When engineering teams evaluate a robotic arm for a new automation line, two performance numbers dominate the conversation: accuracy and repeatability. Both appear on every spec sheet. Both are expressed in millimeters. And both are routinely misunderstood — or worse, used interchangeably — in ways that lead to costly mismatches between the robot purchased and the task it needs to perform.

The consequences are real. Specifying the wrong metric has caused automation projects to underperform from day one: a robot with outstanding repeatability but poor accuracy fails when positions are fed from a vision system or CAD model; a robot specified for absolute accuracy in a task that simply needed consistent cycle-to-cycle motion wastes budget on capabilities that were never required. Getting this decision right doesn’t take a robotics PhD — it takes a clear understanding of what each number actually means and which one your application actually demands.

This guide breaks down robot arm repeatability vs. accuracy from first principles, explains how each is tested and what affects it in real factory conditions, and gives you practical guidance on which specification to prioritize — and how to read manufacturer datasheets critically before you sign a purchase order.

What Is Robot Arm Accuracy?

Robot arm accuracy describes how closely the robot’s end effector arrives at a commanded position — a coordinate defined in the robot’s world frame, not a position the robot has been physically taught. If you instruct the robot to move to X=300mm, Y=150mm, Z=200mm, accuracy tells you how close it actually gets to that point in space. The gap between the commanded position and the measured position is the accuracy error.

This distinction matters more than it sounds. Accuracy is a property of the robot’s kinematic model and its calibration against the real world. Even a robot with excellent mechanical build quality can have poor absolute accuracy if its internal model doesn’t precisely match its physical geometry — and that mismatch is common. There are several sub-types of accuracy worth knowing when reading a spec sheet:

• Absolute accuracy (pose accuracy): The error between a position commanded via Cartesian coordinates and the actual position reached, measured anywhere within the robot’s working envelope.
• Position accuracy: Focuses specifically on the X, Y, Z coordinates of the tool center point (TCP), ignoring orientation. Critical for pick-and-place and material handling tasks where placement location matters.
• Path accuracy: How closely the robot follows a commanded path between two points, not just the endpoints themselves. Relevant for continuous-motion tasks like welding beads, dispensing, or painting.

One frequently overlooked reality: accuracy degrades toward the outer limits of the robot’s reach. As the arm extends further from its base, small angular errors at each joint compound, and axis deflection under load adds to the total error. A robot rated for ±0.3mm accuracy at a reference position near the center of its workspace may perform considerably worse at full extension.

What Is Robot Arm Repeatability?

Repeatability measures something fundamentally different: how consistently the robot returns to the same position when commanded to do so multiple times under the same conditions. The target position is typically one the robot has been physically taught — the operator guides the arm to the desired point and saves those joint angles. Each time the robot returns to that taught point, repeatability captures the scatter, or spread, of all those arrivals around a central cluster.

Critically, the cluster doesn’t have to be centered on the originally intended position for repeatability to be high. A robot could consistently land 0.5mm away from where you intended it to go, but if it does so with a spread of only ±0.02mm across hundreds of cycles, its repeatability is excellent. This is the most important conceptual distinction engineers need to internalize before writing a specification: repeatability does not tell you whether the robot is going to the right place — only that it is going to the same place every time.

In practical terms, this is why most industrial applications — welding, palletizing, machine tending, conveyor pick-and-place — are designed around taught positions rather than computed Cartesian coordinates. Once the robot is programmed by teaching, the consistency of its return trips is governed entirely by repeatability, and absolute accuracy becomes irrelevant to day-to-day operation.

Accuracy vs. Repeatability: The Key Differences

The classic analogy is a dartboard. Accuracy is how close your darts land to the bullseye. Repeatability is how tightly clustered your darts are, regardless of where on the board they land. A player who consistently groups darts in the top-left corner has high repeatability but poor accuracy. A player who scatters darts randomly around the bullseye has decent accuracy on average but poor repeatability — and that inconsistency makes them useless in a production environment that needs the same motion thousands of times a day.

The table below summarizes the operational difference between the two metrics:

Property	Accuracy	Repeatability
What it measures	Error vs. a commanded Cartesian coordinate	Consistency of return to a taught position
Reference point	World-frame coordinate (absolute)	Previously taught joint position
Typical spec range	±0.3mm to ±1.0mm (standard robots)	±0.02mm to ±0.4mm (industrial robots)
Guaranteed by manufacturers?	Rarely guaranteed without calibration	Yes — the primary published spec
Primary applications	Vision-guided pick, offline programming, CAD-driven tasks	Welding, palletizing, machine tending, packaging

Resolution: The Third Metric You Shouldn’t Ignore

Most spec discussions stop at accuracy and repeatability, but there is a third related metric that sits beneath both: resolution. Resolution refers to the smallest incremental movement the robot’s control system and drive train can execute. If a robot’s joint encoder can only distinguish position changes of 0.05mm, then neither its accuracy nor its repeatability can ever be better than that threshold — resolution sets the floor for both metrics.

