Let's say that it's crystal-clear what the user needs to do next and we don't need to test whether or not the user is struggling with finding the action to perform.

Rather, they're seeing the action, and due to the verbiage on the label, they don't want to perform it.

Normally, I'd do an A/B test in production with the proposed better label, and watch the conversion rates. But what if one needed to present data before a change could be made in production (example: you're convincing a different company to make the change on your behalf, not your own development team.)

We could do unmoderated testing, but we're not supposed to ask users to predict their behavior, right? (Example: "Would you proceed? Why or why not?") Is there another way to predict a lift in conversions ahead of making a change?

  • If users don't want to perform an action, then you shouldn't need to convince anyone, that's enough reason. Other than that, one way could be benchmarking the competitors or documenting succesful cases using your proposed solution. Not sure I understand this part we're not supposed to ask users to predict their behavior, this is a valid type of research (conjoint analysis), but it's also used in speakaloud , focus group research and other methodologies
    – Devin
    May 22, 2023 at 18:39
  • @Devin To clarify - I think it's fine to ask, "Which shoes are you likely to click on in this scenario." Trying to avoid larger behavioral predictions like, "If you saw this screen, would you sign up for this program," etc.
    – Izquierdo
    May 22, 2023 at 19:15

1 Answer 1


There are both qualitative and quantitative methods to do this, but there is always a risk to load up questions and metrics only looking from one perspective.

Some starting point would be to look at average time to completion for tasks, or completion rate for specific tasks if you want to focus on the quantitative.

Some starting point for qualitative data would be to ask at the end of tasks or at the end of the activity about the specific ease or difficulty to complete tasks.

Also, I would consider whether A/B testing is the best way to figure out conversion rate, since there are many different things other than the amount of text and the language used that might stop someone from wanting to perform an action. Unless of course your interface is very simple and there are only limited elements or content that the user is exposed to, otherwise your simple A/B test will suddenly become a more complicated multivariate test that is hard to analyze.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.