Describe what you want to learn and the builder drafts the study for you, title, description, screening questions, and the questions or discussion guide for whichever method you pick.
You review and change anything before it goes anywhere.
Getting a good draft
The draft is only as good as what you give it. Three things help most:
Say what decision this feeds. "We're deciding whether to rebuild onboarding or improve the current one" produces sharper questions than "we want to learn about onboarding."
Describe who you need, specifically. Role, industry, company size, and any experience that matters, like having switched vendors recently.
Say what you already believe. If you suspect people abandon at the payment step, say so. Good questions test a belief rather than wander.
Uploading documents
You can upload a brief, a research plan, previous findings or a product document, and have the study built from it.
Accepted formats: PDF, Word, Excel, CSV, plain text, Markdown, and images (PNG, JPEG, WEBP, GIF).
Limits: 20 files per upload, 10MB per file, 50 documents per study.
You can also pull documents in from Google Drive or Notion if you've connected them. See Google Drive and Notion.
What it takes from a document: more than context. It pulls out the practical details: how many participants you need, how long the study should run, what you're paying, who you're targeting, and your objectives.
It also lifts actual content, screening questions, survey questions, and the branching between them, if your document contains them. A written brief with a draft screener in it often comes back close to ready.
Always check what it extracted rather than assuming. It's reading your document, and documents are ambiguous.
Attaching an NDA
You can attach an NDA to a study, which participants see before taking part, common for professional research where you're showing unreleased work. Upload it with your study documents.
Reviewing the draft
Always read the screening questions properly. They decide who gets into your study, and a loose screener is the most common reason a study produces disappointing data.
Check the questions test what you actually want to know.
Cut anything you wouldn't act on.
For AI interviews, preview the interviewer before launching. Hearing the voice and pace tells you more than reading the guide.
Changing things later
Everything is editable before launch. After launch you can still adjust some things, though changing questions partway through means you're comparing answers to different questions.
If you need a substantial change once a study is running, contact us and we'll talk through the cleanest way to handle it.
If the draft misses
Tell it what's wrong in plain language, the same way you'd tell a colleague. Rewriting the request usually works better than editing a draft you don't like.