The social mission of the Platform voor Klantgericht Ondernemen (PvKO [Platform for Customer-focused Entrepreneurship]) is stimulating more customer-focus in the Netherlands for the benefit of our Gross National Happiness (strictly positive customer experiences) and our Gross National Product (successful entrepreneurial organisations). For all organisations that want to work in a customer-focused manner, it is important that they can objectively assess their own level of customer-focus. This assessment is more valuable if an organisation can compare itself with other organisations and can receive input with regards to possible areas of improvement to reach a higher level of customer-focus.
Within the PvKO, the core domain Research, in cooperation with the Rijksuniversiteit Groningen (RUG), has taken a first step in the development of a benchmark tool “Customer-focused Entrepreneurship”. In this article we describe the manner in which we went about this, the results found and the way in which we plan to proceed with the tool.
For the development of the benchmark tool, the steps were undertaken as displayed in Figure 1.
After the PvKO learned that the RUG was prepared to help with the development of the tool, the RUG selected three students who could sink their teeth into this challenge. They saw this as a rewarding subject for the completion of their degree. Cmotions also made an important contribution. At the bottom of the article we provide an overview of the people who helped with the development of the benchmark tool.
The project group agreed that the article “The Path to Customer Centricity” by Denish Shah, et al., should serve as the starting point to develop the conceptual framework of the assignment. A critical analysis of scientific literature in the field of “customer centricity” eventually led to the following definition of customer-focus:
“Customer-focus is the degree to which an organisation, based on individual customer knowledge, meets the needs of individual customers and in doing so, takes the preferences and wishes of the individual consumers as the starting point for making decisions, with the precondition that customer-focus also results in a profitable organisation.”
Published literature supports the assumption that if organisations are more customer-focused, this contributes positively to the company results through higher customer results. Drivers of customer-focus are thought to be: the structure and culture of the organisation, the (customer) processes and the degree to which an organisation focuses on customer-focused criteria and is able to link these to financial ones.
The conceptual model based on the article by Shah, et al., that served as the starting point for the development of the benchmark tool, is displayed in Figure 2.
Each of the seven sections in the conceptual model can be seen as a “construct” that needs to be made quantifiable, or: for which a collection of questions need to be asked which all regard an element of that specific construct and together provide a good overview of it. Based on existing scientific literature, each of the units was filled with good quantifier questions. Most of the quantifier questions found in literature are in the form of a Likert scale. This is a statement for which one can indicate to what degree one agrees with it, or to what extent it applies to the situation of the respondent. These are frequently based on 5 or 7 point scales, varying from “strongly disagree” to “strongly agree”. An example of such a statement is:
“We enable our customers to have interactive communications with us”
The RUG and the project group spent a considerable amount of time discussing and determining the questions per section. The phrasing of each question is important, the goal being that each question is interpreted in the same way by each respondent. This process led to a total of 112 questions, with subjects such as empowerment, responsibility for the customer journey, the presence and use of customer data, and the realisation of goals.
For this basic research the questionnaire may be longer than the list that is used in the final benchmark. After all, analyses of the results show which questions per section significantly contribute to the model found and which contribute (much) less. The final set of questions must be as limited as possible and only contain the most discriminating questions, this in order to maximise the willingness of people to participate in the benchmark.
The questionnaire developed was then tested on several PvKO members, who provided useful feedback with regards to the length of the questionnaire and the phrasing of the questions.
The fieldwork with the final questionnaire was conducted by Marketresponse. Via several PvKO members, a database of approximately 4,000 marketing and customer managers was created of companies with at least 30 FTEs. These managers were invited to participate in the research via e-mail, and in exchange for this were provided insight into their own customer-focus compared to other companies, as well as an invitation to a seminar. After the initial invite several reminders were sent.
In the end, the response rate turned out to be disappointing. Of all the managers that were contacted, 162 managers completed the questionnaire. Of the respondents, 23% worked in the financial services sector, 17% in retail and 16% in industry. Of these, 52% indicated that they worked in a B2B organisation, 21% in a B2C organisation and 27% in an organisation that serves both sub-markets. Approximately 71% of the respondents worked for an organisation with fewer than 200 FTEs.
Because the primary goal here is to determine the collection of questions with which the supposed connections can be determined, the testing of the representativeness of the realized sample is less important.
