Import a survey export

Golden recognizes a survey workbook on upload. A respondent-level export is reshaped into one row per respondent per question with codes and wording joined; an already-summarized results report is imported as-is and turned into a board in Discover instead.

4 min readUpdated October 16, 2026

Survey exports arrive in an awkward shape. One sheet holds the answer labels, another holds the numeric codes, and a third is a data dictionary with the full question wording. Golden looks for that pattern when you upload a workbook and offers to stitch it together for you.

The detection prompt

When Golden spots survey data, a dialog titled "Survey data detected" opens before the import runs. It names your file and offers two choices.

Automatically reshape and apply the data dictionary/codes - the default. Golden imports the labels sheet, reshapes its questions into rows, then joins each answer's code and each question's wording. The wording of this option shortens when your workbook only ships one of the two, so you may see "Automatically reshape and apply the codes" or "Automatically reshape and apply the data dictionary" instead.
Let me choose the sheets - pick the sheets yourself. Nothing is reshaped, and no codes or dictionary are applied.

A preview under the heading "You'll get this" shows the shape you are about to import, so you can compare the two options before committing. Click Import to run it.

What the import creates

The automatic path produces one row per respondent per question, plus a set of new columns describing the question and answer.

Question ID - the identifier of the question the row answers.
Labels - the answer as it was worded in the survey.
Value - the numeric code for that answer.
Wording - the full question text from the data dictionary.
Question Grouping - the battery or block the question belongs to.
Qtype - the question type.

Note: If a column of that name already exists in your data, the new one is suffixed with " (survey)" rather than overwriting yours.

An already-summarized report is different

Not every survey file holds individual responses. A published results report - a vendor's results page copied into a spreadsheet, with one row per question and answer option and each option's share in a column - has nobody's individual answers to stitch together, so the reshape prompt doesn't apply. It imports as ordinary sheets, and Discover offers a ready-made board from it instead of a reshape. See Generate a survey dashboard for that board.

Every step is visible and undoable

The import is not a black box. It lands as separate steps in the data prep history in Prepare, named for what they did - for example "Reshaped 24 questions into rows" and "Added codes from Numbers and the Dictionary dictionary". Each row shows "Applied" or a count of rows affected, and hovering a step reveals a "Remove this step" action.

Tip: Because the steps are recorded rather than baked in, you can remove just the code-joining step and keep the reshape, or roll the whole thing back and re-import with "Let me choose the sheets".

Platform exports

Golden reads the header conventions of common survey platforms, including Qualtrics exports with their ImportId header row. There is nothing to configure - the reader is chosen from the file itself, and detection either fires or it does not.

Heads up: Detection needs real question and answer text to work with. An export whose questions and answers are placeholders like "Question 1" and "Answer 2" gives Golden nothing to recognize, so the prompt will not appear and the workbook imports as ordinary sheets.