Connectivity KPIs

IIoT and SCADA Connectivity KPIs: Benchmark Ranges and Improvement Levers

World-class versus typical benchmark ranges for IIoT and SCADA connectivity KPIs, edge coverage, live connectivity rate, tag coverage, PLC data availability and completeness, plus the levers that move them.

Six KPIs tell you whether an IIoT and SCADA program is actually delivering data: edge gateway coverage, machine connectivity rate, OPC UA tag coverage, PLC data availability, OT data completeness, and telemetry event rate stability. Track them together, because a single high number hides the gaps. A plant can show 95 percent edge coverage while live connectivity rate sits at 82 percent, meaning one in six deployed assets is silent right now. Measure each against its own benchmark range rather than a blanket target, and use the Machine Connectivity Rate and PLC Data Availability calculators to read the current values consistently.

Edge gateway coverage is your breadth KPI. Typical mid-rollout programs run 60 to 80 percent of in-scope assets; a healthy foundation is 90 percent or higher, and world-class programs hold 95 to 98 percent with a documented exception list for the rest. Below 80 percent, analytics dashboards develop blind spots that erode trust. Measure it monthly against the in-scope asset register, not total plant assets, or the number lies. The main lever is prioritizing lines by data value, not by ease of connection, so the assets feeding your predictive models get onboarded first rather than last.

Machine connectivity rate is the live-uptime KPI, and it is stricter. Typical plants sit at 85 to 92 percent of connectable machines reporting in the current polling window; world-class operations hold 98 percent or higher sustained. The gap between coverage and this rate is your silent-asset problem. Measure it on a rolling 5 to 15 minute window so a single missed poll does not distort the day. Levers include watchdog alerts on stale tags, redundant gateway paths, and fixing the 20 percent of assets that cause 80 percent of dropouts, usually old serial links and flaky wireless.

OPC UA tag coverage is the depth KPI that separates connected from useful. Typical assets expose 70 to 85 percent of target tags; world-class equipment classes reach 95 percent or more of the tags a use case actually needs. A machine at 100 percent connectivity but 65 percent tag coverage cannot feed a vibration or energy model. Measure per equipment class against a tag specification, never plant-averaged, because averaging hides the critical asset that is starved. The lever is a tag standard per asset type so every new machine onboards to the same depth instead of whatever the integrator had time for.

PLC data availability and OT data completeness are the quality KPIs. Availability, the share of successful reads per polling window, should run 99 percent or higher for control-critical tags; 95 to 98 percent is common but leaves gaps in trends. Completeness, the fraction of expected samples that actually landed in the historian, should hold 98 percent or better for a reliable analytics dataset; below 95 percent, models trained on the data inherit the holes. Measure both continuously, not spot-checked. Levers are store-and-forward buffering on gateways, sample-rate discipline, and eliminating network contention that drops packets under load.

Telemetry event rate and network latency round out the operational KPIs. A stable telemetry event rate, the messages per second your pipeline sustains, should stay within 10 percent of its expected baseline; sustained spikes signal chatty tags or a publish-on-interval config that should be publish-on-change. Round-trip network latency for OT polling should sit under 100 milliseconds on wired segments and under 250 milliseconds on cellular edge links. Measure latency at the 95th percentile, not the average, because the tail is what breaks control loops and alarming. The lever is edge preprocessing to cut round trips and reduce upstream event volume.

Improvement follows a fixed order regardless of plant. Fix breadth first, get edge coverage to 90 percent so the data foundation exists. Then fix uptime, drive connectivity rate to 98 percent so deployed assets stay live. Then fix depth, bring tag coverage to the use-case standard so the streams carry what models need. Only then chase completeness and latency, because tuning quality on a half-connected plant wastes effort. Each stage has a clear exit benchmark, and moving on before hitting it means the next stage inherits the previous stage's gaps compounded across more assets.

Set benchmarks per use case, not per plant. A predictive-maintenance program demands 95 percent tag coverage and 98 percent completeness on the specific vibration and thermal tags it consumes, while a simple production-count dashboard tolerates 90 percent coverage and 95 percent completeness. Holding every asset to the strictest bar wastes budget; holding the critical assets to the loosest bar wastes the investment. Define the target range per KPI per use case, review monthly, and track the trend rather than the single reading. A KPI improving 2 points a month toward target is healthier than a static number that already looks acceptable.

Published 2026-07-01.