OEE Sensors vs PLC Data Collection: The Architecture Decision That Defines Your Data Quality

Every OEE software platform needs machine data to calculate Availability, Performance, and Quality. But how that data enters the system varies dramatically — and that architectural choice determines what you can and cannot see about your production losses. This guide compares the three main approaches to OEE data collection — sensors vs PLC vs manual entry — with specific implications for data quality, deployment speed, and the types of losses each method can detect.

Approach 1: PLC-based data collection

How it works. The OEE software connects to the machine’s Programmable Logic Controller (PLC) via industrial protocols — OPC-UA, Ethernet/IP, Modbus TCP, PROFINET, or proprietary interfaces. The PLC provides cycle count signals, run/stop states, fault codes, and sometimes speed data. The software translates these signals into OEE metrics.

What it captures well. Cycle times (precise to the PLC scan rate), discrete fault events that the PLC is programmed to detect, production counts from part-present sensors wired to the PLC, and machine states defined in the PLC logic (running, idle, fault, setup).

What it misses. Any event the PLC is not programmed to detect — micro-stops that clear themselves before the PLC fault timer fires, speed losses below the PLC’s threshold sensitivity, and stops on equipment without PLCs. PLC-based collection also misses the context of why a machine stopped if the PLC does not have dedicated fault-code outputs for every possible stop reason.

Deployment requirements. A functional PLC with accessible communication ports. Network infrastructure (Ethernet switch, potential VLAN configuration for security). PLC programming expertise to map signal addresses. IT/OT alignment for firewall and security approval. Typical timeline: 1 to 4 weeks per machine depending on PLC accessibility and organizational processes.

Typical platforms using this approach. SensrTrx, MachineMetrics (also offers hardware), Sepasoft, FactoryTalk Metrics, various MES platforms with OEE modules.

Approach 2: external hardware sensors

How it works. Dedicated sensors are installed on or near the machine — not connected to the PLC. Current clamps around the power cable detect whether the motor is drawing power (machine running) or not (machine stopped). Photoelectric or proximity sensors at the product exit count units produced. The sensors feed data to an edge gateway that calculates OEE.

What it captures well. Machine running/stopped state at one-second resolution or better — independent of the PLC. Every stop of 3 seconds or longer, including micro-stops. True cycle time measured from physical product detection. Production count from an independent sensor (not dependent on PLC counter accuracy).

What it misses. Internal PLC fault codes (the sensor knows the machine stopped but does not know the PLC-reported reason). Detailed process parameters (temperatures, pressures, speeds from the control system). Quality data — sensors detect produced units but cannot determine if a unit is a reject without an additional quality sensor or operator input.

Deployment requirements. Physical access to the machine’s power cable or product exit area. No PLC access needed, no network infrastructure changes, no IT/OT approval for PLC connections. Typical timeline: approximately 1 hour per machine, no production shutdown required.

Typical platforms using this approach. TeepTrak (current clamp + photoelectric/magnetic sensors), XL Productivity Appliance, Factbird (sensors + cameras).

Approach 3: manual operator entry

How it works. The operator logs production counts, start/stop times, downtime reasons, and reject quantities into a software interface — tablet, PC, or paper form that is later digitized. The system calculates OEE from the declared data.

What it captures well. The subjective context that machines and sensors cannot provide — why a stop happened (operator’s judgment), which specific defect type caused a reject (visual inspection), setup activities that did not trigger a machine stop signal.

What it misses. Everything the operator does not log. Industry data shows manual downtime logging completion rates of 40 to 55 percent in typical manufacturing environments. Micro-stops are virtually never logged manually. Timing precision is limited — operators estimate stop durations rather than measuring them. The resulting OEE is an estimate based on partial data.

Deployment requirements. Minimal technical infrastructure — a tablet or PC at each workstation. But sustained management effort to maintain data entry discipline over time.

The data quality hierarchy

In terms of machine-state data quality, the hierarchy is clear: hardware sensors (highest — automatic, second-level precision, 100 percent capture) outperform PLC-based collection (high for equipped machines — automatic but limited to PLC logic, variable stop detection thresholds), which outperforms manual entry (lowest — dependent on operator discipline, no micro-stop capture, estimated timing).

