Bite-Sized UX: The Handiness of Assumptions and Hypotheses

At FreshBooks, our product team talks a lot about assumptions and hypotheses. What’s the difference, and why are they useful?

Assumptions: “We assume…”

An assumption is basically an informed guess… which might be biased, and which is definitively unproven. Assumptions lurk everywhere, and they’re often wrong. What’s really useful, then, is calling out assumptions so that we can turn them into falsifiable hypotheses.

In our design process, assumptions often come up when we’re brainstorming or critiquing a design. “A user would probably click on that button if it was a different colour,” one of us will say, and another will call out, “That’s a testable assumption.” Excellent! We have a test to run, which means we need a hypothesis.

Hypotheses: “We believe…”

A hypothesis is a bit more formal. It’s a way of stating an assumption in a testable form. We also call a hypothesis a “we believe” statement since we start our hypotheses with those words. For example, “We believe that making all our buttons purple will double conversion.” We can test that and learn whether our belief was right. By testing, we can spin our assumption straw into insight gold.

A key part of a sound hypothesis is that it’s falsifiable. For example, “We believe that users will love our purple buttons” isn’t falsifiable; it’s subjective. And (this is going to sound grouchy, but:) why do we care if they love them? Aren’t we trying to achieve a result? Our belief that a purple button will double conversion has a condition and result we can test — if the button’s purple, the conversion rate will double, and that’s quantifiably either true or false.

A hypothesis forms the backbone of our design sprints; the purpose of the sprint is to validate it. That’s why we usually end our brainstorming session at the start of a design sprint by agreeing on a hypothesis. The designer and product owner then test it within the week.

In short, when you identify an assumption, rewriting it as a falsifiable hypothesis will help you get data you can use. And what if your hypothesis fails? That’s great, because you’ll have learned something!

Thanks to Danielle Klein, Kiley Meehan, and Aaron Wright for feedback on this article.