Thus a firm 'producing' marketing studies from survey data tries to control all of the different phases: conceptualization, fieldwork, data processing and interpretation of the final results. Our goal here is to study the question of the data processing phase in relation to different solutions to the problem. Before presenting a historical review we would like to outline the basic steps in processing survey data by defining their main characteristics. Naturally, we cannot undertake a full study of electronic equipment currently being used by marketing firms. Rather we shall talk from our own experiences to illustrate the evolution which has taken place in equipment and in the opening of new perspectives to users.
Cluster analysis has become a very popular multivariate method in market research during the last five to seven years. Its wide propagation does this method probably owe to the simplicity of its basic ideas. But contrary to this simplicity stand some special dangers which are connected with the application of cluster analysis. These crucial points are discussed in this paper, rules are developed for a meaningful use and some methodical questions which are still open will be pointed out.
In many cases, data collected in the same questionnaire are measured at different levels: nominal, ordinal, interval or ratio. For several years, new methods allow to analyse qualitative and/or heterogeneous data. They are based on the optimal quantification principle. In this paper we first give a detailed example of heterogeneous data analysis. Then we describe the optimal quantification principle and the most important methods. Finally we give practical examples in marketing research.
The last few years have seen a great development of the use of clustering in market research, either for specific operations or as part of more universal operations. These two types of approach - specific or universal - meet different objectives and constraints. The universal approaches dealt with in this paper in most cases correspond to a policy wish within the firm to develop strategies within its own scale and not limited to one class or range of products regardless of the others, in the same sector of economic activity.
Without wishing to be openly critical of current methods of research analysis I would like to suggest that data is often accepted at face value because it is thought to be uneconomical or inconvenient to subject the data to more critical testing. This paper suggests that the problem is either an historical one in that quick stabs at the computer used to be very expensive, or that the users are a little wary or even ignorant of the benefits or possibilities of interactive computing. By putting interactive survey analysis into perspective I hope merely to generate some reappraisal of analysis methods. By illustrating ways in which terminals can be used profitably I hope to generate the need for a reappraisal.
The paper discusses the application of modern programmable pocket calculators (PPG) in the analysis of marketing data. It first shows the main difference of the two basic 'languages' of programmable pocket calculators, viz. Reverse Polish Rotation (RPR) advocated by Hewlett- Packard and the natural 'Algebraic Operating System' (AOS) by Texas Instruments. Then the paper explains Solid State Software libraries containing the equivalent of a 25 card library in one small module not greater than a postage stamp.