Why does a survey summary count one answer twice when the wording looks the same?

The displayed wording can look alike while the stored characters differ. One response might have a trailing space, another a different punctuation mark, or one might use different capitalization. A counting tool that treats those text values as distinct can show separate categories. For example, SQLite’s default BINARY text comparison compares string data directly; that describes SQLite, not every survey or spreadsheet program. [S1]

Start by checking the raw responses, not just the chart labels. Copy both values into a plain-text view if available, and inspect their beginnings and endings. Don’t assume that a visible match means the underlying text is identical, or that every odd-looking character should be removed.

Keep the imported response column as your original-value record. Make a separate working column or copy for comparison, and write down each change you test. For instance, you might first compare values after removing ordinary leading and trailing spaces, while leaving punctuation and capitalization alone. SQLite documents trim(X) as removing ordinary spaces from both ends, but that does not establish how it handles every kind of whitespace or other character. [S2]

Compare the original and working values side by side. If they differ, inspect the pair and decide whether the change is appropriate for this survey. Treat a normalized match as a review candidate, not automatic proof that two answers mean the same thing. A response like “yes!” and “yes” may differ in emphasis or intent even if a punctuation-removal rule would combine them.

Suppose the raw answers include North, North , and north. These are illustrative examples, not findings from a particular survey. Create a review list such as:

• Original North: make no change; use it as a reference value.
• Original North : test removing an ordinary trailing space, then compare it with North.
• Original north: compare without changing case, and decide whether capitalization is meaningful for your question.

Keep the proposed change and decision visible for each value. This list is a review record, not an instruction to merge every row. If your tool offers a trim function, check its documentation and test it on a copy; for example, SQLite’s trim(X) is specifically documented for ordinary spaces at both ends. [S2] Neither that detail nor this example establishes how your survey software handles hidden whitespace, punctuation, or Unicode characters. If values still appear identical but count separately, check the tool’s documentation or export the raw values for a closer comparison.

Record the proposed rule, the values it groups, and any exceptions. Review the affected response pairs before using combined counts, especially where punctuation, capitalization, or unusual characters could change meaning. Keep the original column and report counts from the reviewed comparison values; don’t replace the raw data with the cleaned version. If you cannot tell whether two answers mean the same thing, keep them separate or label the decision for follow-up rather than silently merging them.