 50 km makes the drawn area 6.25 times as large, without changing the station. Ask which definition supports the colored region.
Behind that definition lies an investment decision: how much capability comes from station density, and how much from the processing platform? More nearby stations shorten physical baselines and reduce dependence on regional modeling, while increasing deployment and maintenance work. Network processing uses multiple stations and atmospheric models to serve the intervening space, requiring investment in algorithms, computing and validation. Operators can combine these approaches or obtain processing through partners.
Station spacing is therefore one clue to the service strategy. Meeting similar requirements with a more widely spaced network asks more of the model. Adding stations creates more facilities to maintain. Both approaches need abnormal-data detection, fault isolation and usable redundancy. Coverage is delivered jointly by stations and processing, then sustained through long-term operations.
That judgment suggests where to test. A single-base route needs attention at the far reaches between stations and when a station becomes unavailable. Regional modeling deserves scrutiny at network edges, across substantial height differences and during challenging ionospheric conditions. Stations from different sources also need consistent quality across locations. These are test priorities for combinations of infrastructure and processing, rather than a ranking of business models.
RTK Service Reliability: Test Every New Start
To evaluate RTK correction services, turn coverage claims into evidence from the area where the system will work. Begin with repeated starts at one location, extend the observations across locations, then examine accuracy during movement.
A receiver that stays Fixed for a long session shows how the system behaves after initialization. Users also need to know how reliably they can start work in the morning, after lunch or after a power interruption. Repeating the start makes those waiting periods and failed attempts visible.
In our testing, we power the complete device off for one minute, then power it on, invoke the receiver's documented cold-start procedure and observe it for five minutes. Start the clock at power-on, including the wait for the first position and corrections. Repeat around the clock for several days with the same configuration, recording every attempt. Choose an observation window that covers the task's acceptable waiting time. A second receiver running continuously at the same site helps show whether a failed restart coincides with a wider interruption.
Use a known point accurate enough to judge the chosen tolerance, with consistent coordinates and antenna reference. Preserve each startup as a timeline. The first position and the first correction distinguish receiver acquisition from correction delivery; first Fixed and first sustained accuracy show whether ambiguity resolution actually made the position usable. For GGA-dependent streams, the first feedback message also shows when the service received a rover position to work from. Table 1 summarizes the records that make these stages visible.
Mark the time to usable accuracy at the first epoch that begins the chosen hold period entirely within tolerance. The deadline hit rate must use all scheduled attempts as its denominator, including those that never become usable. Report waiting-time medians and 95th percentiles separately for starts that succeed within the observation window, alongside the share still unusable at its end. A report can otherwise look fast simply because its slowest attempts have disappeared. State which epochs enter any position-error percentiles as well.
Look for Fixed but outside tolerance, sometimes called “fixed but wrong”: the receiver reports Fixed while its coordinates miss the requirement. A wrong ambiguity solution, a reference-coordinate problem or a frame mismatch can all produce this symptom. Compare with the known reference and investigate the cause, even when the status display looks reassuring.
Receiver compatibility belongs in this test. Include several mainstream receiver or positioning-engine platforms where practical, holding each platform's firmware and configuration fixed. Compare services on the same receiver, and receivers on the same service, at the same site in overlapping time periods. Keep antenna placement and signal exposure comparable. Present results by service-and-receiver combination so users can recognize the system they would actually deploy.
Spatial Sampling: How Much Can a Few Locations Tell You?
A week at one point tells you about that point. A larger operating area needs samples that cover different station geometry, environments and times.
Include network interiors and edges, sparse station areas, longer single-base reaches and meaningful height differences. Across larger regions, include different geomagnetic latitudes and local times. Ionospheric scintillation varies with latitude, time and geomagnetic activity; one quiet location cannot represent every atmospheric challenge elsewhere.
For the core service comparison, use open sites with controlled antenna installations and record communications and satellite conditions. Add obstructed environments as tests of the complete deployed system. That separation helps interpret whether a difficult result follows the service area or the local reception environment.
Providers often keep their detailed topology private. Build samples around the work area and published boundaries, then add random locations to reduce the bias toward cities and convenient stops. Report results by region, time and receiver. A regional average can hide the one corridor where the machine must work every day.
Network RTK Continuity: Does Accuracy Hold across Grid Boundaries?
Vehicles and agricultural machines can cross several network cells during one job. As a rover moves, the service may select another shared virtual station or regenerate a VRS nearer its new location. That change of virtual reference is a VRS handover. Good results at stationary points leave this transition untested.
A new virtual station brings new reference coordinates and a corresponding observation stream. The network and receiver have to carry the positioning solution across that change. Poor handling can appear as a position step, a temporary loss of accuracy or reinitialization. For moving work, assess whether the error stays within the required tolerance before, during and after the virtual-station handover.
Choose a route through known or suspected cell transitions, and repeat it in both directions at different times. Keep the correction-stream events alongside the vehicle trajectory. A reference change can then be compared with any step in position, rather than inferred from a Fixed flag. The accuracy reference must be independent of the correction service under test; shared correction errors can otherwise make two trajectories agree for the wrong reason. Align their timestamps and antenna offsets so the comparison follows the same physical point.
When cell boundaries are unpublished, longer routes and changes in the virtual reference coordinates can identify candidate handovers for repeat runs. A stream that keeps arriving and a receiver that keeps reporting Fixed provide useful state records; the trajectory comparison establishes accuracy continuity. This turns “that section of road does not work well” into an event that can be repeated and investigated.
