What is a Hypothesis and How to Write It
What is a hypothesis?
A hypothesis is a statement of the format: X leads to Y.
A simple statement : X leads to Y.
This statement has two parts - a Cause (X) and an Effect (Y). These are the basic building blocks of a hypothesis. Let us see how to write a hypothesis using these blocks.
The Cause (X) of the hypothesis usually is some sort of action - like taking a medicine, writing a blog, asking someone out and so on.
The Result (Y) of the hypothesis is usually a change - like getting cured (before - sick, later - not sick) , getting traffic on your site (before - low traffic, after - high traffic), going on a date (before - lonely and desperate, after - well … you get the picture).
So, here it is - a hypothesis is just a statement of fact that links a Driver to a Result.
Do note that a hypothesis cannot be a general statement of fact, like “Life is beautiful”. It must have a cause-effect relationship - like “Presence of puppies (cause) makes life beautiful (effect)” Another thing to note is that the Cause (X) need not always be an action taken by a human. The cause can also be something else - like “the mass of earth” - leading to an effect like “pulls things to itself”
Some Famous Examples of Hypothesis
Gandhi : “My methods of peace (cause) can win the war of independence (effect).”
Newton : “For every action (cause) there is an equal and opposite reaction (effect) ”
Larry and Sergei : “A better search algorithm (cause) can make search more frequent (effect)”
You : “Knowing the scientific method of experiments (cause) can help me grow my business faster (effect)”
How to write a hypothesis that works
Every scientific discovery and every great business outcome, is driven by a hypothesis at its base. This means that hypothesis is not just a theory - it has great practical significance. A good hypothesis shows the experimenter’s clarity of thought. Therefore it must have the following properties:
Rooted in Prior Research / Knowledge : The hypothesis should not be just a guess. If it’s rooted in prior knowledge, then you will need less number of hypotheses to get the outcome you want.
Testable : You must be able to observe the cause and then measure the effect. A bad hypothesis is one where the effect cannot be measured by you.
Pithy : It should have just one cause, and one effect. If you need more causes or effects, create a new hypothesis, and test it separately.
What is a Null Hypothesis?
Well, for all practical purposes, you can ignore this section. Null hypothesis is used by statistical researchers as a matter of tradition (the hypothetical-deductive model), more than anything else
However, for the curious, read on.
A null hypothesis is (typically) the logical opposite of the hypothesis. For example, if the hypothesis is - Cause X leads to Effect Y, the null hypothesis could be - Cause X HAS NO EFFECT on Y.
A slight more complexity occurs when your hypothesis has a direction. For example, “Cause X leads to an increase in Y”. In this case the null hypothesis could very well be “no effect”. But what happens if you did the experiment and X actually led to a decrease in Y. This scenario was not covered in your hypothesis, nor in the null hypothesis. So while you stand red faced in the shame of your incomplete hypothesis space, you know nature has pulled one on you. To avoid such utter shame and discredit, we’ve been told, statisticians create a complete hypothesis space.
For the aspiring programmers and statisticians out there, please note that a complete hypothesis space is important because your function will eventually be pulled out of it. If your program misses out on completion, you may never have a solution.
So here it is. The hypothesis and the null hypothesis
Hypothesis : X leads to Y
Null Hypothesis : X has no effect on Y
Why to Write a Hypothesis
Every business idea is a hypothesis being proven. In day to day business decisions, we are proving and disproving one or the other hypothesis all the time.
Being aware of the hypothesis allows the best teams to take very objective decisions, track the right things, conclude faster and therefore deliver great results in a short span of time. This is the reason why all growth leaders stress so much on writing down the hypothesis or the ideas that you are working upon.
Every feature, sub-feature, blog, website copy design - essentially everything you do in your business is geared towards growth, which is the effect of your effort. Therefore, every decision you make follows a cause-effect relationship, and is a hypothesis.
As a result, if you structure your decisions in the form of a hypothesis, you will keep a tab on all the relevant properties of your decision - Researched, Testable and Pithy.
The best teams do not say - let us build this feature X. Instead, the best teams say, this feature X will increase the retention R by 5%.
Notice how the latter is much more clear in terms of the goal to be achieved (increase retention) and the test to be conducted (the percentage increase). The greatest advantage is this - the moment you are clear about the test, you start thinking about other ways to achieve the same result.
Quick notes on how to write a great hypothesis
Ideally, you could write down 4-5 ideas for the same goal and test. So to “Increase Retention by 5%” you might write down the following:
Build a Feature X
Improve the onboarding
Send a 3-week email course on how to use the product to new users.
Use chat to teach new users, over 3 weeks, on how to use the product.
And so on.
The breadth of your ideas depend upon your willingness to cover all the possible drivers of the effect. The depth of your ideas depends on how many of these drivers are already in place and in need of refinement.
A Framework to Write Hypothesis
An easy framework to cover your bases is this. First write down all the drivers that can affect the intended outcome. For example, greater retention can be achieved by:
More valuable features (in terms of saving people time, money)
Easy to use features (low cognitive load)
Better communication - staying in users’ heads longer.
Better onboarding - training on using the product to reduce cognitive load.
And so on…
Then choose the things you could do for each of the drivers. For example, for Better Onboarding of new users you could:
Improve the product copy.
Nurture new users with chat and support
Create a product walkthrough
Create product videos and support articles.
Ensure that Sales is bringing in only users who are good fit with the product
Manually onboard the largest accounts.
And so on ….
As you can see, you can come up with tons of ideas on what could be done. It’s super important to properly record all these ideas and prioritise them well. Pick the ones that can be done faster. Then deliver them to a limited user base and then keep measuring. If it passed the test, then commit most of your resources to it.
MIT has this presentation on a practical framework for writing a hypothesis called HOS - Hypothesis, Objective and Success Criteria.