← All help articles

Testing a survey

You can try a survey three ways before it goes live, all from the buttons at the top of the builder. Preview lets you click through it yourself with nothing saved. Simulate sends hundreds of made-up respondents through it to show where people go and how long it takes. Test link gives you a real link to take it, or share with colleagues, with the answers saved as tests that never count as real responses.

Preview

Click Preview to open Preview as respondent: the survey exactly as a respondent sees it, run on your draft as it stands. Switch between Phone and Desktop to check both sizes.

Answers are checked with the same rules as the live link, and the routing follows the same logic, so what happens here is what happens live. The This run panel lists the path you have taken and how it ended (completed, screened out, quota full and so on). Start again begins a fresh run, and Show on map jumps to the card you are on. Press Escape or the close button to leave the preview.

Nothing in a preview is saved and nobody is counted. Quotas let everyone in, and each run is a new respondent, so shuffled orders and random splits change from run to run. If the logic has a problem, the preview stops and says so; the Map checks list what to fix.

Simulate

Click Simulate to open the panel, choose how many made-up respondents to send (100, 500, 1,000 or 2,000) and press Run. While the panel is open, their numbers show on the map's lines.

It shows:

  • Where they went: the share who completed, the share who ended for each reason (screened out, quota full and so on), and which cards they ended at. Click a card's name to find it on the map. If anyone got stuck on a logic problem, it says so.
  • Length and feasibility: about how many minutes the survey takes someone who completes (including the panel's own pages), the average number of questions they see, and roughly how many people need to start for your target, with a verdict of Looks feasible or Probably not feasible. Screen-outs the panel handles for you through targeting are left out of that count.
  • How people answer: every answer is equally likely unless you say otherwise. Type the percent of people you expect to give each answer (for a multiple choice, the percent who tick it) to make the screener realistic. These percents aren't saved with the survey.

Quotas close at their share of your target, as they do live. The feasibility line is a simulated estimate; the real reach and price come from the panel on the Audience and Pre-launch tabs. If you change the survey, the panel tells you to run it again: press Run again. Close the panel with its close button, or by clicking Simulate again.

Test link

Click Test link, then Get a test link. Use the buttons beside the link to copy it or open it in a new tab. Take the survey for real, or send the link to colleagues: answers are saved and quotas count, but every response is marked as a test. Nothing is charged and nothing goes to the panel.

  • The link is a snapshot. Changed the survey? Click Update the link for a new link on the new version. The old link keeps working on the old version. If nothing has changed, you get the same link back.
  • Fix map problems first. A test link can't be made while the Map checks show errors.
  • Up to 100 test responses. A survey's test links take 100 test responses between them, then stop. Clear test answers deletes every test answer on the survey (it can't be undone), frees their quota places and starts the count again.
  • Quality checks run on tests too. Test respondents are judged like real ones, so you can see the checks work. The test link panel shows how many passed, were held or were rejected, with the reasons for the latest 20.

Test links are for before launch: the Test link button goes away once the survey is live, and test links stop working when a survey is closed.

Test answers stay apart

Test responses never mix into your real data. On the Results tab, the Responses filter offers Live, Test links or Both. Until a survey has live responses, the page opens on its test responses so you can see the charts working (see survey results).

When you're happy, head to launching a survey.

Still stuck? Sign in and ask Graham, or email support@sortedresearch.com.