The same records should let another engineer reconstruct what happened before judging the result. Keep the service events and the measured positioning outcome together:
| Test | Keep in the record | Judge from the result |
|---|---|---|
| Repeated cold starts | Power-on and cold-start confirmation; first position, connection, correction, Fixed and usable accuracy; first GGA feedback when required. | Deadline hit rate across all scheduled starts; waiting-time percentiles among starts usable within the window; attempts still unusable when it ends. |
| Spatial sampling | Location and time, receiver and firmware, available station/network context, satellite reception and communication conditions. | Accuracy and startup results by location and time. Keep weak corridors visible rather than hiding them in a regional average. |
| VRS handovers | Virtual-reference coordinates and identifiers, GGA feedback, correction age, connection events and solution state, aligned to the reference trajectory. | Position steps, time outside tolerance and recovery before, during and after each reference change. |
CAIRN: Find Possible Services before Testing Their Quality
Selection starts with discovery: which services can you try in the area, and what supports their coverage information? Kalmix maintains CAIRN to help with that step. It brings together public service information, shows published or estimated coverage with its basis and check date, and records access checks for selected services. CAIRN helps build a candidate list; it does not rate service quality. Measuring that takes a much larger program.
You do not need a nationwide program to begin. Use your own receiver at a known point over several days, including the times when you normally work. If the application moves, repeat one representative route and retain its event logs. Surveyed checkpoints provide local accuracy checks; a reference trajectory independent of the tested service is needed to examine continuity between them. This first trial can tell you where a larger investigation is worth the effort.
Read a provider or third-party report the same way. Does its denominator retain failed and timed-out starts? Does success mean coordinates stayed within the required tolerance by the deadline? Finally, were the receiver, reference and test conditions close enough to yours for the result to be useful? For single-base results, the physical baseline is part of that context. A common protocol could also let volunteers contribute comparable startup and route records, filling gaps that one team cannot cover alone.
A Service Has to Be Useful beyond Its Best Accuracy Claim
Similar accuracy claims leave room for very different operating experiences. Station density, processing and maintenance all have to hold up in the places where you work. A fast initialization at one convenient site cannot tell you how that combination behaves across different times or along a moving route.
Begin with services that meet your receiver, access and coordinate requirements. Use CAIRN to find candidates, then ask for evidence you can relate to your own tests. Choose on how often the system reaches and holds the required accuracy on time, including failed starts and difficult handovers in the decision.
Key Takeaway
Use coverage to find candidates. Use repeated starts and route tests to decide whether a service can do the job.
Frequently Asked Questions
How do I choose an RTK correction service?
Begin with services that support your receiver and cover the area where you work. Before the trial, define how accurate the coordinates must be and how long you can wait for them. Then test whether the system meets that requirement repeatedly, including at weaker locations and during movement. Keep failed starts in the results, and compare the full deployment cost once candidates have passed those checks.
Does a larger RTK coverage map mean a better service?
No. Coverage maps can use different station radii, network boundaries and performance assumptions. Ask how the map was drawn and what accuracy and availability it promises. A larger colored area does not establish that your receiver will meet its deadline or accuracy requirement there. Check the parts of the map where you will work.
How should I measure RTK time to fix?
Start the clock at power-on so acquisition and correction-delivery delays remain in the measurement. In the test described here, the device is off for one minute, then cold-started and observed for five minutes. Record first Fixed separately from first sustained accuracy within tolerance. Repeat at different times, keeping every scheduled attempt in the deadline hit-rate denominator, and state that waiting-time percentiles cover starts that became usable within the window.
Will an RTK correction service work equally well with every receiver?
No. A receiver must understand the signals and observation messages in the stream, and its firmware must handle reference changes correctly. Check those requirements before comparing results. GGA feedback and coordinate handling also have to match the service. Test your actual receiver with fixed firmware and configuration, because a successful result on another platform does not establish compatibility with yours.
How do I check accuracy through a VRS handover?
Compare the vehicle trajectory with a reference independent of the service being tested. Keep the reference-change and connection events beside that trajectory, so you can relate a coordinate step to what the service delivered. Check whether the error leaves tolerance and how long recovery takes. Repeat the route in both directions and at different times; a continuing Fixed status alone cannot establish continuity.
How can I evaluate an RTK service without a large test program?
Start at a known point with your own receiver and repeat the trial across several days and working times. Keep failed attempts instead of reporting only the successful waits. For moving work, repeat a route you actually use. Checkpoints can reveal local errors; testing accuracy between them needs an independent reference trajectory. Use this first sample to decide where more testing would change the selection.
Is a paid RTK correction service more accurate than a free one?
Not necessarily. Price does not establish accuracy or reliability. A public network may already meet your requirements, while a commercial plan may offer support, wider access or integration features you need. Compare both under the same receiver and test conditions. Check commercial-use eligibility and outage support before relying on either option; those differences can matter more than a similar centimeter-level accuracy claim.
What should I ask before committing to an RTK service?
Ask the provider to confirm that your receivers and operating area are covered under the proposed plan. For a fleet, check whether every rover can connect at the same time and whether commercial use is permitted. Agree on who responds to outages and how quickly. A useful trial report should also show failed starts and its accuracy reference, so you can judge the evidence before committing.
Continue exploring
How Network RTK Works
Understand how reference networks generate VRS observations and handle moving rovers.
LateralRead Accuracy Claims
Compare error metrics and distinguish repeatability from accuracy against a reference.
ApplyExplore Field Test Evidence
Inspect real-world GNSS test setups, references and recorded results.