In practice, resolution is determined by encoder quality and gearing ratio. High-resolution encoders on precision harmonic drive joints can push resolution well below 0.01mm, while lower-cost systems with coarser encoders or significant gear backlash operate at a coarser increment. When comparing robots, especially at different price points, it is worth checking resolution alongside accuracy and repeatability — because a robot that claims ±0.05mm repeatability but has a mechanical resolution of ±0.04mm is operating at the very edge of its hardware capability, leaving almost no margin for thermal or load-induced variation.

How Accuracy and Repeatability Are Measured

The globally accepted benchmark for measuring these performance characteristics is ISO 9283, titled Manipulating Industrial Robots: Performance Criteria and Related Test Methods. The standard defines the test conditions, measurement equipment, environmental constraints, and calculation methods that manufacturers must follow when publishing accuracy and repeatability specifications. It covers pose accuracy, pose repeatability, path accuracy, path repeatability, and several other performance indices including drift, overshoot, and stabilization time.

Under ISO 9283, repeatability is geometrically defined as the radius of the smallest sphere that encompasses all the end-effector positions reached for the same commanded position across multiple trials. The most common testing tools include:

• Laser interferometers: Use interference patterns in coherent light to measure displacement with sub-micron precision. They compare commanded position to measured position across a range of workspace points, making them the gold standard for absolute accuracy measurement.
• Laser trackers: Portable coordinate measurement systems that follow a reflector mounted to the robot’s TCP in real time. Widely used in aerospace and large-volume applications where the robot arm has significant reach.
• Vision-based tracking systems: High-resolution cameras track the robot’s TCP against fixed reference markers. Software calculates positional error between intended and actual location over many cycles.
• On-site fixture checks: For production floor validation without lab equipment, engineers command the robot to contact a precision pin or datum surface repeatedly and record the variation. Less precise than optical methods but useful for ongoing monitoring.

One important caveat: ISO 9283 testing is conducted under controlled laboratory conditions — specific temperature ranges, defined load configurations, and warm-up periods. The numbers published in a robot’s datasheet reflect those ideal conditions. Real-world performance on a busy production floor with temperature swings, variable payloads, and high cycle rates will typically show some degradation from the published figure.

What Affects a Robot Arm’s Accuracy and Repeatability

Understanding the sources of error helps engineering teams both select the right robot and manage performance over the system’s operational life. Several factors contribute to accuracy and repeatability degradation in production environments:

• Mechanical design and gear quality: Joint stiffness and gearbox backlash have the largest mechanical influence on repeatability. Robots with compact, rigid gearboxes and preloaded joints minimize the backlash-induced variation that causes scatter at the TCP. Smaller robots generally achieve tighter repeatability tolerances than larger ones due to shorter moment arms.
• Thermal expansion: Metal components expand with heat. In a robot running at high cycle rates over long shifts, the arm segments and joint housings warm up and change length by small but measurable amounts. This introduces positional drift — a slow migration of the TCP away from its reference position over the course of a shift — which primarily affects accuracy rather than short-term repeatability.
• Payload and end-of-arm tooling (EOAT): A heavier or eccentric gripper introduces deflection at the wrist and additional joint loading. Both effects shift the TCP position relative to the robot’s internal model. This error is worse at high speeds, during acceleration and deceleration, and at extended arm positions.
• Calibration drift: Over time, the robot’s internal kinematic model drifts away from the real-world geometry of its joints and links due to wear, mechanical settlement, and accumulated small shocks. Without recalibration, the robot’s commanded positions increasingly diverge from its actual positions, reducing absolute accuracy.
• Position within the working envelope: Accuracy degrades at the outer boundaries of the work envelope. Angular errors at each joint compound over longer moment arms, and all error sources reach their worst case at maximum reach. The repeatability figure published in a datasheet is typically measured at a reference point near the center of the workspace, not at the edges where most applications actually place their workpoints.

What to Actually Specify for Your Application

This is the question the article title promises to answer — and the answer depends entirely on how your robot will be programmed and what it is required to do. There is no universal answer, but there is a clear decision framework.

Specify Repeatability When:

Repeatability is the critical metric for the majority of industrial automation tasks. If your robot will be programmed by teaching — an operator physically guides the arm to each required position and saves the joint angles — then absolute accuracy is largely irrelevant to daily operation. The robot only needs to return to that taught position consistently. Applications in this category include:

• Spot welding and arc welding at fixed positions
• Machine tending (loading and unloading CNC machines, injection molding presses)
• Palletizing and depalletizing from a fixed infeed point
• Conveyor pick-and-place where part positions are consistent
• Packaging and carton forming
• Bolt fastening and screw driving at pre-taught locations

For these applications, a robot with ±0.05mm repeatability and ±0.5mm accuracy will outperform a robot with ±0.1mm repeatability and ±0.1mm accuracy — because repeatability governs the consistency of every cycle, while the absolute accuracy number will never come into play once the program is taught.

