The subjects we discuss in this seminar influence major marketing decisions - but are still controversial. Simple, practical assumptions or theories govern most of our decisions in designing, testing and evaluating our marketing activities. But these assumptions are themselves rarely tested and their limitations are almost never spelled out. Our purpose in this seminar is to record where the research industry stands on these issues today. We should start from some common ground. For example, that the effects of marketing activities vary considerably, by execution, across product categories and across brands in these categories. That sales matter more than diagnostics like awareness. That share of category sales is an important objective, but there can be other effects of our activities than on volume share and in the short term. Better, more realistic theories will help us when we assess, after the event, what benefit we got from our marketing investment. We also need theories in order to plan these activities properly, to pre-test them, and to budget for them. The full range of research methods is available to investigate these subjects, from small-scale qualitative work to major numerical analyses. Any solid contribution is welcome. However, we are tackling here what may be the hardest, as well as the most important, of all market research problems.
What strategy should be adopted by researchers who want to help media planning? The traditional route has been to provide databases on media exposure, and these will continue to be required though their design, and access to them, could be improved. A study of actual media decisions and campaign evaluations suggests that media exposure data play a comparatively small part in media choice. The way forward is unlikely to be by providing a greater depth of information in this area. It is suggested that researchers study the users' needs more carefully, for example by observation (media case histories) and by consultation. Eight methodologies in data collection and analysis are recommended.
An aggregate, numerical model is proposed to assess the effects of advertising. It is intended as an aid to advertisers and to an advertising agency. The main decisions it assists are about the size and allocation over time of the advertising budget, the evaluation of its effects - and the price of the brand. Several years of experience in trying to measure how advertising works are summarised in the application of the model. The parameters estimated from data about the brand and its market are used in a series of sub-models which help practical decisions. The research data needed are described, as is the approach to the analysis of each individual case. The examples and references given cover a cross-sectional description of brands' position in a market, the estimation of the return on advertising, different shapes of response function, determining the direction in which to change the advertising budget, and how long it is likely to take for the volume sales and profit benefits to be seen.
The book reproduces and comments on a selection of ESOMAR papers. They have been chosen to represent the most useful research methods and the most illuminating case histories.
The problem considered here is the measurement of advertising effectiveness, especially when the results are to be used to help decide how much to spend and what media types to use. We have found two techniques useful: area tests; and consumer surveys in which the informants media exposure is estimated as well as her behaviour, awareness, etc. Area tests by themselves are often disappointing. Variations between areas make the results hard to interpret; tests may be abandoned or conclusions drawn from them too early. Single-source data, including media-product surveys, are now a well-understood aid to campaign planning; they are less often used but are also very helpful in post-campaign evaluation. Despite technical weaknesses and the caution needed in interpretation we have found their analysis often indicates how advertising is working. A case history is given in which media-product surveys were carried out in addition to area tests of both advertising weight and media type. Measurements made before the campaign allowed us to rule out any natural association which might have existed between media habits and product awareness. The results showed, contrary to a naive interpretation of the area test results, that press was effective in combination with television.
Media research in different countries should be more similar than it is. The variety is due to the different scales on which research can be afforded and to the different media available for advertising and hence the incentives for commercial research. Until it develops techniques closer to measuring advertising effectiveness, and not just exposure, no motive to make media research more uniform will be powerful enough to change this pattern. But there are examples of common techniques (TV meters plus diaries, long self-completed media and product questionnaires) and methods will continue to be imported. This paper struggles against these difficulties. It tries to explain why they exist and why change is slow. But successful exchanges do take place and some are mentioned. And some developments in Europe which may be of general interest have been selected for brief description.
Segmentation is one of several multivariate analysis methods which can be applied to market research data; it can be carried out in several ways. This paper discusses briefly some of the possible techniques. The purposes of segmentation which we have experience of are the practical ones of defining marketing objectives, planning campaigns and briefing creatives. We are not interested in statistical 'explanation' as such, but in helping people, often enumerate, to take better decisions. We find segmentation can do this successfully. In discussing techniques we concentrate not on mathematics or programming but on input strategy. We have used clusters created in four different ways, depending on whether we use the whole sample or a special group, and on whether we input descriptive or product data. The example given is from the UK convenience food market, where product data was used on a large sample to create nine clusters. These are described with examples and the use of such data is discussed.
We have used in our agency two methods particularly for investigating the effectiveness of advertising campaigns. These are area tests and media questions on product surveys. The main object of this paper is to describe a case history in which both these methods were used - this is given in Section K. In the following two sections we outline our recent experience with each of these methods when used on their own.
The purpose of the work upon which this paper is based was to examine the implications, for the immediate purposes of current media planning in the United Kingdom, of an important experiment reported to the ESOMAR-WAPOR Congress 1967. We owe a particular debt of gratitude to John Parfitt and to Attwood Statistics Limited, who prepared for us a special punch card pack based upon their original data. Without their interest and collaboration it would not have been possible to continue this investigation.
Our paper is about the principles of data collection and usage. The integration of media and product research, the construction of a computer data bank, an explicit model of how media reach people, closer blending of media buying with media planning, the evaluation of advertising performance - we believe that all these are desirable. We outline examples of advances in technique but these are not given in detail and in any case are not beyond improvement. We do not describe just one computer program, but a family of programs with a variety of objectives. They are linked by the data they use and by their systematic approach to the media planning problem.
People do not separate neatly into regular readers of a publication and non-readers; some read issues occasionally. This paper shows that it is important to collect data on the regularity of reading, and to know how many people read none of the issues in which a series of advertisements has been placed, how many read just one issue, how many read two and so on. This information is called an exposure distribution. Different models describing this situation have been proposed by Agostini, Prankel, Hyett and Politz. The paper suggests another flexible family of models, which has certain advantages. It shows how the models can be used to help in choosing between publications. The advertiser's real concern, even more than this information, lies in the exposure distribution for all the issues on his schedule of different publications. The paper indicates a way in which this joint exposure distribution can be estimated when the two publications duplicate normally and perhaps even when they do not.