Hypothesis-Driven Design: The Scientific Method for Product Design
Product design decisions are always risky. You never want to spend a long time building something only to find out that it's not what your users want or need. That's where hypothesis-driven design comes in.
Hypothesis-driven design is an empirical approach to making product design decisions based on systematic observation, experiment, and measurement of formulated hypotheses. This method reduces the risks associated with simply relying on assumptions. By conducting structured experiments, we can learn what works quickly and what doesn't, which helps to shorten the learning cycle and build something with the confidence that it is the right product for the job.
There are several reasons why you should use hypothesis-driven design as your research method.
It reduces risk
This method reduces the risks associated with simply relying on assumptions. By conducting structured experiments, we can learn what works quickly and what doesn't, which helps to shorten the learning cycle and build something with the confidence that it is the right product for the job.
It’s great for stakeholder alignment
HDD is also an excellent tool for stakeholder alignment. The process encourages collaboration and consensus-building because it requires input, review, and agreement from various stakeholders. As everyone collaborates to formulate hypotheses, design and run experiments and interpret results, a shared understanding and vision of the project's goals and approach develop. This common ground can help prevent miscommunications and conflicts, making the design process smoother and more efficient.
If set up right, It can centralise your research: HDD's role in creating a centralised repository of hypotheses, their corresponding experiments, and results cannot be overstated. This repository, serving as a single source of truth, ensures transparency and traceability. It keeps everyone informed, aids in tracking progress, measuring success, and provides valuable insights for future projects. Check out the hypothesis tracker further down the page.
It follows the scientific method The scientific method is a systematic approach to gathering knowledge about the world. It involves making observations, forming hypotheses, conducting experiments, and drawing conclusions. Here's how this applies to Hypothesis-driven design (HDD):
Formulate hypothesis: more on this below
Design an experiment: This involves deciding how to test your hypothesis. You'll need to manipulate the variable(s) you're interested in and decide how you'll measure the outcome
Run the experiment: This is where you implement your design and collect data. You'll need to ensure that your experiment is conducted fairly and that your data is reliable.
Conclude: This is where you analyse and synthesise the results of your testing
Where do I start?
An assumptions is a Hypothesis + an expected outcome, so naturally, starting with assumption mapping session will be the best way to approach it.
Start with an Assumption mapping session with key stakeholders
Decide if assumptions need to be tested through hypotheses or if they can be validated through other means. You can do this with this simple matrix:
Write a good hypothesis for the riskiest assumptions (learn what makes a good hypothesis below)
Create an experiment plan - will you test with generative or evaluative qual, or can you test this in product
Store it all in the hypothesis tracker template below
What makes a good hypothesis?
The hypotheses states a causal relationship with a clear actor:
This means your hypothesis should clearly state a cause-and-effect scenario. The "actor" or subject should be explicit, and the action or cause is something that directly affects the actor. For example, if you're designing a new website, your hypothesis might be, "If we increase the size of the product images on our e-commerce site, then our customers will find the site more engaging and spend more time browsing.”
The hypotheses has directionality
This means your hypothesis should predict the direction of an effect. It should not just state that a change will have an effect but also predict what kind of effect. Going back to our website example, the directionality is that customers will spend more time on the site, not less.
The hypotheses states variables that can be manipulated and measured
Your hypothesis should be based on variables that you can control (manipulate) and observe (measure). In our website example, the size of the product images is a variable you can manipulate, and the time customers spend on the site is a variable you can measure.
The hypotheses is testable with an experiment:
This means should be able to design and run an experiment that can test your hypothesis. This involves manipulating your chosen variable and observing the results. In the website example, you would change the size of the product images and then track how long customers spend on the site. This could involve using website analytics or conducting user testing sessions.
The hypotheses is a falsifiable statement:
This is a fundamental principle of the scientific method. Your hypothesis should be stated in such a way that it could be proven false by the results of your experiment. This doesn't mean that your hypothesis has to be false, but it should be possible to imagine results that would refute it. This is important because it allows you to be confident that your results are meaningful. If your hypothesis was not falsifiable, then no matter what results you got, you could argue that they still supported your hypothesis. In the website example, if customers spent less time on the site after you increased the size of the product images, that would falsify your hypothesis.
It’s used by some of the best
Netflix: Netflix used hypothesis-driven design to test different pricing models before launching their streaming service. They hypothesised that users would be more likely to subscribe if they offered a monthly subscription plan, and this hypothesis proved to be correct.
Uber: Uber used hypothesis-driven design to test different ways of matching drivers with riders. They hypothesised that users would be more likely to use the service if they could see how long it would take for a driver to arrive, and this hypothesis proved to be correct.
Airbnb: Airbnb used hypothesis-driven design to test different ways of pricing their listings. They hypothesised that users would be more likely to book a listing if they could see how many other people were interested in it, and this hypothesis proved to be correct.
Conclusion
Hypothesis-driven design is a powerful tool that can help you create the right product for your users. By following the scientific method, you can test your assumptions and learn what works and what doesn't. This will help you to shorten the learning cycle, build a product that meets the needs of your users, and achieve your business goals.