Data quality dashboards: checks, exceptions, and owners

Data quality dashboards show which business-file checks failed, how many records are affected, and who owns the correction. A useful dashboard separates missing fields, duplicate records, invalid values, and reconciliation failures. It keeps the underlying rows available so the owner can resolve the exception.

The aim is fitness for a decision. GAO's data reliability guidance describes reliability through accuracy, completeness, and applicability for the intended purpose. A complete invoice file can still be unsuitable for a customer review if the customer identifiers are missing.

Set the rules before choosing the display

Start with the export, its reporting period, and the decisions it supports. An invoice review needs different checks from a purchase-order review. Do not apply a positive-amount rule to a file that legitimately includes credit memos.

For this worked example, a US distributor reviews an invoice-only export for July 2026. It contains original invoices, excludes credits, and uses USD throughout. The required fields are invoice identifier, customer identifier, amount, and status. Approved statuses are Posted and Pending. An invoice identifier must be unique within its source.

Write these rules beside the measures. The Census Bureau's statistical quality standards illustrate the broader principle that information quality needs explicit requirements. The business checks below are company definitions for this file, rather than a claim to meet federal statistical standards.

Keep metric definitions consistent with the reporting package. If a sales measure excludes pending invoices, use the same filter when checking the inputs to the sales KPI definitions.

Work through the file, row by row

The row number is a stable reference within this exported file. A dash in the customer column represents an empty field, not a customer named Dash.

For duplicate counting, retain the first occurrence in file order and flag later occurrences as surplus. That makes row 3 the surplus duplicate. It does not establish which record is correct; the owner must resolve that from the source document.

Check failure rate = Failed eligible rows / All eligible rows

Affected-row rate = Distinct rows failing any row check / All eligible rows

All eight rows are eligible for each row check here. Percentages are exact; later examples round percentages to one decimal place when needed.

There are four failed checks across three affected rows. Adding the three individual percentages would count row 3 twice. The clean-row count is 8 minus 3 = 5, and the clean-row rate is 5 / 8 = 62.5%.

Reconcile sources separately

The controller supplies independent invoice control totals. Compare them with the file before deciding whether its amounts are usable. Here, a positive difference means the file is above its control.

The North total is 1,200 + 800 + 800 + 500 = $3,300. South is 1,000 + 700 + 600 + 400 = $2,700. Their differences add to 800 minus 100 = $700, which also equals 6,000 minus 5,300.

The North difference matches the surplus invoice amount, making that record a sensible investigation starting point. It is not proof that removing it is authorized. South remains short by $100, even if the North duplicate is resolved. Assign the South difference to its billing owner and request the source documents.

Do not label both differences as erroneous rows: the missing South record is not present in the file. The coverage worked example shows the same distinction between records observed and the population expected.

Change one input and see what improves

Now the billing lead supplies the customer identifier for row 2. Change only that cell. Row 3 remains blank and duplicated; row 4 still has an invalid status.

Completeness improves without any financial reconciliation improvement. That is why the dashboard needs separate checks and a visible explanation of what changed.

Give the review a usable layout

Put reporting period, file name, received date, eligible row count, and rule version above the measures. Below them, show the check results, then source totals, then an exception list with row reference, owner, next action, and review status. An owner should be able to move from a failed count to the records that produced it.

Keep business delivery measures in the operational dashboard. Keep product or service inspection outcomes in the quality assurance dashboard. This review checks whether the files supporting those decisions are usable.

Where it goes wrong

• Counting all duplicate-group members as surplus. State whether the count means all members or only extra records, then keep that definition stable.
• Treating every blank as an error. Some fields are optional or inapplicable. Define eligibility before calculating the denominator.
• Combining errors into an unexplained score. Separate the failed-check count, affected-row count, and business impact.
• Checking only the company total. Source differences can offset each other, hiding the location and scale of problems.
• Editing the export without evidence. Preserve the original reference and get the correction from the responsible owner before rerunning the checks.

Check your own business file in Covirage

Bring an invoice, order, or operating export with the control totals and rules you want checked. Covirage maps the columns, and its tools compute and check the figures; the external AI model explains the results with every figure traced to its rows. Create a free Covirage account and bring a sample of your own data.

See metrics governance for the supporting approach to consistent business measures.