However, each building block of your business model and value proposition represents an area of risk that needs to be tested: Customer segments, key resources, channels, value propositions, revenues and costs. In order to run good experiments to test our business ideas we need to formulate strong hypotheses that are testable, precise and discrete.

Before you start testing you idea you have to identify the risks within each building block of your business model and value proposition by formulating hypotheses. We define hypotheses as: 

• An assumption that your value proposition, business model, or strategy builds on.
• All the things that would have to be true for you business idea to work.

At the core of identifying hypotheses, you have to ask yourself the question: What would have to be true for our idea to work? You capture each hypothesis by writing phrases that begin with “we believe that...” 

We believe that college students prefer mobile banking compared to the traditional banking experience.

Once you’ve identified your hypotheses, you then need to rank them in terms of risk and importance. By using this simple process you can identify which hypothesis you may want to test first. After selecting a hypothesis to test, you then choose the right experiment to run.

But wait! Before you start designing your experiments, it’s important to put on your scientific glasses and refine your hypotheses, to make sure that they’re well formulated.

This helps you ensure that you have a clear connection between the hypothesis you are testing and the results you will get from your experiments. It is difficult to design good experiments if you have badly formulated hypotheses.

Well-formed business hypotheses have three key characteristics; they are testable, precise, and discrete. Let’s look at each one of these in turn. 

Testable

Your hypothesis is testable when it can be shown true (validated) or false (invalidated), based on evidence (and guided by experience).

🙈
We believe that Generation Z prefers pop-up stores over branches.

 😀
We believe that young adults between 18-24 will spend more time in temporary pop-up stores that are placed in co-working spaces, compared to traditional banking branches.

It would be difficult to design and execute an experiment for the first hypothesis. To build a good experiment, we need to be more clear about what several of the terms mean, which then makes it easy for us to be clear about exactly what we are expecting to happen. Clarity on exactly what we expect to happen is what make a hypothesis testable.

• Generation Z → Adults between 18-24
• Prefers → spend more time… compared to
• pop-up store → temporary pop-up stores that are placed in co-working spaces 
• Branches → traditional banking branches
  

Precise 

Your hypothesis is precise when you know what success looks like. Ideally, it describes the precise what, who, and when of your assumptions.

🙈
We believe that young adults don’t plan for their future.

 😀

We believe that the majority of young adults between 18-24 don’t save more than $100 per month for their retirement.

 If you design an experiment for the first hypothesis, different members of your team may end up testing different types of ‘planning for the future’. This is why we have to be more precise about exactly what type of planning we want test. 

The second hypothesis ensures that the team are all testing the same thing: 

• Young adults → majority of young adults between 18-24
• Plan → don’t save more than $100 per month
• Future → retirement
  

Discrete 

Your hypothesis is discrete when it describes only one distinct, testable, and precise thing you want to investigate.

🙈
We believe that our digital platform helps us increase conversion rates and save money in call centres.

 😀
1. We believe that our digital platform will help us increase conversion rates by 25%.

2. We believe that our digital platform will help us save $200M in call centres over 3 years.

With the first hypothesis, we’re mixing together two different hypotheses, that one experiment probably couldn’t really test. Instead, we should split these into two hypotheses. 

The first hypothesis is about testing your channels, which can be made more precise and testable by saying how much you want to increase the conversion rates. 

The second hypothesis is about the viability of your idea: Can it save money?

Technically, we could use one experiment to test both hypotheses. But by separating the hypothesis into two discrete hypotheses we at least now know that these are two very different things we want to test. 

While it may seem bothersome to write hypotheses this way, it is important to do it because it will lead you to design stronger experiments.