Business & Hypothesis Testing. Step-by-step Guide

Tim Ferriss' bestselling book, The 4-Hour Workweek, about the art of managing your life and time, is a great example of hypothesis testing. Testing the best title and description for the book, the author ran more than a dozen ad campaigns in Google Adword. The most clicked result adorned the cover of the future bestseller. This approach paid off: 1.35 million books were sold worldwide! Today, the book is not as popular. But this story of multivariate testing is likely to be written into marketing textbooks.

Without testing, you'll never know which of your actions brought in more customers. It could be a successful ad, an article headline, a sale day, unusual text, price, font colour, photo. Which ad generated more interest? What was the trigger? If you want to earn more and grow faster than the market. If you want to test an idea, an interest, a fascination. Test your ideas and hypotheses.

Let the market choose the best idea
Achieving significant success in dropshipping, and e-commerce in general, is only possible by constantly testing new ideas and optimising all business processes. From content and site speed to delivery clarity and product quality.

Hypothesis testing is the foundation of any business, a systematic process and consistent action: constantly study your audience, your competitors and the market. Set goals, look for options, test, compare, "finish" and repeat.

Important to remember. Testing random ideas that are not based on well thought out hypotheses can lead to a waste of time, money and website traffic.

What to test and how?
By constantly experimenting, observing what your competitors are doing and correctly identifying trends, you can identify whole groups of products for which there is, or is likely to be, increased demand. At the same time, you should take seasonality into account and prepare for seasonal sales at least a month in advance.

To test effectively, you need to understand what you are trying to achieve with the test.
What is your overall objective? (Increase sales, volume or quality of conversions).
Who is your ideal customer?
How do they become your customer (you need to outline this process step by step).
How will your test help you achieve your overall goal? (increase form submissions/purchases/email subscriptions/phone calls).

How do you formulate a hypothesis?
The null (initial) hypothesis and all other hypotheses can be formulated using the formula: "We offer X for Y to solve Z", where X is the product, Y is the audience, and Z is the problem the product solves.

Ideally, you should test each component of the hypothesis. Is the audience right and do they need the product? Is there a problem? Are they willing to pay for a solution to that problem?

Traffic segments
A paying customer is not someone who has money, but someone who is motivated and has a strong desire to buy your product. By segmenting your target audience and analysing each segment, you can measure response by comparison:

New and returning customers
Current and potential customers
specific pages visited
devices used
demographic variations
Geography or languages

Hypothesis testing. Step by step guide
Step 1: The null and alternative hypotheses.
Null hypothesis (H0). Increased sales are not influenced by: advertising channel/seller/target website/ad headline.

Alternative hypothesis (H1): Total sales can be increased by new: promotion channel/vendor/target page of website/commercial headline.

Step 2: Data collection
Collect data on sales, profit and traffic sources. Determine how many sales can be attributed to: each channel / specific product / landing page / ad title. Take into account factors that may affect profitability, such as seasonality, ad campaigns, etc. Select the key metric against which you will evaluate the performance of the test. For tests where you can control the audience, such as sending emails to a customer base, test two or more equal audiences. It is impossible to predict the response to a particular ad in advance, so the length of the test will directly affect the sample size.

Step 3: Statistical test
Choose an appropriate statistical test to analyse the data. For example, you could use a t-test to compare average sales between different promotional channels. To clearly understand what influenced the result, only one variable, one parameter, is tested in an experiment. This does not mean that you cannot test more than one variable. When analysing the response to advertising, the following are tested: headlines, blog posts, images, amount of text, hyperlink clicks.

Step 4: Null hypothesis. Reject or confirm
If the alternative hypothesis is confirmed, we select it. If the null hypothesis is not confirmed, you may want to rethink your advertising strategy or change your data analysis.

Step 5: Conclusions
Which version got a higher response? Which traffic source was the most effective? Which version performed best with certain traffic sources? Once you have completed a test and determined a 'winner' (or found that there was no 'winner'!), you should continue testing. It's important to do everything quickly and systematically: quick test, run a working version, eliminate low converting bundles, test the next version.

Step 6: Iterate the process
Adapt the hypothesis testing process for different advertising channels, different types of ads, different types of content. You can measure the effectiveness of ads on different social networks, or use A/B testing to compare different versions of content.