500 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS The discrimination must adequate. The conditions under which the test is carried out must be realistic. If all these are met and the product survives, then: The technique should be published so as itself to survive peer judgment. which can result in: The technique is accepted as a national or international standard by which we and our competitors will be judged. This catalogue is severe, but I believe more widely achievable than generally realized. The outstanding example is probably in the dental field with reduction of caries incidence, where because the evaluation problem is so difficult, but the goal so worthwhile, the research resource expended by the industry has been by any standard vast. It was recognized early that what- ever laboratory methods might be available for mechanistic screening, the ultimate test must be in use, over long periods, and with large numbers of subjects to cope with biological variations. Research workers tried a variety of protocols, all subject to one or another flaw, either procedural or statistical. Eventually, and with not a little prompting from some of us, an inter- national working party of leading clinicians produced a definitive protocol for the conduct of such trials, which is kept under constant review (1). Improvements are still being made to the statistical treatments so that even more information can be extracted from these very expensive and time- consuming trials. Unfortunately agreement has not yet been reached as to how the results of such trials are to be presented to the public and there is a real difficulty in communicating such a complex set of data without either inadvertently misleading, or only conveying ineffective generalities. To quote Horowitz et al. (1). 'The practice of presenting results as percentage reductions may be misleading. In the interpretation of proportional caries scores, it is essential to pay equal attention to the numerator and denominator of the scores. For example, if the mean increment of new carious surfaces in the control group is only 0.5 per child, and if in the test group the corresponding value is 0.25 per child, this represents percentage reduction of 50•o. There has only been, however, a saving of 0.25 carious surfaces per child. One must

THE PROMISE AND THE PRODUCT 501 therefore critically look at the absolute values involved when one evaluates percentage reductions.' And I might say that similar comments can be made about all simple percentage-type reports, such as are used to communicate anti-perspirant effectiveness. These should relate to the scale of effect which the average consumer can in fact notice. In a very different field, consider hair sprays. As a result of background work and development, let us suppose that an improved product has been made which is designed to be superior to existing own and competitive brands in respect of: Easier brush-out. More efficient brush-out resulting in less build-up on the hair. Improved feel of the hair after brushing-out. Evaluation science can provide the evidence not so much to back a pre- conceived promise as to arrive at the promises which can confidently be made. The evaluation package would include: (2), (3) Combing resistance (instrumental techique). Curl retention ,, ,, ß Sensory assessment of brush-out and feel. Build-up (bench technique). The results are collected in a support package for the benefit of market- ing in its advertising development and relations with control authority. The need for consumer panels for independent uncontrolled judgment can never be, nor should be, eliminated by use of instrumental techniques or expert panels, even when applied to in-use situations. After all, the ulti- mate consumer panel is the market place and we expect to succeed in that most critical of all tests. Much has been published on operating techniques in this area and I do not intend to repeat this. However, such panels are particularly valuable, not only as a confirmation that the functionality sought and demonstrated in the laboratory does in fact survive, but also as a research tool for studying interaction terms otherwise inaccessible to the controlled experiments. Assessment techniques in development programmes Evaluation of functional performance in use calls for techniques and protocols based on the in-use situation and, where appropriate, which it