Research and analysis

Hypothesis generator for research and surveys

Describe a problem or an assumption, and the hypothesis generator writes research hypotheses you can test. Each one comes with a null hypothesis, a way to test it, a survey question, a metric and a rough sample size.

Or take 8 ready questions

  • Free
  • No sign-up to try
Try an example:

Do not paste names, emails or other personal data: the text is sent to an AI model.

Survey questions that test a hypothesis

Each question measures the metric behind a common kind of hypothesis. Replace the brackets with the terms of yours.

  1. How satisfied are you with [the change or feature]?

    Scale

    1 Very dissatisfied … 5 Very satisfied

    The outcome for an "X lowers satisfaction" hypothesis. Compare the average between the group the change affected and everyone else.

  2. Have you [done the action] in the last 30 days?

    Yes / No

    Gives a share you can compare between two groups with a test for two proportions.

  3. What is the main reason you stopped [using the product]?

    One answer
    • Price
    • Missing features
    • Hard to use
    • Switched to another product
    • No longer needed
    • Other

    Tests a "the reason is X" hypothesis. X must be one option among several, with "Other" as an open answer, or the list leads the answer.

  4. How important is [the factor] when you choose [category]?

    Scale

    1 Not important at all … 5 Very important

    Checks whether the factor in your hypothesis matters to people at all.

  5. What is the most you would pay per month for [product]?

    Open answer

    A number you can compare between a group shown the version with [feature] and a group shown the version without it.

  6. Which of these do you use for [task]?

    Several answers
    • [Competitor A]
    • [Competitor B]
    • Spreadsheets
    • Other
    • Nothing

    Defines the groups for the test, for example users of a competitor and people who use nothing.

  7. How often do you [do the task]?

    One answer
    • Every day
    • A few times a week
    • A few times a month
    • Less often
    • Never

    A control question. Heavy users answer differently, so compare like with like.

  8. How did you first hear about us?

    One answer
    • Search engine
    • Social media
    • A friend or colleague
    • An ad
    • An article or review
    • Other

    Splits the answers by channel to test hypotheses about where a problem starts.

Take these questions as a survey: it opens in the SurveyNinja editor, where you replace the [brackets], change the wording and share the link.

Free account, no credit card, 14 days of Premium features. After sign-up the survey opens in the editor.

From a hypothesis to an answer

  • One cause, one outcome

    A testable hypothesis names what differs, for whom and what you measure: "Customers who used live chat rate support higher than customers who used email." A vague hypothesis cannot fail, so it cannot be tested.

  • Write the null hypothesis

    The null hypothesis says there is no difference or no effect. The data either lets you reject it or it does not. Fixing it before you collect answers keeps you from reading the results the way you hoped.

  • Choose the metric before the survey

    A share, an average on a scale or a count. The metric decides the question type: yes/no for a share, a 1–5 or 0–10 scale for an average, a number field for a count.

  • Count the sample per group

    The sample size in the result is a rough estimate. Check it with a sample size calculator for the margin of error you need, and count it for each group you compare, not for the whole survey.

Questions and answers

What does the hypothesis generator do?

You describe a problem, an observation or an assumption. An AI model writes several research hypotheses and, for each one, a null hypothesis, a way to test it, a survey question, the metric to look at and a rough sample size.

Is it a null hypothesis generator too?

Yes. Every research hypothesis comes with its null hypothesis: the statement that there is no effect or no difference, such as "The price change did not affect renewals". A statistical test tells you whether your data is strong enough to reject it.

Can I use it for a thesis or academic research?

As a starting point. The hypotheses are drafts: check them against the literature and your research question, and agree on the final wording with your supervisor.

How do I test the hypotheses with a survey?

Take the survey question suggested for each hypothesis, add the questions that split people into the groups you compare, and build the survey in SurveyNinja for free. Then compare the metric between the groups.

Is the hypothesis generator free?

Yes. You can generate hypotheses without an account; the number of generations per hour is limited.

Is it safe to describe my problem?

The text you type is sent to an AI model to write the hypotheses. Describe the problem in general terms and do not paste names, emails, customer data or other personal data.

The SurveyNinja ninja takes off on a rocket

Need the whole survey, not only the questions?

Describe it in one sentence, and Ninja AI builds a complete survey: questions, answer options and the start and end screens. You check it and publish.