Skip to Content

From Word document to fielded survey in ten minutes

Your questionnaire is still a Word document, and that’s fine.
17 November 2026 by
From Word document to fielded survey in ten minutes
ANNE-MARIE MOIR

Most questionnaires still start life in Word. The client comments in the margin, the account director tracks changes, someone writes “PN: ROTATE” in capitals, and the final version is signed off as a .docx.

Then someone has to script it. They retype every question, rebuild every answer list, and translate “IF S3 = 3–5, SKIP TO Q7” into the logic of whatever platform you’re using. Then someone else checks it, because a scripting error found after fieldwork is paid for twice.

That step is the one Redge takes out. You bring the Word document you already have, and Redge turns it into a survey with the routing wired, which you then check before anyone is paid to answer it.

If you’ve been looking for a way to import a Word questionnaire into survey software without survey scripting, this is it. The video below shows the whole process on a real questionnaire, including what the import got wrong and how it was fixed.

Why the questionnaire is still a Word document (and that’s fine)

There’s a school of thought that says the questionnaire should be built straight into the survey tool. For some studies, it should.

But Word is how research teams and clients agree what’s being asked. It’s where the comments, the sign-off and the version history live. Asking a client to review a survey inside a platform they don’t use is asking them to change how they work so that you can change how you work.

So the Word document stays. What goes is the retyping.

1. Write the questionnaire the way you already do

You don’t need a special template. Redge reads a questionnaire written the way researchers normally write one, with one question per paragraph:

  • Sections as Word headings, such as “Section B: Usage”.
  • Numbered questions: S1, Q1, B3, whatever your house style is. If you already use reference IDs, put them at the start of the question line and Redge will pick them up.
  • Answer options listed one per line under each question.
  • Question types in words where they aren’t obvious: “(open end)”, “(single code)”, “select all that apply”, “on a scale of 1 to 5, where 1 is Very poor and 5 is Very good”, “rank these in order of preference”.
  • Routing as a plain-English note directly under the question it belongs to: “If Under 18, screen out. Otherwise continue.” “If No, skip to Q7.” “Split randomly between Concept A and Concept B.” “Everyone else continues to the Wrap-Up section.”
  • A probe brief next to any question you want followed up, written the way you’d brief a moderator: “Ask what specifically drove the score. Don’t accept one word.” That sentence becomes the instruction the interviewer follows. Open ends are the usual place, but a probe works on closed questions too: why they ranked one option first, or what’s behind a rating.

Three things to do before you import:

  • Accept all tracked changes and resolve the comments. The importer reads the text, not the markup, so a document still in review imports the wrong wording.
  • Take your answer options out of tables. Anything inside a Word table is dropped, so a question laid out in a table arrives with no answers. Grids are the usual culprit: write the statements as a plain list.
  • Strip out what isn’t questionnaire. Cover pages, background notes and appendices count towards the import limit of about 2,500 words. A questionnaire on its own rarely gets near it; the surrounding document often does.

2. Import it

From your Surveys list, choose Import document and upload the file. Redge also reads PDF and PowerPoint, but Word gives the cleanest result.

What comes back is a draft to review before it becomes a survey: the sections, every question with its type and answer options, the question numbers you used, and probing switched on wherever you wrote a brief. Fix any wording or types, then create the survey.

That’s the easy half.

3. Check the routing it compiled

Generating logic is easy. Proving it’s right is the job.

Routing is where most survey errors live: a filter pointing at the wrong code, a question nobody can reach, a screen-out that never fires. A tool that writes logic for you and hides it in rules question by question has only moved the checking, not removed it.

Redge compiles your routing notes into a map you can read. Open the Workflow tab and every section, every branch and every screen-out is a line on the graph, with every rule sitting against the question it belongs to. You can see at a glance whether everyone who should reach Q7 does, whether the people you screened out actually leave, and whether both arms of a split come back together where you intended.

Routing is inferred from your wording, so an ambiguous note produces a rule that is valid but not what you meant. If a rule is wrong, rewrite it in plain English (“if they don’t do the grocery shopping, screen them out”) and Redge compiles it again. Or set it by hand. Either way, you can see what you’re about to field.

