This book introduces the reader to the niceties of samples (random or stratified random), averages (mean, median or modal), errors (probable, standard or unintentional), graphs, indexes and other tools of democratic persuasion.
There are plenty of lessons about how we should interpret the numbers we come across every day in adverts and (potentialy biased) news reports and there is nobody living in the developed world who can't benefit from the enlightenment that this brings.
The only disappointing aspect of this book is that it's so short, an accomplished reader with some knowledge of statistics could get through the book in a single (if lengthy) sitting.
It is not a long book, and some of the examples are dated (physicians recommending brands of tobacco, for instance), but the meat of the book is both accurate and extremely readable. It covers the ways that statistics can be made to show pretty much anything, both through deliberate manipulation and through simple sloppiness. The main chapters cover issues such as inadequate and biased samples, how to provide subtly and not-so-subtly misleading (though technically accurate) visual charts and representations, how to manipulate perception by eliminating inconvenient precision and adding spurious precision, how to manipulate perception by supplying numbers without context or by simply leaving inconvenient facts out, and how to confuse people thoroughly about correlation vs. cause-and-effect. The final chapter provides a nice summary: the questions you absolutely must ask about any figure you are presented with, in order to judge its worth.
As the author himself says, it may read something like a graduate text on dishonesty, but one can assume that people who deliberately wish to mislead have figured out how already; this is to educate the honest person who wishes to be alert. It is frequently used as a text in undergraduate statistics courses, for good reason.