Write a validation rule that flags a total gone wrong

A report's grand total and the sum of its own rows have drifted apart, and nobody would notice until someone does the math by hand.

IntermediateRead 5 minDoing: 45 minutes to write and prove the rule once; it runs on its own after that6 steps

▶ Start

What you'll have at the end

One rule, written as a SQL query (or a spreadsheet formula), that runs on its own and flags the exact period where a reported total stops matching the sum of its own rows.

Before you start

You need

A detail table you can query directly, like the orders you already pull with SQL, and a second number that claims to summarize it (a totals table, a dashboard tile, or a report export), sharing a date or key you can group both sides by.

Not covered

Reconciling two independently maintained reports that never shared a common source of rows, or tracking down a total that's exactly double from a join fan-out, are both a different check.

Leans on

6 steps

Where to go next

Two reports that should reconcile, and currently do not

when the two numbers come from independently maintained reports that never shared a common source of rows, not a report and its own detail table

Catch a join fan-out that doubles your totals

when the total is too high because a join duplicated rows, rather than too low because rows went missing

Write a reproducible data cleaning log

for the standing log format a flagged row from this rule is worth recording in

Checked 26 Aug 2026

Part of the Data Cleaning and Validation for Analysts cookbook