A percentage can arrive in a decision paper looking settled. The calculation may be reproducible, the chart may be clear, and the number may still be answering a different question from the one the reader thinks it answers.
Which population does it describe? Over what period? What was excluded? Has anyone checked the source, or only the arithmetic?
In my metric-passport note, I argued that a number should retain its origin, method, boundaries, review status and dependencies as it moves between organisations. Here, a metric passport means a linked explanation of a metric and its permitted use—not a certificate that the number is correct.
The same observations can answer different questions
The issue is not to declare that one aggregation is always right. It is to state which question the calculation answers. A report about the average plot and a report about the combined planting cohort should not silently exchange their indicators.
Neither result, by itself, describes an entire landscape. Nor does seedling survival establish a biodiversity, livelihood or carbon outcome. Those would be separate claims requiring their own definitions and evidence.
What needs to travel with the metric?
The W3C PROV data model provides a way to describe entities, the activities that produced or transformed them, and the agents associated with those activities. Its Data on the Web Best Practices also address metadata, provenance, version identifiers and change histories.
The public example below builds on those ideas by making the intended decision use explicit. It is an editorial illustration, not a W3C-defined passport or an institutional review instrument.
In a real record, the plot boundaries, observation dates, method version and responsible people would need to be identifiable to the appropriate reviewer. Public visibility is a different question: personal information, sensitive locations and restricted source material should not be exposed merely to make a reference look complete.
Now change one source value
Suppose the Plot B reviewer discovers that two replacement seedlings were counted as original survivors. The original-cohort count is eight, not ten. The corrected combined result is (80 + 8) ÷ 120 = 73.3%, rounded to one decimal place.
Updating a dashboard cell would fix the display, but not necessarily the monitoring note or decision paper that used 75%. The correction needs a route through those dependencies.
First, retain B-01:v1 and create B-01:v2 with the reason for the correction and the reviewer’s identity. Then derive M-07:v2 from the corrected input. Mark the earlier versions as superseded for current use while retaining the history needed to understand previous work.
Next, notify the owners of N-03:v1 and D-02:v1. Ask what the correction changes in each document: a descriptive sentence, a threshold assessment, a proposed action, or nothing material. Record that assessment rather than treating every dependency as automatically invalid.
A historical decision should not be silently rewritten. Preserve what its reviewer saw at the time, then link the new evidence and any reconsideration. Where continued use needs to be restricted, the responsible authority should state the scope and next review step.
A source can change without changing the number
Imagine that the arithmetic remains 75%, but the observations turn out to cover a different visit date. The displayed value has not moved; the claim about six-month survival no longer has the same basis.
This is why the correction route should consider changes to definitions, dates, geographic coverage and review status—not just numerical differences. A change log that records only values could miss the very change that matters to the decision.
Likewise, a list of known downstream uses is not proof that every copy has been found. An exported slide or forwarded document may sit outside the maintained record. Give exports a visible version reference and a correction contact, and record the limits of the dependency check.
Keep the record proportionate
The passport should not require the same evidence to be retyped into every report. A concise reference can point to controlled source records, calculations and review decisions. The purpose is to make the route inspectable, not to create another disconnected form.
For a low-consequence internal discussion, a linked calculation note may be sufficient. A consequential external claim may need additional review and a specific authority to release it. The relevant institution determines that requirement; the existence of a passport does not settle it.
Where this sits in the framework
Fracture is the gap between a number and what it purports to show. Signal gives the discrepancy an identifiable record. Spine connects the correction to the people responsible for dependent uses. Assurance asks whether a later reviewer can reconstruct both the earlier basis and the response to change.
Before carrying a metric into another document, ask: Can its next reader identify the question it answers—and find the person responsible when that answer changes?
About this version
This article was prepared on 24 September 2026. It develops the author’s linked public writing with new explanatory examples. The figures, identifiers and scenarios are fictional; they are not programme evidence. References were checked on the preparation date.
Fracture → Signal → Spine → Assurance