A surprising pivot total is easier to investigate when you can see the records behind that one number. Start with a precise result: a particular category and period, together with the aggregation being shown. A sum, count and average answer different questions even when they use the same source rows.

Save a working copy and note the selected value and its row and column labels. Keep the report’s filter context with that note. This prevents a later comparison from accidentally using a different group after someone changes the pivot layout.

In Numbers on Mac, select a populated pivot cell, Control-click it and choose Create Table for Source Data. Numbers adds a table on the same sheet containing the original source headings and the values used to calculate that selected result.

Reconcile the result with its contributing records

Inspect record identifiers in the new table. Look for an unexpected category, a repeated record or a value that belongs to a different period. These observations are more useful than simply confirming that a detail table was created.

For a small practice sum, use a pivot result whose contributing quantities are 4, 7 and 9. The expected total is 20. Compare that arithmetic with the selected pivot value and read all three record identities. If the real report uses an average, calculate the relevant mean instead of adding values and expecting agreement.

If an individual source value is wrong, locate the authoritative record in the original source before correcting it. Do not assume editing the newly created explanatory table will repair the source or the pivot. Keep the extracted detail as evidence of what you inspected at that point.

Follow the discrepancy to its cause

If the contributing rows are all correct but the result seems wrong, revisit the aggregation and the meaning of the requested figure. For example, a count of entries is different from a sum of quantities, and a record may legitimately appear more than once for separate events.

After an authorised correction in the appropriate source and report workflow, repeat the selected-value investigation. Compare the new result and contributing records with the original note. Label any retained extracts by the report state they explain, and remove obsolete working copies from the release package. The useful outcome is a traceable explanation of one figure, not merely another table containing similar data.

Reference: Apple: source data for a Numbers pivot value.