Count survey words by mentions or response lines

Decide what one count means

A word cloud can summarize literal words in a small set of answers, but its meaning depends on the counting rule. A count of ten might mean ten occurrences, ten response lines, or a manually assigned display weight. State the unit beside the artwork so readers do not have to infer it from font sizes.

The frequency counter offers Every occurrence and Once per non-empty line. Use the first to inspect repeated wording. Use the second when each line is one response and repeated emphasis within a response should count only once. Neither setting identifies people, deduplicates submissions or determines whether a comment is positive. Open the local word frequency counter.

Reproduce a three-line example

Friendly friendly helpful
Friendly clear
Not clear

These are invented answers for demonstrating the method. Load Try survey example, which turns off stop-word removal, and count in Words mode with Merge letter case enabled. Keep both minimums at one. The complete comparison below can be verified by reading the three lines, without relying on the visual size of a word.

Prepare one response per line

Remove the question heading, spreadsheet column labels, blank boilerplate and personal information before pasting. Replace line breaks inside a single response with spaces. Keep a separate source copy so you can check context later. If an answer was wrapped into two lines, the counter would treat those as two input boundaries.

1. Paste one complete response per non-empty line; do not paste respondent identifiers.
2. Choose Words, merge case if appropriate, and review the stop-word settings.
3. Count once with Every occurrence, then once with Once per non-empty line. Compare the changed weights.
4. Download the report to preserve both count columns and the method. Import the selected weighted CSV into the generator.

Label the denominator

In this example, friendly accounts for 3 of 7 counted occurrences, about 42.9 percent. With line weights it occurs on 2 of 3 non-empty lines, about 66.7 percent. These values answer different questions. The line percentages for all words can total more than 100 percent because each answer can contain more than one word.

The tool includes every non-empty line in the line denominator, even if all its words are filtered out. Blank lines are excluded. This is not a survey response-rate calculation: missing answers, participant eligibility and repeated submissions are outside the tool. Publish only denominators that fit your actual data collection and preparation.

Keep conclusions connected to the original answers

Not clear contains clear, so a high count for clear cannot establish satisfaction. Removing not can make the visual especially misleading. Read source comments and keep complete expressions as reviewed phrases when needed. A frequent term is a useful starting point for discussion, not a validated theme or evidence that a sample represents a population. Review stop words and negation. Keep complete phrases together.

When you share the cloud, add the counting unit, filters, number of non-empty response lines and a table of values. Identify any manually edited weights. Review unplaced words before publishing; a word absent from the artwork may still exist in the imported list. The existing survey workflow covers the remaining layout steps. Continue with the survey word cloud workflow. Add a usable text alternative.