When assessing a health-care intervention the main question is--does it work? Often, a more difficult question needs to be answered--how well does it work? Systematic reviews and meta-analyses help to provide answers to both questions. Too often though, the results are expressed in a way that leaves the reader asking, 'what is the result?' Numbers-needed-to-treat is a simple method for conveying the answers. It can be applied to any chosen clinical outcome with dichotomous data, and the results can be understood by doctor, patient and the public. This paper discusses the concept of numbers-needed-to-treat and gives worked examples using trials in H. pylori eradication and lowering of serum cholesterol.