Evidence needs provenance, not just attachment

Organisations collect large amounts of evidence.

Inspection reports.

Photographs.

Laboratory results.

Engineering assessments.

Certificates.

Emails.

Spreadsheets.

System records.

Historian data.

Meeting papers.

Many systems allow those items to be attached to an incident, action or decision.

That is useful.

But attaching a file is not the same as understanding the evidence it contains.

When evidence supports a material governance decision, context matters.

An attachment is not necessarily evidence

Consider a PDF called:

Inspection_Report_Final_v3.pdf

It may contain exactly the information required to support a decision.

But the application may know very little about it.

Which equipment was inspected?

When?

By whom?

Using what method?

Was the inspector competent?

Was the instrument calibration current?

Which revision is authoritative?

What operating condition existed at the time?

What conclusion from the report is being relied upon?

Those details are often inside the document.

If they are not represented structurally, every later user has to reconstruct them.

That becomes difficult when several pieces of evidence contribute to the same decision.

Evidence needs identity

A useful evidence record might describe:

• source,
• subject,
• observation,
• effective time,
• method,
• author or producing system,
• version,
• authority,
• quality,
• and provenance.

For example:

Subject: P-101 elbow E17 Observation: wall thickness 6.8 mm Method: ultrasonic thickness measurement Measurement time: 31 August 2026 Instrument: UT-17 Calibration status: current Source system: inspection management Report: IR-771

The document can still exist.

But the organisation now has a structured statement about what that document contributes.

This allows evidence to participate in the operating model rather than remaining only an attachment.

Provenance matters when information crosses systems

Suppose two applications contain different values for the same condition.

One says:

Risk status: Medium

Another says:

Risk status: High

The immediate temptation is to use the latest timestamp.

But the more important question may be:

Which application has authority to make this statement?

The answer depends on provenance and semantics.

One system may contain a copied value for convenience.

Another may be the authoritative source.

The Orchestrator can preserve that distinction when information moves.

A governance application can then know whether a decision relied upon:

• an authoritative record,
• a replicated record,
• a manually entered value,
• a calculated value,
• or an external document.

That difference becomes important if the decision is later challenged or reviewed.

Evidence has a period of validity

Evidence is not always permanently true.

A geological interpretation may remain relevant until new information changes it.

A gas test may be valid for a very short period.

A calibration certificate has an expiry.

An engineering assessment may remain valid only while operating assumptions stay unchanged.

A risk assessment may become invalid when a temporary control fails.

A contract clause may be authoritative until the contract is varied.

A useful evidence model therefore needs to ask:

When was this evidence valid?

That may be different from:

When was this record created?

The distinction matters.

A decision made today may rely on evidence collected last week.

That may be perfectly reasonable.

Or it may be unacceptable because the underlying condition changes hourly.

Evidence should be referenced, not buried

There is another architectural benefit to treating evidence as a first-class object.

Instead of putting copies of the same document into several workflows, different records can point to the same evidence.

For example:

Evidence E102 — inspection result Evidence E147 — engineering assessment Evidence E221 — operating history

A decision can reference all three.

An obligation verification may later reference E147 and E221.

An auditor can see exactly which records relied upon each item.

That creates an evidence graph rather than a collection of attachments.

It also preserves provenance.

If Evidence E147 is superseded, the system can identify which decisions relied upon the earlier assessment.

That can be far more valuable than simply replacing a file.

Evidence quality can itself become governable

Not all evidence has equal weight.

Some may be:

• provisional,
• independently verified,
• manually observed,
• automatically measured,
• estimated,
• externally certified,
• disputed,
• superseded.

That does not mean every organisation needs an elaborate evidence scoring model.

It means the status of important evidence should be visible where it affects decisions.

A provisional laboratory result and a certified result may support different levels of confidence.

An inspection completed outside the intended method may still be useful, but its limitation should travel with it.

The objective is not to create bureaucracy around every data point.

It is to preserve the context that matters when evidence becomes material.

Evidence tells us what we know, not what we should do

Once provenance is preserved, the organisation has a much stronger basis for governance.

It can ask:

What did we know at the time?

How reliable was it?

Which assumptions depended on it?

Which decision used it?

That improves traceability.

But evidence alone still does not decide what should happen.

The same observation may be acceptable in one operating context and unacceptable in another.

A minor technical deviation may create a major compliance consequence.

A serious-looking alarm may be well controlled by redundancy.

Understanding the evidence is therefore only one part of the governance chain.

The next question is risk.