A green quality bar over an imported column is useful evidence only when you know which rows it describes. An error near the end of a large export can be absent from the small preview you first inspected.

Choose the field whose correctness matters to the report, such as a required numeric quantity. Keep a copy of the source and identify a known record late in the file. Record the transformation step being examined: an earlier text column and its later numeric conversion answer different questions.

Make the inspected population explicit

In Power Query Desktop, enable the profiling tools from View, Data Preview. Microsoft documents a default scope of the first 1,000 rows. Use the profiling-scope message in the lower-left corner to switch to the entire dataset. Column Quality distinguishes valid, error and empty values, with additional unknown or unexpected-error states.

Wait for the requested profile to finish and inspect its result at the relevant step. Do not record a full-file conclusion while the displayed scope still says the top 1,000 rows or the result is unresolved.

In a disposable source with more than 1,000 rows, place one clearly identifiable nonnumeric quantity after the first 1,000. Inspect the numeric-conversion step with the full profile. This deliberate trial checks whether your inspection covers the problem record; it is not a recommendation to alter a real report's input.

Read the count labels literally

Column Distribution reports distinct and unique values. Distinct counts different values; unique counts values that occur only once. For a small column containing A, A and B, there are two different values, but only B occurs once. Neither label by itself confirms that a business identifier satisfies the intended key rule.

For the real import, inspect the records behind unexpected errors, empty fields or repeated identifiers. Compare them with the source before choosing a correction. A value can be valid for its data type and still describe the wrong item or reporting period.

Record the scope, query step and findings with the import check. Remove any disposable test source from the working process, then confirm the real result. A useful acceptance note states what was inspected rather than simply saying that the preview looked clean.

Sources: Microsoft: Power Query data profiling.