4. Set what a document can’t carry

Some things can’t be read out of a Word document, and it’s better to know that than to discover it in field. After every import, work down this list:

  • Answer option behaviour: shuffle, pin, exclusive (“None of these”) and “please specify”. Write them in brackets in your document anyway, as “(exclusive)” or “(randomise)”, so they’re sitting in the question as a reminder when you set them.
  • Images and video: stimulus, concept boards and image-choice options.
  • Grids and image choice, which arrive as simpler question types and need converting.
  • Piping and carry-forward, such as showing only the brands chosen earlier. You can set these in the builder, or ask Claude to do it through the Redge connector.
  • Quotas, because they depend on how you’ll buy the sample.
  • Survey settings: branding, introduction and completion messages.

5. Fix what the import got wrong

No importer reads every questionnaire perfectly, and it would be odd to pretend otherwise. In the recording, the import failed to make 'None of these' exclusive at A2. This took a minute fix.

6. Run the pre-field check

Before anything fields, Redge reads the whole study and reports what’s wrong with it:

  • Blockers: things that will cost money or produce data you can’t use, such as a screener that can’t screen anyone out or a quota that can never fill.
  • Warnings: things that will field but may not do what you meant, such as a multi-select with no exclusive “None of these”.
  • An interview length estimate, including the time added by probing. Panels price on length, so this is the number your sample quote depends on.

Nothing in the check stops you fielding; it tells you what you’re fielding. A clean check says the design holds together. It doesn’t say the questions are the right ones; that’s still your job.

7. Walk every route with a dry run

Reading a design isn’t the same as walking it. A question nobody can reach, or a branch that never rejoins, often only shows up from the inside.

So after the check, run the dry run. It takes every route through the questionnaire as a different respondent, answers the open ends and the follow-ups, and reports anything that didn’t happen the way the design says it should. It runs on a test link, so nothing it collects is counted: your bases and quota cells are untouched.

8. Test it yourself, then field it

The dry run walks the routes. You still need to feel the interview. Run the survey in test mode a few times yourself, through each branch. Read the follow-up questions the interviewer asks wherever you’ve switched probing on: are they asking what your brief asked for, and stopping when they should? Then remove the test responses.

To field with a panel, get a quote for your specification and buy the sample inside Redge. Set a soft launch so the panel stops at a small share of the sample, then check the data again before you release the rest: the screen-outs, the split between routes, the interview length, and a sample of the open ends and follow-ups. All of it runs from the study’s Fieldwork panel. Or send the link to your own list.

If the client wants one last look before launch, export the survey back to Word. It comes out as a questionnaire they can read and sign off, built from what will actually field.

What the ten minutes doesn’t include

Ten minutes is the time from document to a survey that’s checked, tested and ready to field, for a typical questionnaire of 20 to 30 questions. A questionnaire with heavy routing, grids or stimulus will take longer, mostly in the checking and the settings a document can’t carry.

It doesn’t include writing the questionnaire, getting it signed off, or fieldwork. Those are still yours. What it removes is the day or two of scripting and checking between “approved” and “live”.

A checklist for a Word document that imports cleanly

  • Accept all tracked changes and resolve all comments.
  • Keep it to the questionnaire: under about 2,500 words, with no cover pages or appendices.
  • Use Word headings for sections.
  • One question per paragraph, numbered, with its answer options listed underneath, never in a table.
  • Say the question type where it isn’t obvious, and give scales their range and end labels.
  • Put each routing note directly under its question, name sections exactly as their headings read, and say where every branch goes afterwards.
  • Write a one-sentence probe brief under each question you want followed up, open or closed.
  • Mark “(exclusive)”, “(randomise)” and “(specify)” in brackets as reminders, then set them after import.

Try it on your own questionnaire

The quickest way to judge this is with a questionnaire you’ve already written. Open a free account, import it, and look at the routing graph. It costs nothing until you field.

If you’d rather see it done with your document on screen, book a 30-minute walkthrough.