Writing useful hypotheses in the product design process helps understand user and business problems better while fostering critical thinking about the product team’s beliefs and bets.
☕ Writing useful hypotheses
Hypothesis! Gotta love them. They provide scaffolding to how we think about our process and what we want to achieve. Also, since the hypothesis is a term scientist use, stating one must also mean our team is data-driven, doing science.
Well as it often goes, hearing about or using a term doesn’t mean it’s understood well. Or used the right way. Just writing a statement and calling it a hypothesis, doesn’t make it a hypothesis. Much less science.
I’ve written earlier how stating a design intent helps in articulating shared goals not only for the designer but the whole product team. While hypotheses are great in stating the intent, their role goes beyond, as they give a broader clarity to the product design process.
In the context of product development, a hypothesis refers to a proposed explanation that can be tested to validate or invalidate a specific product-related idea. While its use is based on the scientific method, it’s more of an inspiration than an actual implementation of the scientific process.
Because of this, the more years pass, the less I like the terms hypothesis (and also assumptions). While they are scientific terms and carry some similarities, the similarity hides the differences in rigor, which leads to misunderstandings. So recently I started to use terms like bets and beliefs which are closer to how people talk and think about these things.
Still, hypotheses are incredibly useful to clarify the thinking of the product team, if defined well. A couple of years ago one of my ex-colleagues, Jonas came up with a pretty neat framework that I’ve been using since:
- We observe …
- We believe that (assumption) …, because (evidence) …
- If we (do this) …, we’ll observe (result) …
We observe…: First, we start with a business problem. This is something we observe in our systems or the world that we think is worth solving.
- Business problems should be understood. If it’s data, there needs to be some exploration of what it means, user research to figure out how it affects our user’s behavior, and market research to understand what other companies and the broader industry are doing to address the problem.
- This allows us to transform the business problem into something that’s connected to what the users are doing, which puts it into the space where product development can do something about it.
- The result should be a list of user problems we identified, connected to the business problems, and compiled into a single observation statement.
We believe that (assumption) …, because (evidence) …: Second, we write a list of hypotheses and beliefs based on what we know at this point. Our knowledge is still not certain. We want to solve the problem but still need to figure out the right way to solve it and also solve it the right way.
- Beliefs should be about what we think of the user problems and a list of supporting evidence. This is the most important part, as the solutions should be ultimately about solving user problems that drive the right business outcomes.
- Beliefs and evidence might be lacking, which points to more research that might be necessary.
- Based on this list we can prioritize which of the user problems we found can we also solve, and seems we can come up with effective ways of solving them.
- Based on this list we can clarify the design intent, what is the problem, and the constraints of the problem we will work with.
If we (do this) …, we’ll observe (result) …: Third, based on the beliefs we can create a list of bets for each belief, possible ways that reflect our beliefs, and ultimately solves the initial problems.
- Each bet is a possible solution to the users’ problem that we also think would drive the initial business problem.
- Since we don’t know yet which bet has the best potential (in the most rough sense for example impact / effort), we need to learn about each of the bets.
- Sometimes a list of small impact / small effort bets is better than a large one, but usually, it should be clear which bet(s) are the most lucrative ones.
- To learn about a bet, further discovery is needed - this mostly means explorations, experiments, and anything that lets us learn about the bet and the belief, and helps us prove, we made the right prediction with the bet.
This structure is also similar to the formulation of Teresa Torres’s opportunity solution trees and gives a high-level framework for product discovery and a hypothesis-driven design process.
One way I found the above structure helpful, is how earlier I often struggled on hypothesis writing sessions. It turns out, generating is hard if there is not enough data about the problem space. “Just writing something” in these cases won’t lead to useful statements, only to muddy design processes with too much back and forth on small UI details.
Another way this formulation helps is that it encourages more active looking to disprove the beliefs. Often product teams are focused on progress, validating their ideas fast and starting building, rather than looking for signs they might be wrong. By actively looking for evidence and not just justification for beliefs, solution ideas will be less likely to become wishful thinking.
The terms beliefs and bets also allow for a simpler way of teaching hypothesis-driven design to the design team. Beliefs highlight how the team might not have enough data, while bets show that outcomes might be uncertain. This fosters a culture safer for taking risks, experimenting, making data-informed decisions, and most importantly critical thinking about what the team knows.
🥤 To recap
- Hypotheses provide scaffolding and clarity to the product design process, helping teams articulate shared goals and think in a data-driven manner.
- A hypothesis in the context of product development refers to an assumption that can be tested to validate or invalidate a specific product-related idea.
- A framework for writing a hypothesis is “We observe…, We believe that (assumption)…, because (evidence)…, If we (do this)…, we’ll observe (result)…”.
- This framework emphasizes the importance of understanding business problems, connecting them to user problems, and prioritizing them based on beliefs and supporting evidence.
- Using hypotheses encourages a mindset of actively looking to disprove beliefs, fostering critical thinking, experimentation, and a culture of taking risks and making data-informed decisions.
This is a post from my newsletter, 9am26, subscribe here:
🍪 Things to snack on
Already at its third edition, Jeff Gothelf and Josh Seiden’s Lean UX book is a good way to get started with hypothesis-driven design and product development and is generally a recommended read to product designers. This edition describes the Lean UX process based on the Lean UX Canvas.
I loved Tal Raviv’s That’s Not a Hypothesis, as it gives a pretty good critique of some of the cargo-cult usage of hypothesis and describes a good way of thinking about them. The most important point is to separate the problem (the objective), the hypothesis (the thing we believe to be true), and the prediction (the thing that will happen).
Thinking your design with hypotheses is a nice overview of how to formulate hypotheses starting from the scientific process, also showing a good framework. There is also a good discussion on the definition of success, which is what result the team hopes from the proposed solution.
Andrew Duckworth writes about how hypotheses can act as the articulation of intent in Hypothesis driven design. The simple presumption here is, that most designers are already thinking as if they would have a hypothesis, but as it’s not explicit, it’s not becoming something to learn from. The article also describes a simple way of working with hypotheses.
One challenge with hypotheses is, that while it’s easy to generate more, validating them takes time and resources. This is where Jeff Gothelf’s The Hypothesis Prioritization Canvas helps, as it gives a simple way to decide what to do with each hypothesis. The canvas is a simple matrix around risk and perceived value, and the resulting quadrants are Test, Ship & Measure, and Don’t Test. Usually Don’t Build, and Discard.
One way to group research activities is how they relate to hypotheses, as Laura Klein writes in Hypothesis Generation vs. Validation. Generation happens after collecting data about users and the team has an idea of what needs to be done. Validation follows once there is a tangible solution to the problems found that need to be tested in some way.