Organize information

Analyze missing values in Excel

Measure missingness by field and group before deciding whether to leave, recover, or exclude it

7 min setupTested with:ChatGPTReviewed: 2026-08-28

Add your context

Your text stays in this browser. AILesson Prompts does not send it to a model or server.

Analysis or import purposeFields and missing-value meaningsCounts or representative rows

Your prompt

Unfilled fields remain visible as placeholders, so you can still copy and edit the prompt

Analyze and plan treatment for missing Excel values.

Dataset purpose:
[purpose]

Fields and meanings of missing values:
[fields]

Counts, groups, and representative rows:
[profile]

Distinguish true blank, unknown, not applicable, zero, empty string from a formula, and missing because of a data failure. Do not fill values with averages, zero, or guessed categories unless the decision rule and impact are justified. Return a field-by-field missingness profile, group comparisons, likely causes labeled as hypotheses, recommended action, validation formulas, and sensitivity checks showing how exclusions or imputations change the result.

Private by defaultPrompt assembly happens locally in your browser. Avoid placing confidential information into any AI service unless your organization allows it.

From input to outcome

A worked example

See how concrete context turns this recipe into a usable result

Actual input

Analysis or import purpose: Report average ticket resolution hours by support channel. ResolutionHours is used only for closed tickets.
Fields and missing-value meanings: Status is required; Channel is required; ResolutionHours is required for Closed, not applicable for Open, and may be blank from an export failure. Text 'N/A' means not applicable; numeric 0 means resolved immediately and is valid.
Counts or representative rows: 1,000 rows: 800 Closed, 200 Open. Closed has 40 blank ResolutionHours: 32 are Chat, 8 Email. Open has 200 N/A. Three Closed rows have 0. A source-system change occurred midway through the month.

Example output

Profile missingness only among rows where the field is applicable. Open/N/A rows are not missing resolution data and must not enter the average denominator. The three numeric zeros are valid. Among Closed tickets, the missing rate is 40/800 = 5%; its concentration in Chat is evidence of uneven missingness, but the source-system change is only a causal hypothesis until dates are compared.

Add MissingClass: =IFS([@Status]<>"Closed","Not applicable",ISNUMBER([@ResolutionHours]),"Observed",[@ResolutionHours]="","Missing-required",TRUE,"Invalid-marker"). Build a PivotTable of MissingClass by Channel and by period before/after the system change.

Recommended action: do not mean-fill. Attempt recovery from the source for 40 Missing-required rows; until then, report observed averages with the observed/missing count for each channel. Sensitivity check each channel using at least three bounds: observed-only mean, missing values set to the channel's observed 25th percentile, and set to its 75th percentile. If channel ranking changes, state that the comparison is not robust. Reconcile 800 Closed = Observed + Missing-required + Invalid-marker.

Why this works

1
Separating missing meanings prevents zero and not-applicable values from being treated as unknown.
2
Sensitivity checks show whether a cleanup choice materially changes the reported result.