PI System Data Governance: Best Practices for Reliable Operational Data

Data governance in a PI System environment is not mainly a documentation exercise. It is the set of controls that keeps operational data understandable, traceable, and safe to use as the system changes.

A useful governance program covers source data, PI Points, AF models, calculations, displays, ownership, and change control.

1. Define data ownership

Every important data domain should have an accountable owner.

Ownership should be clear for:

• Source-system signals

• PI Point creation and configuration

• AF templates and hierarchies

• Calculations

• Critical displays and reports

• Data-quality exceptions

The owner does not need to perform every task. The owner must know who can approve changes and who must respond when data is unreliable.

2. Standardize PI Point creation

Define minimum requirements for new points.

Include:

• Naming rules

• Description

• Engineering units

• Source and point source

• Expected update rate

• Exception and compression approach

• Security

• Owner or system of record

Do not allow projects to create large tag sets without a handover and retirement plan.

3. Govern AF as a semantic model

AF should provide stable operational meaning above source-specific tag names.

Use templates for repeatable equipment. Keep attribute names and units consistent. Validate data references before large deployments.

Control template changes because one change can affect many assets and analyses.

4. Control calculation changes

Calculations can create trusted business and operational values. Treat them as governed logic.

For critical analyses, record:

• Purpose

• Inputs

• Output

• Owner

• Schedule

• Expected behavior for missing or bad data

• Change history

Before a major change, identify downstream dependencies.

5. Maintain lineage and impact visibility

Teams should be able to trace an important value from its source through AF, calculations, and displays.

This supports troubleshooting and safer change management.

Lineage is especially important during migrations, server retirement, tag cleanup, and AF redesign.

6. Monitor data quality continuously

Governance must include operating controls, not only standards.

Monitor conditions such as:

• Stale data

• Bad values

• Unexpected flatlines

• Data gaps

• Broken references

• Analysis failures

• Configuration changes

Prioritize issues by asset criticality and downstream use.

7. Define a retirement process

Old tags, analyses, displays, and AF objects create risk when nobody knows whether they are still used.

Before retirement:

1. Identify dependencies.

2. Confirm the owner.

3. Record the reason for retirement.

4. Define the effective date.

5. Retain required history.

6. Verify downstream systems after the change.

8. Treat compression and metadata as governance

Engineering units, span, exception settings, and compression settings affect how a signal is interpreted and stored.

These settings should not change without review for important points.

9. Make governance practical

A governance program fails when every small change requires a large committee.

Use clear standards, automated checks, and risk-based approval. Reserve formal review for changes with material operational or downstream impact.

The goal

Good PI governance gives engineers confidence that they can understand a value, trace its source, identify its owner, and change the system without unknown consequences.

That is the foundation required for reliable dashboards, analytics, cloud integration, and industrial AI.