For the analysis of designed experiments, Taguchi uses performance criteria that he calls signal-to-noise (SN) ratios. Three such criteria are here denoted by SNT, SNL, and SNS. The criterion SNT was to be used in preference to the standard deviation for the problem of achieving, for some quality characteristic y, the smallest mean squared error about an operating target value. León, Shoemaker, and Kacker (1987) showed how SNT was appropriate to solve this problem only when σy was proportional to μy. On that assumption, the same result could be obtained more simply by conducting the analysis in terms of log y rather than y. A more general transformation approach is here introduced for other, commonly met kinds of dependence between σy and μy (including no dependence), and a lambda plot is presented that uses the data to suggest an appropriate transformation. The criteria SNL and SNS were for problems in which the objective was to make the response as large or as small as possible. It is argued here that these predecided “portmanteau” criteria can provide an inadequate summary of data and that, regarded as measures of location, they can be extremely inefficient. In preference to such performance criteria, the merits of simple methods of data analysis that can uncover information both expected and unexpected are urged. A reanalysis of an interesting experiment due to Quinlan (1985) illustrates the value of this approach and its contribution to the art of discovery. It is argued that improvement of quality will best be catalyzed by engineers using elementary data analysis with computer graphics rather than by those trained only to employ more rigid predecided criteria.