Over the course of this chapter, we've seen how to conceive of, create, and analyze the results of an A/B test.

The statistics themselves are really a continuation of the null-hypothesis testing that we saw in Chapter 7, Null Hypothesis Tests – Analyzing Crime Data. A/B testing provides a nice, complete, useful example of the workflow involved in using null-hypothesis testing and of the power and the help in the decision-making that it provides.

This allows us to use a standard and widely used way of testing exactly what variations on a website drive more interactions and allow us to identify and serve the site's users in a better manner. It allows us to decide on changes to the site in structured, testable ways.

Of course, in actuality, we'll probably want to use an existing service. There are several services out there, from bare bones but free services such as Google Analytics Content Experiments to full-featured for-pay services that cover all aspects of A/B testing, such as Optimizely, Visual Website Optimizer, or Maxymiser. However, knowing what's involved in A/B testing and what the best practices are means that we can evaluate and use these services and get the most from them in a better manner.

In the next chapter, we'll look at applying the data analysis to another part of the Web; we'll analyze how people participate in social sites by looking at patterns of participation in the Stackoverflow (http://stackoverflow.com/) data dumps.