In order to determine which questions should be used to be able to research the supposed connections, it is first important to find a (latent) construct within each section that properly represents it. One way of doing this, is factor analysis. Factor analysis is a statistical technique that makes it possible to determine whether there is a hidden shared factor behind a string of questions (or aspects). Under factor analysis, a researcher checks whether separate questions can be traced back to one or a limited number of factors. If so, it can be turned into a linearly weighted composite variable.
Take for example the construct “customer-focus”. In order to measure this, the respondents were presented with two statements:
The combination of these two statements yielded a factor that was very usable.
In this manner, factor analysis was executed for all seven sections. Subsequently, the relationships between the various sections were assessed by means of regression analyses.
The results of these estimates were disappointing. Only a limited amount of connections could be found. We were therefore unable to determine the conceptual model in this manner. On the one hand, the reason for this can be found in the relatively low number of returns and in the item non-response, i.e. the fact that respondents did not answer all the questions. On the other hand, the relevant literature may not have provided sufficient direction for the measurement of the constructs.
In a subsequent step, several alternative analyses were performed. For this, the original seven sections were first subdivided into three groups, namely:
All the questions within these three groups were assessed integrally. As a result, the a priori division in the constructs of Shah, et al., was abandoned. This is particularly true for the driver and the result variables.
In particular within the result variables group, there was a relatively high degree of item non-response. This concerned questions with regards to having specific goals and/or the measurement of various customer metrics. Metrics such as customer feedback, Customer Equity and Share of Wallet were included. Because the factor analysis did not prove to be very successful for these variables, the decision was made to use a Mokken scale analysis. The Mokken scale analysis determines which questions form a hierarchy. The scale that results from such an analysis can be seen as a hierarchy of a specific attribute or skill. When a company is rated higher on this scale, the company is more skilled with regards to the aspects that form the scale than with a lower score. It can be compared to a numeracy test at elementary school. For this, several questions can be used, organised from easy to hard. The more correct answers a student gives, the more skilled he or she is. For example, if the child correctly answers the fifth question, it is likely that he or she also got questions 1-4 right. If another question is answered correctly, it is most likely to be question 6.
For the result variables we found a strong Mokken scale, which includes, among others, the following elements:
The percentages indicate how many companies in the sample do this. Many companies aim for customer satisfaction, but few companies do this for customer equity. The scale implies that when a company indicates it aims for customer equity, this company also aims for customer retention and customer satisfaction.
The Mokken scale did not only provide the best results for the result variables, but also for the driver variables. In the end, ten scales were constructed:
LISREL was used to test the relationships between the scales. LISREL can be seen as a combination of factor analysis and regression analysis. When using all the questions, this yielded usable models. Subsequently, the question was which of the questions contributed most to the scale in question, and whether a good model could also be produced with this set of questions.
After all: for the benchmark tool, we want to work with a limited number of questions in order to maximise the willingness to participate. For this purpose, the most significant items were selected and Mokken scale and LISREL analyses performed once again. As with the previous analyses, this again resulted in ten scales. These scales resulted in the easily interpretable LISREL model that is displayed in Figure 3.
The Mokken scales on which the constructs are based, can be used relatively easily to determine how high an individual company ranks with regards to customer-focus. However, from a scientific point of view, the following should be noted: as indicated, the sample set for the research was on the low side. In order to validate the model, more responses are necessary. This should provide motivation for many companies to participate in the benchmark.
The following representatives participated in the project group of the PvKO: Linda van Zomeren (Post NL), Ton Timmermans (Board of PvKO), Wolter Kloosterboer (Marketresponse), Victor Blom (Efficy) and Ronald Wiekenkamp (Cmotions). Contributions from the University of Groningen were forthcoming from: Dr. Janny Hoekstra and students Lisette Klaver, Tomas Geerts and Sicco Hempenius. Prof. Dr. Ton Kuijlen played an important role in the development of the final model. PvKO is extremely grateful to Marketresponse for conducting the field work.
19 February 2020
Quality within customer service Just like the world around us, customer service is evolving at a... read more
12 July 2019
Cmotions and Broad Horizon join forces with the launch of Tailo. Combining data science with the... read more
17 July 2018
As if our daily portion of “working with data” wasn’t enough, at Cmotions we are also... read more