A ten-item check to run before any report is issued, designed to be completed in minutes rather than hours. It is the practical companion to Data-Quality Checks for Report Production, and it exists because the defects that reach readers are almost always found by a short, repeated check rather than a thorough, occasional one.

Run it on the frozen snapshot, sign it, and keep it with the report. The record is what turns the check from a good intention into a control.

A report is issued on a snapshot. Check the snapshot, not the live system, and check it every cycle.

The checklist

Before you draft

  • 1. Source. Does every headline figure have a named source and extract?
  • 2. Definition. Is each metric defined in one agreed way, per the dictionary?
  • 3. Period. Is the period explicit and consistent across every figure?
  • 4. Completeness. Does the data cover every expected entity and line?

After you draft

  • 5. Totals. Do subtotals sum to totals, and do the totals agree across exhibits?
  • 6. Reconciliation. Do the figures agree with the prior period, and with any sibling report?
  • 7. Units and currency. Are units and currency stated and consistent?
  • 8. Outliers. Are the unusual figures explained, or flagged for review?
  • 9. Labels. Are estimates, forecasts and provisional figures labeled as such?
  • 10. Version. Is the snapshot recorded, and the report versioned before issue?

Ten checks. If a report fails one, fix or disclose it before issue – never override silently.

How to use it

  • Run it in order. Checks 1-4 catch the errors that are expensive to fix later.
  • Automate what you can. Totals, completeness and thresholds can be scripted.
  • Keep the judgment checks human. Whether a cause is true or an outlier is real is a judgment.
  • Sign and retain. An unsigned checklist is a habit; a signed one is a control.
  • Record failures, so the cause is fixed and the check improves.

The full rationale for each check, and the six dimensions of data quality behind them, is in the linked guide.

Automation

The mechanical checks are the ones to automate first, because they are the ones that get skipped under deadline.

  • Totals and subtotals – compare each total to the sum of its parts.
  • Completeness – compare the data to an expected list of entities or lines.
  • Prior-period comparison – flag changes beyond a threshold.
  • Outlier detection – flag values outside a normal range.
  • Units and format – validate against the metric dictionary.

Leave the judgment checks to a person, and automate the rest. A check that runs in seconds will run every cycle.

Frequently asked questions

How long should the check take?

About ten minutes on a frozen snapshot. The aim is a habit run every cycle, not an audit run occasionally.

Can the checklist be automated?

The mechanical checks can – totals, completeness, prior-period comparison, outlier detection and format validation. The judgment checks remain human.

What should you do if a check fails?

Stop, trace the failure to its source, decide whether to correct the data, correct the report or disclose the limitation, and record the failure and the fix. Never override silently.

Should the checklist be signed?

Yes. A signed checklist, retained with the report, is what shows the report was checked and who checked it.

Does the checklist replace a second reviewer?

No, it reduces what the reviewer has to catch. The mechanical check finds the broken total; the reviewer finds the sentence that contradicts the number.

Can it be used for reports you inherit?

Yes. The checks do not depend on knowing the report’s history, only on it having named sources and a stated period. Where those are missing, establish them first.

Next step

Run the ten checks on your next cycle, sign it, and retain it with the report. If you would like the checks installed in your process, book a pilot call to run two cycles with your team.

Sources

  • The checklist reflects standard data-quality dimensions – accuracy, completeness, timeliness, consistency, validity and uniqueness – applied to report production.

No statistic in this asset is invented; where a figure appears in the linked guides, it is cited there with its source and date.