Part I
BASIC ASPECTS OF CUSTOMER SATISFACTION SURVEY DATA ANALYSIS
1
Standards and classical techniques in data analysis of customer satisfaction surveys
Silvia Salini and Ron S. Kenett
Customer satisfaction studies are concerned with the level of satisfaction of customers, consumers and users with a product or service. Customer satisfaction is defined as āThe degree of satisfaction provided by the goods or services of a company as measured by the number of repeat customersā (www.businessdictionary.com). Customer satisfaction therefore seems to be an objective and easily measured quantity. However, unlike variables such as revenues, type of product purchased or customer geographical location, customer satisfaction is not necessarily observed directly. Typically, in a social science context, analysis of such measures is done indirectly by employing proxy variables. Unobserved variables are referred to as latent variables, whilst proxy variables are known as observed variables. In many cases, the latent variables are very complex and the choice of suitable proxy variables is not immediately obvious. For example, in order to assess customer satisfaction with an airline service, it is necessary to identify attributes that characterize this type of service. A general framework for assessing airlines includes attributes such as on-board service, timeliness, responsiveness of personnel, seating and other tangible service characteristics. In general, some attributes are objective, related to the service's technical characteristics, and others are subjective, dealing with behaviours, feelings and psychological benefits. In order to design a survey questionnaire, a set of observed variables must be identified.
In practice, many of the customer satisfaction surveys conducted by business and industry are analysed in a very simple way, without using models or statistical methods. Typical reports include descriptive statistics and basic graphical displays. As shown in this book, integrating a basic analysis with more advanced tools, provides insights into non-obvious patterns and important relationships between the survey variables. This knowledge can significantly affect findings and recommendations derived from a survey.
After presenting classical customer satisfaction methodologies, this chapter provides a general introduction to customer satisfaction surveys, within an organization's business cycle. It then presents standards used in the analysis of survey data. Next it gives an overview on the techniques commonly used to measure customer satisfaction, along with their problems and limitations. Finally, it gives a preview and general introduction to the rest of the chapters in this book.
1.1 Literature on customer satisfaction surveys
Survey questionnaire design, data collection approaches, validation of questionnaires, sampling problems, descriptive statistics and classical statistical inference techniques are covered in many books and papers. This book presents such topics, but also provides a large range of modern and non-standard techniques for customer satisfaction data analysis. Moreover, these various techniques are compared by applications to a common benchmark data set, the ABC 2010 annual customer satisfaction survey (ACSS). For details on the benchmark data set and the ABC company, see Chapter 2.
A non-exhaustive list of relatively advanced books dealing with customer satisfaction data analysis includes Grigoroudis and Siskos (2010), Jacka and Keller (2009), Hayes (2008), Allen and Rao (2000), Johnson and Gustafsson (2000), Vavra (1997) and Biemer and Lyberg (2003). Grigoroudis and Siskos (2010) describe service quality models and the Multicriteria Satisfaction Analysis (MUSA), with examples of satisfaction barometers. Hayes (2008) gives special attention to reliability and validity of questionnaires with a link to customer loyalty. The book by Allen and Rao (2000) is most comprehensive in terms of statistical methods. Although not written by statisticians, it provides a useful and well-written description of techniques of descriptive analysis of univariate, bivariate and multivariate data; it also describes dependent models (linear and logistic regression), explanatory techniques (factor analysis, principal component analysis), causal models (path analysis), and structural equation models. Appendix C of Johnson and Gustafsson (2000) presents an interesting comparison of alternative data analysis methods, in particular considering (1) gap analysis, (2) multiple regression, (3) correlation, (4) principal component regression and (5) partial least squares (PLS). Vavra (1997) covers theories of customer satisfaction and loyalty with several examples of scales, analytic procedures and best practices. Biemer and Lyberg (2003) provide a comprehensive treatment of classical design and analysis of sample surveys.
This book is focused on statistical models for modern customer satisfaction survey data analysis. It addresses modern topics such as web surveys and state-of-the-art statistical models such as the CUB model and Bayesian networks (BN). The book chapters, written by leading researchers in the field, use practical examples in order to make their content also accessible to non-statisticians. Our ultimate goal is to advance the application of best practices in the analysis of customer satisfaction survey data analysis and stimulate new research in this area. As stated in the book foreword by Professor David Hand, we aim to make a difference.
1.2 Customer satisfaction surveys and the business cycle
Statistical analysis is a science that relies on a transformation of reality into dimensions that lend themselves to quantitative analysis. Self-administered surveys use structured questioning designed to map out perceptions and satisfaction level, using a sample of observations from a population frame, into data that can be statistically analysed. Some surveys target all customers; they are in fact a type of census. In others, a sample is drawn and only customers in the sample receive a questionnaire. In drawing a sample, several sampling schemes can be applied. They range from probability samples such as cluster, stratified, systematic or simple random sampling, to non-probability samples such as quota, convenience, judgement or snowball sampling. For more on the different types of surveys, see Chapters 3 and 7.
The survey process consists of four main stages: planning, collection, analysis and presentation. Modern surveys are conducted with a wide variety of techniques, including phone interviews, self-reported paper questionnaires, email questionnaires, internet-based surveys, SMS-based surveys, face-to-face interviews, and video conferencing.
In evaluating the results of a customer satisfaction survey three questions need to be asked:
1. Is the questionnaire properly designed?
2. Has the survey been properly conducted?
3. Has the data been properly analysed?
More generally, we ask ourselves what is the quality of the data, what is the quality of the data analysis, and what is the quality of the information derived from the data (for more on information quality, see Kenett and Shmueli, 2011). Addressing these questions requires an understanding of the survey process, the organizational context and statistical methods.
Customer satisfaction surveys can be part of an overall integrated approach. Integrated models are gaining much attention from both researchers and practitioners (Rucci e...