Blog · Data quality and reconciliation
A dashboard audit checks whether reported measures match their source rows and agreed definitions. Follow a reconciliation from exported records to a corrected dashboard total, with spreadsheet formulas and an evidence checklist.
A dashboard audit checks whether reported measures agree with their source rows, definitions, and intended decisions. Start by fixing the reporting period and filters, then recompute the figures independently. A matching grand total is necessary, but it does not prove that every customer, queue, or percentage is right.
Keep the dashboard export and its source files together with the extraction time, selected filters, reporting cutoff, and responsible owner. Record whether the export contains transaction rows, invoice lines, or already summarized results. A total copied from another report is not independent evidence if both reports use the same incomplete extract.
For each measure, write down its business question, formula, unit, period, inclusion rules, and source columns. State how credits, cancellations, blanks, and transfers are handled. Use versioned metric definitions when a definition changes so the audit can explain which version produced each result.
Decide what constitutes a pass before comparing numbers. Financial totals should reconcile to the appropriate precision; a known display-rounding difference needs a stated tolerance. An unexplained difference does not pass because it looks small.
Consider a US distributor reviewing September 2026 net revenue for its East business unit. The source file has one row per document line, including credits. Its four relevant rows are:
| Document, column A | Line, B | Net amount (USD), C | Posting date, D | Unit, E |
|---|---|---|---|---|
| INV-101 | 1 | 120,000 | September 5, 2026 | East |
| INV-102 | 1 | 80,000 | September 10, 2026 | East |
| INV-103 | 1 | 60,000 | September 15, 2026 | East |
| CR-104 | 1 | -10,000 | September 20, 2026 | East |
In Google Sheets, place those headers in row 1 and records in rows 2 through 5. Dates must be real date values. Calculate the independently selected total:
=SUMIFS(C2:C5,D2:D5,">="&DATE(2026,9,1),D2:D5,"<"&DATE(2026,10,1),E2:E5,"East")
This returns $250,000: $120,000 + $80,000 + $60,000 - $10,000. Google's SUMIFS documentation describes summing rows that meet multiple criteria. The exclusive next-month boundary also includes records with a time component on the final reporting date.
Now inspect the dashboard's exported detail:
| Dashboard row | Document | Line | Net amount (USD) |
|---|---|---|---|
| 1 | INV-101 | 1 | 120,000 |
| 2 | INV-102 | 1 | 80,000 |
| 3 | INV-103 | 1 | 60,000 |
| 4 | INV-103 | 1 | 60,000 |
| Reported total | 320,000 |
The dashboard overstates net revenue by $70,000. Its detail repeats an invoice line and excludes the credit. Neither problem is visible from the headline total alone.
Build an explicit bridge rather than changing the dashboard until the totals happen to agree.
| Reconciliation step | Adjustment (USD) | Running total (USD) |
|---|---|---|
| Start with dashboard | 320,000 | |
| Remove extra INV-103 line | -60,000 | 260,000 |
| Include CR-104 | -10,000 | 250,000 |
| Compare with source | 0 | 250,000 |
The identity is $320,000 - $60,000 - $10,000 = $250,000. Keep the duplicated and missing row identifiers with the evidence. For a broader workbook review, use the five checks before trusting a spreadsheet total.
To flag repeats, create a document-line key in F2 on a detail sheet where document and line occupy A and B:
=A2&"|"&B2
Fill down, then put this in G2 and fill down:
=COUNTIF($F$2:$F$5,F2)
Google's COUNTIF documentation explains the conditional count. A result greater than one flags a repeated key; it does not authorize removing every flagged row. Confirm the source grain and preserve one valid occurrence. Include entity in the key if document numbers repeat across companies, and choose a separator that cannot appear in the identifiers.
Covirage is not affiliated with Google.
Trace a percentage to both of its components. A returned-order rate needs returned orders and eligible orders from the same population. Inspect whether the dashboard counts lines while its label says orders. Compare the unrounded calculation with the displayed percentage before investigating a rounding difference.
Check classifications too. A sales pipeline can contain the right total value while assigning opportunities to inconsistent stages; pipeline stage definitions covers the business rules behind that classification.
Finally, ask the reader what action a flagged result triggers. Record the exception owner, evidence required, and next review date. A measure with no agreed response may belong in supporting detail rather than the main view. The audit committee reporting dashboard addresses the separate question of organizing oversight information for that audience.
Bring your dashboard detail and the corresponding business export, with the period and definitions you want to check. Covirage maps the columns, and its tools compute and check the figures; the external AI model explains the differences 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.
Audit a dashboard before its first business use and after a source, definition, or reporting filter changes. Repeat reconciliations during normal reporting cycles. The frequency should reflect how often the underlying files change and how consequential the decisions are.
A data audit examines the underlying records and their quality. A dashboard audit also tests the presentation: which rows a filter includes, which definition a measure applies, and whether the displayed result supports the decision readers believe it supports.
Yes. A sales total can reconcile while a margin uses gross billings instead of net revenue. A dashboard can also apply a different date or customer population to the numerator and denominator. Audit both the arithmetic and the business definition.