Blog · Board and management reporting
An EDA dashboard checks the shape and completeness of a business export before you use it for decisions. Follow a worked invoice example with spreadsheet formulas for missing values, distributions, segment differences and unusual transactions.
An EDA dashboard shows whether a business export is complete, how its values are distributed and which segments deserve a closer look. Build it before relying on averages or publishing comparisons. Each finding should lead back to the records behind it.
For this example, a US distributor reviews September 2026 invoices. Each row is one posted invoice, sales are net of credits, and all amounts are USD. The export is already restricted to the same posting period and legal entity.
Put the headers below in spreadsheet cells A1:D1 and the invoices in rows 2:11. Leave the cost cell for INV-104 genuinely empty; the word "blank" in the table is an instruction, not a value to type.
| Invoice | Region | Net sales (USD) | Recorded cost (USD) |
|---|---|---|---|
| INV-101 | East | 200 | 120 |
| INV-102 | East | 250 | 150 |
| INV-103 | East | 300 | 180 |
| INV-104 | East | 350 | blank |
| INV-105 | East | 400 | 240 |
| INV-106 | West | 450 | 270 |
| INV-107 | West | 500 | 300 |
| INV-108 | West | 550 | 330 |
| INV-109 | West | 600 | 360 |
| INV-110 | West | 2,400 | 1,440 |
Keep an unchanged copy of the export. Put the rows and calculations together on a review worksheet so the references below match. Agree on dashboard standards for period, currency and exclusions before comparing this file with another month's extract.
These Excel formulas return the record count, missing cost count and missing cost rate:
=COUNTA(A2:A11)
=COUNTBLANK(D2:D11)
=COUNTBLANK(D2:D11)/COUNTA(A2:A11)
The results are 10 invoices, one missing cost and 10.0% missing cost. Microsoft documents COUNTBLANK as counting empty cells, including formulas returning empty text; numeric zero is not blank. Also check numeric coverage with =COUNT(C2:C11) so a sales amount stored as text does not escape review.
In E2, enter the following duplicate flag and fill down. Every result should be zero for this invoice-level export:
=IF(COUNTIF($A$2:$A$11,A2)>1,1,0)
Repeated invoice IDs require investigation, rather than automatic deletion: a different export might legitimately contain several lines per invoice. Reconcile the 10 accepted invoices to the source register. The reconciliation definition explains why a matching total alone cannot prove that the correct records were included.
Use these formulas on the sales column:
=SUM(C2:C11)
=AVERAGE(C2:C11)
=MEDIAN(C2:C11)
=QUARTILE.INC(C2:C11,1)
=QUARTILE.INC(C2:C11,3)
| Measure | Calculation from the invoice rows | Result (USD) |
|---|---|---|
| Total sales | Sum of all sales | 6,000 |
| Mean invoice | 6,000 / 10 | 600 |
| Median invoice | (400 + 450) / 2 | 425 |
| First quartile | 300 + 0.25 × (350 − 300) | 312.50 |
| Third quartile | 500 + 0.75 × (550 − 500) | 537.50 |
| Interquartile range | 537.50 − 312.50 | 225 |
The quartiles use the inclusive percentile convention in Microsoft's QUARTILE.INC documentation. Keep that convention consistent across periods. Currency summaries are shown to the cent where needed; percentages below are rounded to one decimal place after calculation.
A sorted invoice chart would show a long upper tail. The mean exceeds the median because the largest invoice pulls the mean upward. Neither statistic replaces the total when the question is revenue delivered.
The NIST box plot guide describes fences based on the interquartile range. Use them to identify records for review. The upper fence here is 537.50 + 1.5 × 225 = $875; the lower fence is 312.50 − 1.5 × 225 = −$25.
In F2, flag sales outside those fences and fill down:
=IF(OR(C2>QUARTILE.INC($C$2:$C$11,3)+1.5*(QUARTILE.INC($C$2:$C$11,3)-QUARTILE.INC($C$2:$C$11,1)),C2<QUARTILE.INC($C$2:$C$11,1)-1.5*(QUARTILE.INC($C$2:$C$11,3)-QUARTILE.INC($C$2:$C$11,1))),"Review","")
Only INV-110 is flagged. Its $2,400 is 2,400 / 6,000 = 40.0% of total sales. Ask sales operations to confirm its contract and posting before interpreting the regional difference.
| Region | Invoices | Sales (USD) | Mean invoice (USD) | Missing costs |
|---|---|---|---|---|
| East | 5 | 1,500 | 300 | 1 |
| West | 5 | 4,500 | 900 | 0 |
| Total | 10 | 6,000 | 600 | 1 |
For East, use =SUMIF(B2:B11,"East",C2:C11) for sales and =COUNTIF(B2:B11,"East") for invoices; divide those results for the mean. Change the criterion to West for the other region. The regional totals reconcile: 1,500 + 4,500 = 6,000. Without INV-110, the remaining invoices total $3,600 and average $400; that sensitivity helps explain the total without erasing a valid sale.
Covirage is not affiliated with Microsoft.
Upload a sample of your own invoice workbook to Covirage, which maps the columns so its tools can compute and check the figures. The external AI model explains the results with every figure traced to its rows. Create a free Covirage account, which takes about a minute, then set up your business and bring your file.
See self-service analytics for the supporting business workflow.
Include the file's scope, record count, missing-value counts, duplicate checks, numerical distributions and segment comparisons. Show the unit and denominator beside each measure. An exception list should identify which original records need review and who can confirm them.
Keep unusual records until someone verifies them. A large invoice can be a valid contract order, a duplicate, or an entry error. Record the reason for any correction or exclusion and retain a reconciliation to the original export.
It can identify a difference and suggest questions to investigate. It cannot establish a cause from a segment comparison alone. Check customer mix, period coverage, transaction types and missing information before attributing the difference to a manager or process.
The analyst documents the checks and exceptions. The business owner confirms whether unusual records are valid, while the source-file owner corrects missing or inaccurate fields. Finance approves changes that affect reported revenue, cost or margin.