How to Determine Customer Satisfaction
Building the Best Customer Satisfaction Survey
Customer satisfaction is a variable that can be represented as a continuum ranging from "not at all satisfied" to "completely satisfied." Some value between these extremes represents the level of satisfaction for a particular customer. Generally, market researchers consider the agreement between the point on the customer satisfaction scale and the actual opinion of the customer to be inexact.
However, it is feasible and even likely that the point on the scale approximates the customer's satisfaction. As it is an approximate value, the market researcher allows for a small degree of error. Because of this small approximation error, a market researcher would consider customer satisfaction to be a latent variable.
Latent Variables vs. Manifest Variables
Latent variables are those concepts in psychology, sociology, economics, and other social sciences that cannot be measured explicitly. For instance, market researchers are often interested in consumers' motivations or attitudes. But these concepts, like the concept of satisfaction, cannot be measured directly in the same way as, for example, age, weight, or level of education. These demographic attributes are referred to as manifest variables, as they can be measured explicitly; they manifest in a tangible form.
Theoretically, scientists generally agree that for every latent variable being measured, several manifest variables should be associated with that variable. In this way, it is possible for the market researcher to explore the relationship between a latent variable, which cannot be measured directly, and several manifest variables, which can be measured directly.
Developing Survey Questions
Customer satisfaction can be measured well through the use of survey questionnaires. It is helpful to craft a number of questions that measure on a scale the degree of satisfaction or dissatisfaction experienced by a consumer. Though satisfaction is infinitely variable, for practical reasons, a satisfaction scale would need to be limited. The customer should be afforded sufficient flexibility in his or her response so the match between the customer experience and the response on the scale closely relate.
Customer Satisfaction Scales
Scales used to indicate customer satisfaction are often 5-point, 7-point, or 10-point, such that zero always represents the highest degree of dissatisfaction. On a 5-point scale, a customer would be asked to select a response to a question from the following set of alternatives: (1) Very unsatisfied, (2) Moderately unsatisfied, (3) Neutral, (4) Moderately satisfied, or (5) Very satisfied.
For each component of the satisfaction survey that respondents are asked to consider, there should be three related questions that represent manifest variables. The questions should be written so that it is easy to match the question language to the aspects of the survey components. For instance, if the market researcher is interested in measuring the component ease of doing business with a company, then the questions could address the speed of transactions, the usability of the website, and the live-chat customer service experience.
Satisfaction Survey Length
The survey questionnaire should range from about 15 to 35 items, each of which addresses some aspect of the components of customer service being measured. In addition, some of the questionnaire items should be directed toward learning more about the customers, not just their opinions, in order to support market segmentation analyses.
Strong analyses of customer satisfaction will include mathematical and statistical methods of data analysis. An objective of the analysis is to estimate the relationship between the manifest variables and the latent variables, and among the latent variables.
A commonly used method to conduct this type of analysis is a structured equation model (SEM). The fit between the model and the data will be gauged against some criteria or a single criterion, such as the capacity to minimize deviation from the actual observed data. This statistical method determines the weighting attributed to the relationship between the latent variables, rather than the subjective opinion of the market researcher.
The reliability of each manifest variable is calculated, the contents of the latent variables are derived, and the relationship between the latent variables are calculated. At this point, the market researcher is able to see whether the estimated model actually fits the data to an acceptable extent, typically by using the coefficient of determination, which is the level of explained variability in a data set.