Since the 1980s, diversified products and services have developed worldwide, replacing the early economic and market environment dominated by a large number of ordinary daily necessities. After the 1990s, the electronics and Internet industries began to develop at a high speed, and various high-tech products changed with each passing day life Everywhere in the world. For these huge changes, we need more scientific methods to company And the actual benefits of products and services, not just considering the traditional mass production indicators.
Therefore, in the 1990s, scientists began to make more scientific and effective assessments of different industries, different fields, different products and services in the new economic environment, and began to pay attention to the feedback of consumers or users, because no matter enterprises, industries, departments or products, there are "customers" they face directly customer satisfaction Customer Satisfaction becomes an intangible asset of these contents, which can directly or indirectly reflect the actual benefits, so the Customer Satisfaction Index (CSI) has become a new measurement standard. Here is a brief introduction to the CSI model:
CSI model is not a specific actual satisfaction model, but a methodology to measure satisfaction, and is a set of models to evaluate customer satisfaction. The set of overall concepts reflecting customer satisfaction can be called "macro model", such as value, perceived quality, customer expectation, customer complaint behavior, brand/product impression, customer loyalty, etc. These concepts can all be "macro" indicators to measure overall customer satisfaction, It is called "Latent Variable" in the model. The following figure is a simple example of the CSI model. The one-way arrow shows the relationship between the concepts.
(Source: C. O’ Loughlin and G. Coenders, 2002)
For different applications, different studies have given a variety of model structures. For example, in 1996, Woodruff and Gardial proposed to take the concept of value as the driving force in the relationship between product selection and satisfaction, and form a value chain as a simple psychological response. According to this goal, they established a hierarchical model from customer value chain to customer satisfaction.
The literature on market research points out that different macro concepts need more detailed factors (or elements) to support them, because macro concepts are relatively vague and emotional definitions, which cannot be directly calculated. Therefore, each macro concept can be decomposed into different factors, called "Micro Model". For example, the perceived quality of products can be measured by reliability, overall evaluation after purchase, and functional requirements satisfaction. In the model, it is also called "MV, Manifest Variable". A simple microscopic model can be expressed as follows:
(Source: C. O’ Loughlin and G. Coenders, 2002)
Customer satisfaction index model of countries (regions):
Since the 1990s, researchers have begun to collect data based on the actual situation of countries (regions) to study and establish satisfaction models. Among them, the famous ones are the Swedish Customer Satisfaction Barometer (SCSB), the American Customer Satisfaction Index (ACSI) and the European Customer Satisfaction Index (ECSI). Here, ACSI is used as an example.
The preconditions for establishing the ACSI model include: 1 Satisfaction is derived from customer evaluation and cannot be directly observed, so it is measured as a potential variable with multiple indicators (measurement variables); 2. ACSI not only considers the actual consumption experience, but also pays more attention to the prediction of future prospects. The specific macro structure model is shown as follows:
(Source: Fornell, et al, 1996)
Since ACSI and macro latent variables cannot be directly measured, they need to be calculated according to multiple measurable variables. The following table describes the measured variables in the model. (Source: Fornell, et al, 1996)
The establishment of ACSI model is based on the method of "structural equation modeling" (SEM) to establish a structural relationship equation for different measurement variables and potential variables. For the SEM method, refer to . In other words, potential variables can be obtained through various measurable factors, and then the overall customer satisfaction can be calculated. Different models are based on different actual data. The quantitative data is obtained through questionnaires. The determination of the structure and relationship of the model also comes from actual data, which is obtained by fitting and estimating some mathematical algorithms (such as PLS). As shown in the figure below, before the model is determined, we first assume that all potential variables and measurement variables are related, and then select effective variables for modeling through actual data. In addition, different measurement variables correspond to different questionnaires. At first, we were not sure about the number and content of effective questionnaires. The common method to determine was to find out the questionnaire data most relevant to potential variables through correlation analysis as the source of measurement variables. This analysis method, also known as CFA (Confirmatory Factor Analysis), determines the model step by step by continuously verifying the compatibility of the model and data. For details, please refer to .
(Reference: Ref. [2,3])
Analysis of model characteristics and effects:
Scientific satisfaction model can bring good results. The ACSI model is also taken as an example. In order to measure economic output more accurately and comprehensively, predict economic profits, provide useful information for economic policies, and become an indicator of economic health, ACSI model must meet some standards: accuracy, effectiveness, reliability, predictability, coverage, simplicity, diagnostics, and comparability.
These contents can also be used for reference in product satisfaction assessment, but specific product characteristics and industries need to be improved. Therefore, in the whole process of modeling, comprehensive and careful thinking must be carried out in the selection of potential variables/measurement variables, data collection, data processing, algorithm application, conclusion analysis and solution extraction, so that the model can be continuously optimized in the long-term horizontal (different products, different industries) and vertical (different time) accumulation, This makes the results more comparable. Therefore, in the process of modeling, we need to conduct specific analysis and investigation on each link.
The questionnaire design is specifically mentioned here. The questionnaire is the source of measurement variable data, and the data directly determines the validity of the model. Therefore, the design of the questionnaire is very important in the early qualitative analysis. Necessary adjustments should be made under the large structure of the model, and the design should be carried out according to the specific situation of the product itself, so as to avoid using too general examples. The length of the questionnaire is also determined by the number of measurement variables. For the measured variables of the model, it is enough to select as accurate and relevant questions as possible, because even too many irrelevant questions may be removed from the analysis. This requires screening according to the researcher's own experience and actual situation in the process of designing the questionnaire. If the questions in the questionnaire are lengthy, not only the accuracy of the analysis results may not be improved, but also the abandonment rate of customers in the process of answering questions may be increased and the reliability may be reduced.
In addition, when discussing satisfaction indicators, we need to propose several Precautions ：
1. According to Oliver's definition in 1997, "satisfaction is a reflection of the satisfaction of consumers (or users). It is a pleasant level of consumer satisfaction provided by a product or service, including lower or higher satisfaction……". When modeling the user satisfaction index of products, our target audience should be users, not customers. Because users are users of products or services and customers are buyers, customers may not be direct users of products or services. Therefore, when selecting target groups, we should consciously eliminate inappropriate "customers", so as to improve the information to noise ratio of data and improve the reliability of data.
2. Satisfaction usually has an upper level threshold (beyond the level of user psychological satisfaction) and a lower level threshold (below the level of user psychological satisfaction). That is to say, when users get "too many good things", user satisfaction may decline. Many users focus on the lower level threshold and ignore the upper level threshold.
3. Satisfaction is a feeling. Although various methods can be used to make this feeling quantifiable, measurable and more accurate in modeling, it is always a relatively short-term attitude and may change with the environment and time, Therefore, the product satisfaction needs to be constantly improved according to the suggestions given by the evaluation results, and its model should also be constantly optimized.