The book discusses the design of experiments, like A/B tests or response surface methodology, in idealized conditions: ex., measurements are independent, there is one clear business metric, and good quality randomization possible
Later, it looks into the kinds of non-ideal situations that present themselves in practice: measurements are correlated, you having competing metrics, truly random samples are not trivial, and more.
Also, the book discusses ways to manage the risks of running experiments. If you're trying something new, you might make a mess, so
you should take care to limit that For example, you could start each experiment in small size, then scale up as you gain confidence that risks are low.
Dave