Specify Accuracy When:

Accuracy becomes the governing specification when the robot must move to positions it has not been physically taught — when target coordinates arrive from an external source, an offline program, or a real-time sensor system. If any of the following conditions apply to your application, absolute accuracy must be a primary selection criterion:

• Positions are generated offline from CAD models and downloaded to the robot without on-site teaching
• A vision system or laser scanner detects part location in real time and transmits Cartesian coordinates to the robot controller
• The task involves a very large number of positions (thousands of drill points, for example) where manual teaching of every point is impractical
• The robot must interact with fixtures, jigs, or CNC machines where positional alignment is defined by physical geometry rather than a taught point
• Multiple robots must collaborate on a shared workpiece and their coordinate frames must align

In these scenarios, a robot that cannot accurately translate a commanded Cartesian coordinate into a correct physical position will fail the task regardless of how tight its repeatability is. For vision-guided applications in particular, you may also need to budget for external calibration — a kinematic calibration process that measures the robot’s actual geometry against a reference and updates the kinematic model to reduce absolute accuracy error significantly.

When You Need Both:

Some applications demand both high repeatability and high accuracy. Electronics assembly — placing fine-pitch components on PCBs — requires the robot to arrive at the correct absolute position (accuracy) and to do so with the same precision every cycle (repeatability). Precision medical device assembly, surgical robot calibration, and aerospace part machining share this requirement. In these cases, expect to pay a significant premium, and plan for regular kinematic calibration as part of the system’s maintenance schedule.

Manufacturer Spec Sheets: What They Don’t Tell You

There is an important industry-wide pattern every automation buyer should understand: robot manufacturers consistently guarantee repeatability, not absolute accuracy. A robot datasheet may publish both numbers, but only the repeatability figure is a firm performance commitment backed by ISO 9283 testing. The accuracy figure, where quoted at all, is often a best-case or representative value measured under favorable conditions with full kinematic calibration applied — conditions that may not reflect out-of-box performance.

A robot with ±0.05mm repeatability may have ±0.5mm or worse absolute accuracy without external calibration. This gap is not a defect — it reflects the fundamental difference between the precision of a robot’s joint encoders (which drives repeatability) and the accuracy of its kinematic model (which drives absolute accuracy). For taught-position applications, this distinction is harmless. For offline-programmed or vision-guided applications, it is a critical procurement risk if not addressed at the specification stage.

When evaluating manufacturer claims, ask these questions:

• Is the repeatability figure measured at the center of the workspace or across the full working envelope?
• At what payload and speed was the specification tested?
• Is the accuracy figure measured with or without kinematic calibration applied?
• What is the thermal warm-up time assumed in the specification?
• Does the figure represent unidirectional or bidirectional repeatability?

These questions will quickly reveal whether a published specification is a genuine engineering commitment or a marketing figure measured under the best possible conditions.

Accuracy and Repeatability in AMR-Integrated Systems

For factories deploying robotic arms alongside autonomous mobile robots (AMRs) — a configuration that is rapidly becoming the standard in modern warehousing and manufacturing — the accuracy and repeatability specification for the arm must be considered in the context of the mobile platform’s own positioning error. When a robotic arm is mounted on or coordinating with a mobile platform, the arm’s TCP accuracy in world coordinates is the sum of the mobile robot’s docking precision and the arm’s own accuracy.

This is why Reeman’s approach to integrated automation focuses on the full system: our IronBov Latent Transport Robot and Ironhide Autonomous Forklift use laser navigation and SLAM mapping to achieve precise, repeatable docking positions — providing the stable, consistent base that arm-on-AMR configurations require to maintain acceptable TCP accuracy at the work point. Similarly, the Reeman Robot Mobile Chassis platform is engineered to support stable, vibration-controlled deployment of end-of-arm tooling in logistics environments.

For operations that integrate robotic arms with material transport, consider the following approach to specification:

• Characterize the docking repeatability of the mobile platform separately from the arm’s repeatability spec
• For tasks where the arm must pick from a fixed station while the AMR docks, the docking precision of the mobile platform feeds directly into the effective accuracy of the arm — so prioritize docking repeatability in the AMR spec
• For applications where the arm operates from a fixed base with AMRs delivering parts to a consistent presentation point (such as a conveyor or locating fixture), the arm’s own repeatability is the governing spec and the mobile platform’s positioning only needs to be good enough to place parts within the arm’s reach zone
• Vision-guided arm systems can compensate for moderate docking variation, shifting the burden from navigation precision to sensor and control system quality

Whether you are operating a standalone arm on a factory floor or building a fully integrated mobile-arm system, the fundamental principle remains: define your task type first, then determine which performance metric governs it, then specify accordingly — and read the manufacturer’s datasheet with the critical eye it deserves.

