→ Finance
FinanceIntermediate

Spreadsheet Anomaly Review

Use when you need a structured second pair of eyes on a spreadsheet to catch anomalies, broken formulas, and suspicious numbers before they reach a decision-maker.

Finance LeadersFP&A Teams
Save to My LibraryAgent Skills compatible

Spreadsheet Anomaly Review

A wrong number in a spreadsheet becomes a wrong number in a board deck. This skill runs a structured anomaly review so errors are caught before they propagate.

When to use this skill

Use this skill when:

  • A model or report is about to be used for a decision or shared upward.
  • You inherited a spreadsheet and do not fully trust it.
  • You want a repeatable QA pass before month-end close.

Inputs needed

  • The spreadsheet data (exported as text, CSV, or pasted ranges).
  • The expected relationships (e.g., "totals should sum the rows above").
  • The plausible ranges for key metrics, if known.
  • The period and any prior-period values for comparison.

Process

  1. Identify the key totals and subtotals and the rows they should reconcile to.
  2. Check that each total equals the sum of its components.
  3. Compare each key metric against its plausible range and prior period.
  4. Look for sign errors, transposed digits, and unit mismatches.
  5. Flag blanks where a value is expected and constants where a formula is expected.
  6. Rank findings by potential dollar impact.
  7. Recommend a specific fix or check for each high-impact finding.

Prompt or workflow

You are reviewing a spreadsheet for anomalies before it is used to decide.

Period: [PERIOD]
Expected relationships:
"""
[e.g., "Revenue total = sum of product lines"; "Margin = (Rev - COGS) / Rev"]
"""
Plausible ranges / prior values:
"""
[OPTIONAL]
"""
Data:
"""
[PASTE RANGES OR CSV]
"""

Produce:
1. RECONCILIATION CHECKS: each total vs the sum of its parts (pass/fail).
2. RANGE CHECKS: metrics outside plausible range or far from prior period.
3. STRUCTURAL ISSUES: blanks where values expected, hardcoded values where
   formulas expected, sign/unit/transposition errors.
4. RANKED FINDINGS: ordered by estimated dollar impact.
5. FIX FOR EACH: a specific check or correction.

Rules:
- Show the arithmetic for any reconciliation you claim fails.
- Do not flag a number as wrong without stating the expected value or range.
- Rank by impact, not by how easy it is to spot.

Quality checklist

  • Every claimed error shows the expected value or the failing arithmetic.
  • Totals are reconciled against their components.
  • Key metrics are checked against ranges or prior periods.
  • Structural issues (blanks, hardcodes, sign errors) are checked.
  • Findings are ranked by dollar impact.
  • Each high-impact finding has a specific recommended fix.

Common mistakes

  • Flagging a number as "wrong" without saying what it should be.
  • Checking formatting while missing a broken subtotal.
  • Treating every anomaly as equally urgent.

Example output

Reconciliation: Q2 revenue total 4.80M != sum of lines 4.62M (FAIL, -0.18M).
Range check: gross margin 71% vs prior 58% — implausible jump, verify COGS.
Structural: cell for July headcount is hardcoded 42, breaks the rollforward.
Ranked: revenue mismatch (~$180k) first; margin spike second.
Fix: re-link revenue total to the product-line range; confirm COGS mapping.
  • Finance Variance Narrative Generator — once the numbers are trusted, explain them.
  • Board Deck Insight Extractor — to vet the numbers a deck presents.

Attribution

This skill was created by Vectory and is licensed under CC BY 4.0.

Source: https://vectory.io/skills/spreadsheet-anomaly-review

Attribution: "Spreadsheet Anomaly Review" by Vectory.