I remember a US colleague commenting, in the mid 1980s, on the predilection of deans and other university managers for assessing academic statisticians’ performance in terms of the numbers of papers they published. The managers, he said, “don’t have many skills, but they can count.” It’s not clear whether the management science of assessing research performance in universities has advanced greatly in the intervening quarter century, but there are certainly more things to count than ever before, and there are increasingly sophisticated ways of doing the counting. The paper by Adler, Ewing and Taylor is rightly critical of many of the practices, and arguments, that are based on counting citations. The authors are to be congratulated for producing a forthright and informative document, which is already being read by scientists in fields outside the mathematical sciences. For example, I mentioned the paper at a meeting of the executive of an Australian science body, and found that its very existence generated considerable interest. Even in fields where impact factors, h-factors and their brethren are more widely accepted than in mathematics or statistics, there is apprehension that the use of those numbers is getting out of hand, and that their implications are poorly understood. The latter point should be of particular concern. We know, sometimes from bitter experience, of some of the statistical challenges of comparing journals or scientists on the basis of citation data—for example, the data can be very heavy-tailed, and there are vast differences in citation culture among different areas of science and technology. There are major differences even within probability and statistics. However, we have only rudimentary tools for quantifying this variation, and that means that we can provide only limited advice to people who are using citation data to assess the work of others, or who are themselves being assessed using those data.