FAQs

Is repeatability always more important than accuracy for industrial robots?

For most industrial applications programmed by teaching, yes — repeatability governs production quality and consistency more directly than absolute accuracy. However, when robots receive position commands from external sources such as vision systems, CAD-generated programs, or collaborative multi-robot setups, absolute accuracy becomes equally or more critical. The correct answer depends entirely on your programming method and task type.

What is a typical repeatability specification for an industrial robot arm?

Most industrial robot arms fall in the range of ±0.02mm to ±0.4mm repeatability. Smaller, lighter-payload arms designed for precision electronics assembly tend toward the tighter end. Heavy-payload arms used for palletizing or automotive body handling typically sit between ±0.05mm and ±0.2mm. Collaborative robots (cobots) generally range from ±0.03mm to ±0.1mm. These figures are measured under ISO 9283 conditions and may degrade slightly in real-world production environments.

How often does a robot arm need to be recalibrated to maintain accuracy?

For most production environments, annual kinematic calibration is the baseline recommendation. High-precision applications — electronics assembly, medical devices, aerospace — may require calibration every six months or following any significant mechanical event such as a collision or joint replacement. Repeatability, being primarily a function of encoder and gearbox quality, degrades more slowly and can often be monitored through periodic on-site fixture checks rather than full kinematic calibration.

Can software updates improve a robot’s accuracy?

Yes, to a degree. Controller firmware updates can improve motion planning, thermal compensation algorithms, and kinematic model accuracy. However, software cannot overcome fundamental mechanical limitations such as encoder resolution, gear backlash, or link rigidity. Software improvement is most effective when the underlying hardware is already capable and the inaccuracy is driven by a suboptimal kinematic model rather than mechanical wear or design limitations.

Does robot arm position within the workspace affect accuracy?

Yes, significantly. Accuracy degrades toward the outer boundaries of the working envelope where joint angular errors compound over longer moment arms and where payload-induced deflection is greatest. Repeatability is also somewhat worse at extended positions than near the center of the workspace. If your application places workpoints near the edge of the robot’s reach, test or request accuracy and repeatability data measured at those specific positions rather than relying solely on the datasheet’s reference-point specification.

The Bottom Line

Robot arm repeatability and accuracy are not interchangeable. Repeatability describes consistency — how reliably the arm returns to a taught position cycle after cycle. Accuracy describes correctness — how closely the arm reaches a commanded coordinate in the real world. For the majority of industrial automation tasks, from welding and palletizing to machine tending and packaging, repeatability is the specification that drives production quality. Absolute accuracy only becomes the governing requirement when positions are generated externally or when the robot must align with the physical world without prior teaching.

The practical implication is straightforward: before you compare spec sheets, define your programming method. If you teach, specify repeatability. If you compute or sense, specify accuracy — and ask manufacturers the hard questions about how and where that figure was measured. In either case, factor in thermal effects, payload, and workspace position, and treat published specifications as a starting point for engineering evaluation rather than a performance guarantee.

For operations integrating robotic arms with autonomous mobile platforms, the system-level positioning picture matters as much as the arm’s own spec. Understanding both layers — mobile platform docking precision and arm endpoint performance — is the foundation of a well-engineered, reliable automated workflow.

Ready to Build a More Precise Automated Workflow?

Reeman’s industrial-grade autonomous mobile robots, forklifts, and robot chassis platforms are engineered to work together as a complete automation ecosystem — giving your robotic arm the stable, precisely docked base it needs to perform at specification. With over 200 patents, laser SLAM navigation, and deployments across more than 10,000 enterprises globally, Reeman brings proven industrial automation expertise to your factory floor.

Explore our full range of autonomous mobile platforms designed for demanding warehouse and manufacturing environments:

• IronBov Latent Transport Robot — precision latent AMR for high-density logistics
• Ironhide Autonomous Forklift — laser-navigated autonomous forklift for heavy-load handling
• Stackman 1200 Autonomous Forklift — compact stacking forklift for tight warehouse environments
• Rhinoceros Autonomous Forklift — high-capacity truck for demanding throughput operations
• Robot Mobile Chassis — industry-ready chassis platforms for custom arm-on-AMR integration
• Big Dog Delivery Robot and Fly Boat Delivery Robot — autonomous delivery platforms for last-mile intralogistics

Talk to a Reeman Automation Expert →