Skip to main content

General

Data Quality & Metrics

How Proxus tracks the health and reliability of every connected device — and what to do when quality drops.

When a PLC connection drops or a sensor returns corrupt readings, you need to know immediately — before a dashboard shows the wrong number or a rule triggers a false alarm.

Proxus tracks five quality dimensions for every connected device. This page shows you what each dimension measures, how to read the scores, and what to do when quality drops below acceptable levels.

After reading this page, you will be able to:

  • Interpret the 5 data quality scores in Device Details
  • Understand what causes each score to drop
  • Set up rules that react to quality drops
  • Diagnose the most common causes of low freshness or accuracy

The 5 Dimensions of Data Quality

Every device in Proxus is continuously scored across five independent dimensions, each on a 0–100% scale.

DimensionWhat It MeasuresDrops When…
AccuracyPercentage of successful read cycles with complete dataPLC returns errors, partial reads, or protocol fails
CompletenessPercentage of expected tags that are actually receivedSome tags return null or are missing from the response
IntegrityPercentage of received data that passes validationData types mismatch, values out of expected range, corrupt payloads
ConsistencyRolling window stability of recent success/failureIntermittent connection drops, flapping device
FreshnessHow recently the last data was receivedDevice goes offline, network outage, polling delay

Reading the Scores Together

ScenarioAccuracyFreshnessWhat It Means
Healthy device95%+95%+Everything is working normally
Network outage95%40%Device was reading correctly but has not sent data recently — check gateway connectivity
PLC configuration error30%90%Device is reachable but most read attempts fail — check address mapping and protocol settings
Damaged sensor80%95%Reads are coming in but some values fail validation — inspect the physical sensor
Flapping connection60%70%Intermittent drops affect both accuracy and freshness — check network stability

Freshness Thresholds

Proxus calculates freshness using time-based interpolation. The score decays as data ages:

Time Since Last DataFreshness Score
≤ 30 seconds100%
≤ 1 minute95%
≤ 2 minutes90%
≤ 5 minutes80%
≤ 15 minutes60%
≤ 1 hour40%
≤ 12 hours10%
> 24 hours0% (completely stale)
lightbulb
Practical Threshold

For most industrial environments, a freshness score below 80% (data older than 5 minutes) warrants investigation. For safety-critical applications, set your alert threshold at 90% (2 minutes).


Device Diagnostic Metrics

Beyond the 5 quality dimensions, every device maintains detailed diagnostic counters:

MetricWhat It TracksWhy It Matters
Total RequestsTotal communication attemptsBaseline activity — if this stops increasing, the device is not being polled
Successful RequestsSuccessful communicationsCompare with Total Requests to calculate success rate
Failed RequestsFailed communicationsRising count signals protocol or network issues
Last Communication TimeWhen the gateway last attempted to reach the deviceIf this is old, the gateway may be down
Last Data Received TimeWhen valid data was last receivedIf this is old but Last Communication is recent, the device is reachable but not returning data
Average Response TimeAverage response time in msIncreasing values suggest network congestion or overloaded PLC
Connection Error CountNumber of connection failuresHigh count = network or firewall issue
Read Error CountNumber of read operation failuresHigh count = wrong address mapping or protocol config

You can view these metrics in Device Details.


What Happens When a Device Goes Offline

When a device disconnects:

  1. Freshness decays over time based on the threshold table above
  2. Accuracy reflects the last known success rate (it does not retroactively drop)
  3. Consistency decreases as the rolling window captures more failures
  4. The device status changes to Offline in the Operations Center
info
No Automatic Bad Quality Packets

Proxus does not inject artificial "Bad" quality values for tags when a device is offline. Instead, the freshness score provides clear visibility into data staleness. The last known value remains available until new data arrives.


Using Quality in Automation

You can use data quality metrics in your Rules Engine logic:

Example: Create a rule that triggers an alert when any device's Data Freshness drops below 60% (no data for 15+ minutes):

  1. Navigate to Automation > Rules and create a new rule
  2. Set the condition to monitor the freshness metric
  3. Add an Alert Action to notify the operations team
  4. Assign the rule to the relevant gateways

This lets you catch failing devices before operators notice missing data on dashboards.


Common Quality Problems and Solutions

SymptomCheck FirstSolution
Freshness drops to 0% suddenlyGateway status in Management ConsoleGateway may have disconnected — see Gateway Issues
Accuracy drops graduallyFailed Requests count in Device DetailsReview Protocol Issues for your protocol type
Completeness is low from the startDevice Profile tag mappingVerify all tag addresses match the PLC configuration — use Tag Browser
Consistency fluctuates rapidlyNetwork stability between gateway and deviceCheck cabling, switch health, and IP conflicts
Integrity drops for specific tagsData type configurationEnsure the tag data type in the profile matches the PLC register type

Where Quality Metrics Appear

  • Device Details: Per-device breakdown of all 5 dimensions and diagnostic counters
  • Operations Center: System-wide Data Freshness and Connectivity insights
  • UNS Explorer: Quality indicators alongside live values
  • REST API: Query device metrics programmatically for custom dashboards

Next Steps