In terms of contextual data (why a stop happened, which defect type), the hierarchy reverses: operator input provides the richest context, PLC fault codes provide partial context, and external sensors provide no context for the stop reason. This is why the best-performing OEE systems combine automatic machine-state detection (sensors or PLC) with operator-provided context (reason coding).

The micro-stop question

Micro-stops — stops lasting 3 seconds to 3 minutes — are the single most important differentiator between these approaches. On high-cadence production lines (packaging, filling, labeling, assembly, stamping), micro-stops typically accumulate to 15 to 25 percent of scheduled production time. They are invisible to manual entry, mostly invisible to standard PLC configurations, and fully visible to second-resolution sensor systems.

TeepTrak’s experience across 450-plus factories shows that in the majority of first-time deployments, micro-stops are the number-one Pareto item — larger than any single equipment failure category. Yet the plant had no awareness of this loss before sensor deployment because no prior system could detect it.

The mixed-equipment reality

Most manufacturers do not have a uniform equipment base. A typical shop floor includes machines from multiple decades, multiple manufacturers, and multiple control architectures. PLC-based OEE platforms can only cover the subset of machines with accessible, compatible PLCs. External sensors cover everything that uses electricity. For a plant with 30 machines where 12 have modern PLCs and 18 do not — a PLC-based platform covers 40 percent of the floor. A sensor-based platform covers 100 percent.

Frequently asked questions

Can you combine sensor-based and PLC-based approaches?

Yes. Some manufacturers deploy sensor-based OEE on machines without accessible PLCs and PLC-based collection on newer machines. TeepTrak’s sensor approach is often used as the universal baseline across all machines — including those with PLCs — because the sensor data provides consistent second-level resolution regardless of the PLC configuration.

What about Industry 4.0 and IIoT platforms?

Industry 4.0 and IIoT architectures typically assume PLC or edge-device connectivity. They work well in greenfield installations or plants with modern, uniform equipment. For brownfield plants with mixed equipment — which is the vast majority of manufacturing worldwide — external sensors provide a faster path to OEE data than waiting for a complete IIoT infrastructure upgrade.

Do sensors require network infrastructure changes?

TeepTrak sensors communicate with a local gateway that transmits data via cellular or local network. There is no requirement to connect sensors to the plant’s industrial network or to make firewall changes. The OEE system operates on a separate network path from the OT infrastructure — which is often a requirement from IT security teams.

How accurate are current-clamp sensors compared to PLC signals?

Current clamps detect motor current draw — when the motor runs, current flows; when it stops, current drops. The detection is binary (running/stopped) with one-second resolution. This is comparable to or better than most PLC-based run/stop signals, which often have configurable debounce timers that filter out short events. The key advantage is not higher precision per event but broader coverage — every stop is detected, including those the PLC ignores.

What about quality data — sensors cannot detect rejects?

Correct. External sensors count produced units but cannot distinguish good from rejected parts. Quality data in TeepTrak comes from operator input — the operator enters reject quantities and types on the Field V4 tablet. This is the one area where operator input remains essential. However, machine-state data (availability, performance) — which typically accounts for 80 to 90 percent of total OEE loss — is fully automated.

Is the sensor-based approach more expensive?

The sensor hardware adds a per-machine cost that PLC-based software-only platforms do not have. However, the sensor approach eliminates PLC integration engineering costs — which can be substantial for legacy or proprietary control systems. For mixed-equipment plants, the total cost of deploying sensors on all machines is often lower than the cost of PLC integration on the subset of machines where it is feasible.

What is TeepTrak’s position in this landscape?

TeepTrak is a sensor-first OEE platform. It deploys proprietary non-invasive sensors (current clamps, photoelectric, magnetic) on any electric machine, provides second-level resolution, detects all micro-stops, and adds AI-powered pattern detection via JEMBA. It is deployed in over 450 factories across 30 countries. The typical deployment is 48 hours for a full production line.

Find out what your PLC is missing — Request a TeepTrak demo

For a specific comparison between TeepTrak and SensrTrx, see our SensrTrx alternative OEE analysis. For small manufacturers evaluating OEE for the first time, read our OEE software guide for small discrete manufacturers.