J. Soc. Cosmet. Chem. 22 211-229 (1971) ¸ 1971 Society of Cosmetic Chemists of Great Britain Sensory testing--a approach statistician's W. A. PRIDMORE* Presented on 22nd April 1970 at the Symposium on "Perfumery", organised by the British Society of Perfumers and the Society of Cosmetic Chemists of Great Britain, at Eastbourne, Sussex. Synopsis--This paper shows ways in which the statistician can assist the cosmetic chemist in SENSORY TESTING by (1) employing his knowledge of PROBABILITY, (2) stressing the importance of a fuller understanding of the purposes for which the sensory tests are being used, and (3) stressing the importance of EXPERIMENTAL DESIGN in getting the full benefit from any set of tests. It sets out the theory behind certain forms of tests of differences and preferences. INTRODUCTION This paper deals with some techniques devised by one group of statisticians to help the cosmetic chemist in the appraisal of new product formulations, in the development and modification of existing products, and in the routine control of raw materials and of production. It is always difficult in any experimental or investigational process with which a statistician is associated to distinguish between that part of the plan which is directly attributable to the statistician's intervention and the basic experience, scientific knowledge and technical know-how contributed by the rest of the team. It is therefore to be understood that, whilst this paper will stress the statistical and probability aspects of the work which link together the various parts, most of the ideas have come from the statistician's technical colleagues, sometimes as a spontaneous contribution, sometimes in response to the statistician's awkward questions. *Reckitt & Colman Products Ltd., Hull, Yorks. 211

212 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS The statistician can only work as a junior member of a team he takes the ideas of others and helps them to put those ideas to work as efficiently as possible. At the start, therefore, this statistician acknowledges that most of the good ideas presented have originated in the first place from other people any nonsense is his own responsibility. We shall be confining ourselves to that important but limited field of investigation in which the research chemist seeks to supplement his own judgement about the sensory characteristics of the products he is dealing with by calling upon the judgement of other judges or members of panels who are available at short notice in the laboratory area or accessible to it. THE STATISTICIAN'S ROLE A statistician may be asked to provide some analysis of results obtained by going to outside groups of potential users of products scattered through- out the country on a nationwide basis. Such market research tests, market placement tests, consumer panel tests, call them what you will, have an unavoidable tendency to be time consuming, costly and difficult to organ- ise. The inevitable delays imposed by the problems of packaging for distribution over the whole country, of the distribution of product and the collection of reports are all difficult to accept at a time when commercial executives are breathing heavily down the formulator's neck. Some quick and simple approximations are absolutely necessary. On the other hand, the statistician has known of the opposite dangers the instant decision achieved by the chemist's own personal choice, which so often leads to expressions of acute surprise when larger scale tests fail to confirm the results from that sample of one. A very similar phenomenon concerns the use of the managing director's wife she rarely constitutes a very typical market for the products in question. In the first instance, the statistician may be required to advise on cheap and speedy versions of the national consumer panel of testers in the second instance he may be required to provide some systematic substitute for the one or two judgements by the cosmetic chemist and his colleagues at the next laboratory bench. From whichever direction the approach is made, we tend to end up with a small-scale sensory testing panel, in some cases composed entirely of the non-technical office or factory staff immediately available on the work site, in other cases composed of specially recruited outside groups of people called together for special sessions to some con-