Statistics may be defined as the collection, analysis and interpretation of numerical data. As market research is concerned mostly with counting and measuring, it is not surprising that the theory of statistics can play an important part in assisting researchers to collect valid samples of data, and in helping them to draw correct conclusions from those data. In this chapter, the emphasis will be on the analysis and interpretation aspects of statistics dealing mainly with simple descriptive measures calculated from survey data and the testing of hypotheses about those data. The techniques and significance tests described below are those which have been found most useful in interpreting survey tabulations. This is not an exhaustive list, by any means, and the reader who wishes to know more about statistical analysis in market research may consult the texts given in the references at the end of the chapter.
This chapter describes the data processing and data analysis stages of the quantitative market research process. These steps follow after data have been collected by interviewers, respondents or via automated registration systems. They are followed by in-depth analysis and reporting of results. Data processing concerns activities and technologies which prepare the collected data for analysis: data checking, entry, coding, and editing. Data analysis concerns activities and technologies which provide statistical insight in the collected data: weighting, tabulations, response analysis and analysis of interviewer performances.
Factor Analysis (like the similar technique of Principal Components Analysis) is a multivariate statistical method used primarily for data reduction. Itis usually employed to reduce the columns of our data matrix, that is the variables measured for each respondent. The reduction Is done by grouping together those variables which are intercorrelated as measured by the coefficient of correlation.
Clustering normally operates on the rows of the data matrix. It seeks to find groupings or clusters of respondents who exhibit similar patterns in terms of the variables measured.
With the expanding popularity of "brand-image" and market segmentation studies, the use of statistical data processing methods has increased. In particular, one can note an increase in the application of multivariate analysis techniques to quantitative investigations of the relationships between consumer typologies, attitudes and purchasing behaviour. In this paper, we discuss various problems associated with the use of "packaged" computer programs designed to effect the required multivariate analyses. Although these programs can and often do provide adequate results, it should be remembered that each set of survey data has unique properties. In some cases, these properties can jeopardise the validity of the whole analysis. The techniques we discuss are: a) Principal component based analyses. b) "Cluster-analysis" type techniques. The paper falls into two sections.
The statistical sampling technique is nowadays generally used in surveys on the particular conditions and problems of the market for several products as well as in opinion surveys. The sampling technique is to supply the "most efficient" results by complying both with the necessity of exact information and of a minimised cost. The latter condition can be realised using the smallest sample size for the requirements of the research. For this purpose, the stratification appears to be a useful means by which the survey's field can be divided in sufficiently homogeneous segments, so as to obtain a smaller sample that the one obtainable with a simple random sample at the same level of statistical significance. More detailed aspects are to be taken into consideration, such as the necessity of attributing to the single strata a non-proportional weight; this situation can also exist after effecting the survey since more interviews can fail in some strata: A non-proportional, i.e. a biased sample, entails the problem of evaluating the reliability of the results obtained, for the purpose of assessing the correct estimate. This particular aspect will be dealt with in this note.
In the field of marketing, a prediction is only of value if it has a practical use and this is dependent upon the following concepts: 1.The suppositions underlying the prediction are clearly defined, and; 2. they are realistic. These points are well known. However the main problem of prediction is to find suppositions which are realistic and which occur often enough in the real world to be of value. From the study of numerous instances of new product launches, a special type of prediction relating to new products has been developed by the Attwood Group of Companies arid is the subject of this joint paper. The basis of this technique is the analysis of purchasing histories of members on a consumer panel. When a new product is launched, the following types of questions are usually asked: 1. What percentage of the population will buy the product in a period, say a month, at this time next year?; 2. What will be its brand share, at that time ? The first part of this paper, my contribution, deals with the first point. The second part, developed by my collegue Mr. Dennis, deals with the latter point. In it we intend to demonstrate how, with the analysis of consumer panel data, it is possible to give answers to these questions both accurately and sufficiently quickly after the introduction of the new brand, to ensure that valuable information on its future success is obtained at the earliest possible time. The statistical methods used for both parts of the paper are different and fully independent.