1.04 (V) 1.3 Hypothesis and Preliminary Experiments
So how do you know what ideas to test and what kind of MVPs to build? So the way you do that is through hypothesis testing. Just like if you were doing science, you're going to do science to find out whether your startup is a good or bad idea and a pivot if you need to do so. So let's talk about this. Hypothesis tell you what to test by your idea. So in order to do that you need to go through two steps. The first way to do it, you need to figure out what your assumptions of your business are. Finally, your assumptions in the surface sources of uncertainty in your business. You learned a lot of ways to do this in the first lectures of this class where we talked about things like discovery-driven planning and business planning. All of those processes gave you methods of thinking about uncertainty and ways of figuring out what source uncertainty exist in your business idea. What do you do once you have these sort of uncertainty? You then develop those and turn those into assumptions. So you need to go through each of the key assumptions in your business and using either discovery-driven planning or other methods look for the three to five key assumptions in your business by rating them. Once you found those three to five key assumptions, you'll develop those into hypotheses. So what's it mean to develop hypothesis? Hypothesis, just like hypothesis in science, has a number of characteristics. It has to test the key assumptions of business, it needs to be falsifiable and it needs to testable. Let's go through each of those and explain what those mean. So the first thing your hypothesis needs to do is test at least one key assumption. The key assumptions are the things that need to be true about the world for your business idea to succeed. So you may think that customers will be interested in your taste, your cookie, your fast food, Ethiopian restaurant chain, your brand new app. But what do you actually know about what they're interested in? What are you assuming about what they're interested in? You'll have surfaces key assumptions through discovery-driven planning, through your business planning process, and now you know there are three or four key issues that we need to test. Maybe, we need to find out whether are operationally we can run the restaurant the way we think we are, whether our customers actually value the key issue that we're making as the part of our app that we're developing. But whatever those key assumptions are, those are the core of your hypothesis. Once you've figured out what key assumption you're testing, the next step is that you need to create a testable hypothesis. What does it mean to be testable? Testable means that you actually get an answer in a reasonable period of time whether your hypothesis is right or wrong. So if your hypothesis requires you to build your full product and launch into the world, that's not testable. If your hypothesis says 80 percent of people will love our product but you don't know how you're going to figuring out whether 80 percent of people love it or not, that's not testable. So testable hypothesis needs to be one where you actually can run an experiment and find out an answer. Once it's testable, it also needs to be falsifiable. I happen to really love this particular image. It's very useful for true-false tests. I highly recommend teaching your children or yourself to write this. But you don't want this ambiguity between true and false in your own hypothesis testing. It needs to be clearly a falsifiable answer, where you know whether or not you're right or wrong. So it can't be something ambiguous like people will like our product, it needs to be direct number or direct comparison. So let's put this together. You need to make sure your hypotheses are clear, that they don't let you fool yourself, and you want to make sure you're not testing hypotheses you already know the answer in advance. You need to have high levels of uncertainty when we test the key assumption. So you want to make sure your hypotheses are not ambiguous like people will like a product, but that they are clearly indicated. Forty percent of our customers who tried to prototype will agree to pre-order the product for delivery in the next three months. That's clear, it tests a clear issue, it is falsifiable if 40 percent of people who tried the product don't agree then our hypothesis is wrong. It is also clearly testable, we're going to give people a prototype, an MVP and then they going to use that to do the tests of their product. You might ask where we get that 40 percent number from, the idea that it's 40 percent as opposed to 50 percent or 30 percent. That 40 percent needs to be a key number for our products to succeed and a way of getting that is through discovery-driven planning approach that was discussed in a prior lecture. In conclusion, you need to first identify your assumptions using whatever method you want to use. You can do a qualitatively through your planning process or quantitatively through discovery-driven planning process discussed in a prior lecture. You use it to generate three to five key hypotheses that are both falsifiable and testable, that you can then build MVPs for to test and to find out whether you need to pivot or keep going in the same direction.
