Make each dashboard number traceable by giving it a written definition, a saved query, a timestamp and a click path to the underlying records. When a reader can go from "14 overdue" to the 14 findings, with their dates and owners, the dashboard becomes evidence instead of decoration.
Why traceability matters
Auditors and customers do not accept a number they cannot check. A screenshot shows what a tool displayed once. A traceable figure shows how it was computed, from which records, at what time. It also protects you internally: when two leaders quote different numbers, the trail tells you why.
The four parts of a traceable metric
| Part | What it contains |
|---|---|
| Definition | Plain language rule, clock start, inclusions, exclusions |
| Query | The saved filter or code that produces the number |
| Snapshot | Timestamp, and ideally a stored copy of the result |
| Drill down | A link from the number to the individual records |
Steps
- Write a definition card for each metric on the dashboard. Keep it short: name, formula, data source, owner, refresh time.
- Store the query in version control or in the tool, with a change history.
- Stamp each view with the data refresh time, not the page load time.
- Add drill down links, so every tile opens the filtered list.
- Snapshot the numbers at fixed points, such as month end, and store them with a hash or in a write protected location. Reports for the period should be built from the snapshot, not recalculated later.
- Record changes to definitions, with dates, so trend lines do not silently shift.
Completeness and accuracy
Auditors often ask whether a report from a system is complete and accurate. Be ready to explain: what population the report covers, what filters were applied, and how you know records were not missing. A reconciliation, such as repositories scanned compared with repositories in the source control inventory, supports the claim.
Common failures
- A metric defined differently by two teams.
- A chart built from a spreadsheet copy that nobody can refresh.
- Rounded or capped values with no raw data behind them.
- Definitions changed after a bad quarter.
A quick test
Pick any number from last quarter's report. Can a colleague reproduce it in ten minutes using only the definition and the saved query? If not, fix that metric first.
Quick check
Is a dashboard screenshot enough evidence?
Usually not alone. Pair it with the definition, the query and the underlying records or a stored snapshot.
What if we change a metric definition?
Record the date and reason, and restate prior periods or mark the break in the trend line.
Safeguard supports this with compliance and evidence automation, and you can run a free scan at gold.safeguard.sh/scan.
Last reviewed: 19 October 2026.
