A/B test for questions

Test the wording. Keep the one that gets more people to the end.

Show respondents different versions of a question and see after which one more of them finish the survey. SurveyNinja marks the leader once it is at least 80% confident.

The A/B test works on every plan. Comparing completion needs Standard or higher.

The A/B test in the report: Formal finished 17 of 38, Friendly 28 of 39 with the Leader tag and about 98% confidence
A Formal
The survey on a phone with the question "How would you rate your visit today?"
B Friendly
The survey on a phone with the question "How was your coffee today?"
  • Runs on every planThe Leader tag needs Standard or higher.
  • One person, one versionBack in the same browser, they see the same variant.
  • A leader with its oddsMarked from 80% confidence, with the number shown.
  • The variant in your dataExport on every plan, filters from Basic.

One word in a question can change how many people finish your survey. Show two wordings and see which one gets more people to the end.

How to A/B test a survey question.

Question settings with the A/B test switch on and the Manage variants button, next to a question with the Formal and Friendly tabs
The A/B test variants window: Formal and Friendly at 50%, the variant name, its share and its title
The A/B test block in the report: Formal finished 17 of 38, Friendly 28 of 39 with the Leader tag and about 98% confidence
  1. 1

    Turn on A/B test.

    Open the question settings and switch on A/B test. Two variants appear, 50% each, and the question shows their tabs.

    Question settings with the A/B test switch on and the Manage variants button, next to a question with the Formal and Friendly tabs
  2. 2

    Write the variants.

    Give each variant its own title and description, and set the share of people who see it. Add as many as you need.

    The A/B test variants window: Formal and Friendly at 50%, the variant name, its share and its title
  3. 3

    Read the leader.

    The report shows how many people saw each variant and how many of them finished. The better one gets the Leader tag.

    The A/B test block in the report: Formal finished 17 of 38, Friendly 28 of 39 with the Leader tag and about 98% confidence

What to test.

  • Formal or friendly"How would you rate your visit?" or "How was your coffee?" The tone can change how many people finish.
  • Short or detailedOne line, or a line with an example in the description.
  • Broad or specific"Why did you choose us?" or "What made you order today?"
  • The intro messageA Message block can have variants too: test how you open the survey.
The ninja sorts response cards marked A and B into two boxes

The variant goes everywhere.

Each respondent's variant is saved with their response, so you can use it like any other answer.

  • Logic by variantSend people who saw one variant to their own follow-up question.
  • Answers per variantFilter the report and the responses by variant to see what each group answered. From Basic.
  • The wording each person sawEvery response shows the title its respondent saw.
  • A column in the exportCSV and Excel files get the name of the variant.
  • Close to your sharesWith Autosave progress on in the survey settings, a lagging variant is shown more often, so the split stays near the shares you set.
  • Inside groups tooQuestions inside a group can have their own test, and so can Message and HTML blocks.

Agent skills

Do this with your AI assistant

Free agent skills for Claude Code, Codex, Cursor and Gemini CLI that do this job in your SurveyNinja account.

All agent skills
The ninja holds a five-star rating card

Already paying for another survey tool?

Send us a screenshot of your bill and get a month of Premium free. Our team will help you move your surveys too.

Get a free month

Questions and answers

What is A/B testing in a survey?

Showing different respondents different versions of the same question to see which works better. In SurveyNinja a variant changes the question's title and description, and the report compares how many people finished the survey after each one.

Which plans include the A/B test?

The test itself works on every plan, Free included, and so do logic by variant and the variant in the export. Filters need Basic or higher. Comparing completion needs the All tab of the report, which is on Standard and higher.

What can a variant change?

The question's title and description. Answer options, images and the other settings are the same for every variant.

Can several questions have tests?

Yes, each with its own variants. Every question with a test picks its variant separately.

How is the leader picked?

SurveyNinja takes the two variants with the highest share of finished respondents and compares them with a two-proportion z-test. When it is at least 80% confident the difference is real, the better one gets the Leader tag. It compares completion only, not the answers. To check your numbers at the significance level you choose, use the free A/B test significance calculator in our tools.

Will one person see different versions?

No. A respondent who comes back in the same browser gets the same variant, even if you change the shares.

How many variants can a question have?

As many as you need. Each new variant gets an equal share, and you can change the shares so they add up to 100%.

Didn't find the answer? Write to team@surveyninja.io or use the contact page.

The ninja rides a rocket

Keep the wording more people finish with.

Run the test on any plan. The completion comparison and the Leader tag come with Standard and higher.