Research
Word clouds for open-ended survey responses, compared by segment
Import a CSV or Excel file with one response per row. WordCloud+ finds the words and phrases people repeat, keeps the columns that say who answered, and lets you look at one group at a time or put two groups side by side. Then export the coded responses back to a spreadsheet.
The file is read in your browser and is never uploaded. A Free account is needed to open the editor.
Open-ended answers get stuck in the spreadsheet
A survey export has a column of free-text answers next to the columns you can chart. The closed questions get tables. The open-ended column gets read once, a few quotes are copied into the report, and the rest stays in the file.
Pasting that column into a word cloud tool loses the other columns, so you cannot ask the obvious next question: do younger respondents say something different from older ones, or one region from another?
WordCloud+ reads the file itself and keeps each answer tied to its row. If you only have pasted text and no segments, the free word cloud generator on the home page takes it without an account.
What you need
- A .csv or .xlsx file with one response per row.
- One column that holds the open-ended answers.
- Optionally, up to 3 columns that describe the respondent, such as age band, region or customer type.
Import a CSV or Excel file of responses
- 1
Choose the file. On the new project form, choose a .csv file or an Excel .xlsx workbook with one response per row, up to 10,000 rows and 20 MB. For a workbook, the first sheet is read.
- 2
Pick the responses column. Say whether the first row holds column names, then pick the column with the answers. The column with the most text is selected for you. Each row becomes one line of the project, so a phrase is never joined across two respondents.
- 3
Pick the segment columns. Tick up to 3 columns that describe who answered, such as age band or region. Their values are kept with each response. A segment column can hold up to 50 different values.
- 4
Create the project. Rows with an empty response are left out and the form tells you how many. A line break inside an answer is closed up so the answer stays on one line. The cloud is then built from the responses, with phrases of up to four words.
Not read today: the older .xls format, sheets other than the first, Google Sheets links, SPSS files and PDFs. Save the data as .xlsx or .csv first. How phrases are found is explained on the phrases page.
Filter by segment, or compare two groups side by side
Show one group.
On the Cloud tab, choose a segment column and one of its values. The cloud is rebuilt from only those responses, exactly as if you had made a project from them alone. The colours, fonts, stop words and codes stay the project's.
Compare two values.
“Compare two values” draws two clouds next to each other, one for each value of the same column. Each has its own Download, so both can go into a report. Each download follows the same rules as the Cloud tab's Download.
See what differs.
Under each cloud is a list headed “More common in A than in B”: the phrases a larger share of that group mentions, with both percentages and the counts behind them.
The filter and the comparison are on the Cloud tab. The Summary, Coding and Text tabs show the whole project. A filtered view is not saved: it is gone when you leave the tab. If you edit the text after importing, the segments stop applying until the text is back to what was imported.
What the comparison is, and what it is not
The comparison is descriptive. It counts, and it shows you the counts. It is not a significance test. The app says so itself, under every comparison:
These figures describe the two groups; they are not a test of statistical significance.
Small groups give large differences. Six mentions out of twenty looks like a big gap next to two out of twenty, and it may not be one. Use the lists to decide what to read, then read the responses.
How a row is worked out
- A response is one row of your file. A response that says a phrase three times counts once.
- The share is mentions divided by responses. For each group it is the number of responses that mention the phrase, divided by the number of responses in the group.
- A phrase needs at least 3 mentions in a group to be listed. Each list shows up to 10 phrases, ordered by the gap between the two groups in percentage points.
- Every row shows its counts. Both percentages come with the counts they were worked out from, so any row can be checked by hand.
Export the phrase table and the coded responses
Coding is how you turn phrases into themes: you create a code, such as “Price”, and apply it to the phrases that belong to it. Coding needs Premium or Research. The two downloads on the Coding tab are part of Research. Both are CSV files that open in Excel. How to code open-ended survey responses.
- Phrase table. One row per ranked phrase, with these columns: phrase, count, rank, score, codes.
- Coded responses. One row per response, in the original order: the line number, the text, a 1 or 0 column for each code, and a column for each segment. A response gets a 1 for a code when it contains a phrase you coded with it.
Your responses stay on your computer
Survey responses are often covered by a promise you made to respondents. The file you import is read, parsed and turned into a project in your browser. It is not uploaded to us, and neither are the responses, the segment values or the column names. The project is stored on your own device.
Our analytics receive counts, such as how many rows an export had, never the text. Text leaves your browser only if you turn on CloudSync, and then it is encrypted before it is saved to your own Google Drive. Privacy policy.
Price
The Research plan includes everything in Premium.
Coding needs Premium or Research. The two CSV exports are part of Research.
Questions about survey responses in WordCloud+
- Is my survey file uploaded anywhere?
- No. The CSV or Excel file is read, parsed and turned into a project in your browser. The file, the responses, the segment values and the column names are not sent to us, and the project is stored on your own device. Our analytics receive counts, never the text. Text leaves your browser only if you turn on CloudSync, and then it is encrypted before it is saved to your own Google Drive.
- Which files can I import?
- A .csv file or an Excel .xlsx workbook with one response per row, up to 10,000 rows of responses and 20 MB. For a workbook, only the first sheet is read. The older .xls format is not read: save the file as .xlsx or .csv first. Comma, semicolon, tab and pipe separated files are recognised.
- How many segments can I use?
- Up to 3 segment columns per project, such as age band, region or customer type. A segment column can hold up to 50 different values. A column with more than that is refused, because a segment is a few groups, not a respondent id.
- Is the comparison a significance test?
- No. It is descriptive. For each phrase it shows the share of each group's responses that mention it, with the counts behind the percentages, and lists the phrases that are more common in one group than the other. It does not test whether a difference is statistically significant, and small groups give large differences.
- What happens if I edit the text after importing?
- Segment values are tied to the exact text they were imported with. If you edit the text, the segments stop applying and the filter and comparison are hidden. They are not deleted, and they apply again if the text goes back to what was imported. Nothing is guessed or shifted.
- What is in the exports?
- Two CSV files from the Coding tab. The phrase table has one row per ranked phrase with these columns: phrase, count, rank, score, codes. The coded responses file has one row per response, with its line number, its text, a 1 or 0 column for each code, and a column for each segment.
- Does it use AI to find themes?
- No. Phrases are found by counting runs of adjacent words, and codes are ones you create and apply yourself. The same file gives the same result every time.
- Does it work in other languages?
- Phrases are found in languages written in the Latin alphabet, with default stop word lists for 36 of them. Text written entirely in scripts such as Cyrillic, Arabic or Chinese does not produce terms yet.
- What does it cost?
- Where the Research plan is on sale, its price is shown on this page and on the billing page, in your currency. Research includes everything in Premium. Coding needs Premium or Research. The two CSV exports are part of Research.
Try it on your own responses.
Take one survey export, import it, and compare two groups. The file does not leave your computer.