IIoT and SCADA Connectivity KPIs: Coverage, Uptime, Tag Depth and Data Quality Benchmarks

Use a benchmark as a first reference point, not as a target. The ranges below are most useful for spotting a gap worth investigating, and your own monthly trend will say more about your program than any single industry figure.

Every figure here shows its basis. A superscript number links to a public source in the Sources list at the end, and the sentence states that source's context. Most figures are marked no cited source: they are kept exactly as originally published because none of the public documents reviewed for this guide states them, so treat them as illustrative.

At a glance

Read the six KPIs together

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. Measure each KPI against its own benchmark range rather than a blanket target, and read the current values the same way every time.

For example, a plant can show 95 percent edge coverage while its live connectivity rate sits at 82 percent. That leaves 18 percent of deployed assets silent right now, about one in 5.6, or roughly one in six.

Edge gateway coverage

Edge gateway coverage is the breadth KPI: the share of in-scope assets connected through an edge gateway. Typical mid-rollout programs run 60 to 80 percent of in-scope assets no cited source.

A healthy foundation is 90 percent or higher no cited source, and world-class programs hold 95 to 98 percent with a documented exception list for the rest no cited source. Below 80 percent, analytics dashboards develop blind spots that erode trust no cited source.

How to measure

• Measure it monthly against the in-scope asset register.
• Leave total plant assets out of the denominator, or the number lies.
• Keep a documented exception list for assets deliberately left out.

What moves it

• Prioritizing lines by data value, not by ease of connection.
• Onboarding the assets that feed your predictive models first rather than last.

Check your own figure with the Edge Gateway Coverage calculator.

Machine connectivity rate

Machine connectivity rate is the live-uptime KPI, and it is stricter than coverage: the share of connectable machines reporting in the current polling window. Typical plants sit at 85 to 92 percent no cited source, and world-class operations hold 98 percent or higher sustained no cited source.

The gap between coverage and this rate is your silent-asset problem. Dropouts cluster on a few assets: fix the 20 percent of assets that cause 80 percent of dropouts no cited source, usually old serial links and flaky wireless.

How to measure

• Use a rolling 5 to 15 minute window so a single missed poll does not distort the day no cited source.
• Count only connectable machines in the denominator.
• Compare the rate with edge coverage to size the silent-asset gap.

What moves it

• Watchdog alerts on stale tags.
• Redundant gateway paths.
• Replacing old serial links and flaky wireless on the worst offenders.

Track the live rate with the Machine Connectivity Rate calculator.

OPC UA tag coverage

OPC UA tag coverage is the depth KPI that separates connected from useful. Typical assets expose 70 to 85 percent of target tags no cited source, while world-class equipment classes reach 95 percent or more of the tags a use case actually needs no cited source.

For example, a machine at 100 percent connectivity but 65 percent tag coverage still cannot feed a vibration or energy model.

How to measure

• Measure per equipment class against a written tag specification.
• Never average across the plant, because averaging hides the critical asset that is starved.

What moves it

• A tag standard per asset type, so every new machine onboards to the same depth.
• Holding integrators to that standard instead of to whatever they had time for.

Score each equipment class with the OPC UA Tag Coverage calculator.

PLC data availability and OT data completeness

These are the quality KPIs. PLC data availability is the share of successful reads per polling window, and it should run 99 percent or higher for control-critical tags no cited source; 95 to 98 percent is common but leaves gaps in trends no cited source.

OT data completeness is the fraction of expected samples that actually landed in the historian. It should hold 98 percent or better for a reliable analytics dataset no cited source, and below 95 percent, models trained on the data inherit the holes no cited source.

How to measure

• Measure both continuously, not by spot checks.
• Count availability as successful reads over attempted reads in each polling window.
• Count completeness as samples that landed in the historian over samples expected.

What moves it

• Store-and-forward buffering on gateways.
• Sample-rate discipline.
• Eliminating network contention that drops packets under load.

Price the gateway and SCADA work that closes these gaps with the SCADA Upgrade Cost calculator.

Telemetry event rate and network latency

Telemetry event rate is the messages per second your pipeline sustains, and a stable rate should stay within 10 percent of its expected baseline no cited source. Sustained spikes signal chatty tags, or a publish-on-interval configuration 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 no cited source. Measure latency at the 95th percentile, not the average¹, because the tail is what breaks control loops and alarming. That practice comes from IT service reliability rather than OT: Google's site reliability engineering book prefers percentiles to the mean, charts the 95th among other percentiles, and notes that some teams watch the 99th or 99.9th instead¹.

How to measure

• Log messages per second against an expected baseline for each pipeline.
• Measure round-trip latency separately for wired segments and cellular edge links.
• Report latency as a high percentile rather than an average.

What moves it

• Edge preprocessing that cuts round trips.
• Publish-on-change configuration that reduces upstream event volume.
• Cleaning up chatty tags.

See what event volume costs with the MQTT Data Cost calculator.

Improve in a fixed order

Improvement follows a fixed order regardless of plant. 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.

Stages and exit points

• Breadth first: get edge coverage to 90 percent so the data foundation exists no cited source.
• Uptime second: drive connectivity rate to 98 percent so deployed assets stay live no cited source.
• Depth third: bring tag coverage to the use-case standard so the streams carry what models need.
• Quality last: chase completeness and latency only then, because tuning quality on a half-connected plant wastes effort.

Set targets per use case

Set benchmarks per use case, not per plant. Holding every asset to the strictest bar wastes budget, and 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 climbing 2 points a month toward its goal is healthier than a static number that already looks acceptable.

Test the business case for closing a gap with the IIoT ROI calculator.

How to build your own baseline

• Fix the in-scope asset register and each KPI's definition before you measure.
• Collect several months of readings before you set a target.
• Split every KPI by equipment class and use case, never one plant average.
• Compare against your own trend first and published ranges second.
• Record what changed, such as a new gateway or tag standard, whenever a KPI moves.