Enhancing Service Quality—A Customer Opinion Assessment in Water Laboratories through Artificial Neural Networks

Abstract

Existing literature presents multiple perspectives on quality within organizational contexts. Although these perspectives may differ, they universally emphasize the importance of meeting customer expectations regarding products/services. Consequently, organizations are dedicated to addressing customer requirements to foster elevated satisfaction levels. This study aims to assess customer satisfaction in water laboratories and develop a predictive model using artificial neural networks to improve service quality. A methodology was devised, integrating principles from thermodynamics with logic programming for knowledge representation and reasoning. Data were collected from 412 participants of both genders, aged 22 to 79 years old, using a questionnaire covering six specific areas, i.e., customer service, quality of service provided, support documentation, technical support, billing and payment, and online services and tools. While customer opinions were largely positive, the study identified areas for improvement, including clarity and effectiveness in responses to inquiries, reliability of results, clarity of analysis reports, usefulness of test interpretation guidelines, inclusion of legal information, billing options, and online services. Differences in satisfaction were noted based on socio-demographic factors such as age and academic qualifications. The findings offer a framework (an ANN-based model) for future evaluations and improvements in services, highlighting the importance of addressing specific customer needs to enhance satisfaction.

1. Introduction

Water quality is a critical aspect of public health, environmental well-being, and sustainable development. Safe drinking water must be free from chemical, biological, and physical contaminants to ensure it is safe for consumption and use. Water quality directly impacts human health. Waterborne diseases, such as cholera and dysentery, are prevalent in regions with inadequate sanitation and limited access to treated water. Ensuring clean water significantly reduces disease rates and improves the quality of life. Furthermore, good water quality is essential for agriculture and industry. In agriculture, irrigation with contaminated water can compromise food production, while in industry, water quality affects process efficiency and equipment longevity. Thus, water monitoring and treatment are essential priorities.

In Portugal, public water systems serve 96% of the population, but this coverage falls to 80% in rural areas [1]. Despite this, many rural inhabitants use private water sources for drinking and recreational activities. Thus, water laboratories are essential for owners to verify water quality. In addition to performing analyses, some labs offer to collect and transport water samples from wells, boreholes, and pools.

In recent years, laboratories have sought to respond to economic, technological, and social challenges by adopting new organizational models and creative, innovative technological solutions to satisfy customers and become competitive against the competition [2]. Indeed, with increasing competition, customer satisfaction is a key factor in retaining and fostering client loyalty, as dissatisfied customers may not return [3]. Even without making complaints, dissatisfied customers can spread negative opinions and deter other potential clients [4]. Defining satisfaction is not an easy task, as it is a dynamic and subjective concept. According to Jones et al. [5], customer satisfaction is defined as the evaluation of performance, which can be either positive or negative, depending on previous interactions with the organization. Kotler et al. [2] suggest that the level of satisfaction is related to perceived quality, with satisfaction being a function of perceptions and expectations. If perceptions fall short of expectations, the customer will be dissatisfied; if perceptions meet expectations, the customer will be satisfied; and if perceptions exceed expectations, the customer will be very satisfied.

ISO/IEC 17025 standard plays a pivotal role in enhancing quality and customer satisfaction within testing and calibration laboratories. This adherence to quality management principles directly impacts the credibility of the laboratory’s results, fostering trust and confidence among customers. The standard mandates rigorous procedures for method validation, equipment calibration, and staff competency, which collectively reduce errors and improve the consistency of outcomes. Moreover, the emphasis on impartiality and confidentiality protects customers’ interests and data, further enhancing their trust in the laboratory’s services. In essence, ISO/IEC 17025 helps laboratories to consistently meet or exceed customer expectations, thereby driving customer satisfaction and loyalty through a demonstrable commitment to quality and excellence [6].

In today’s competitive landscape, understanding customer satisfaction is crucial for any business. For a water laboratory, this is particularly important as it reveals how well the laboratory meets customer needs and expectations. Analyzing feedback allows the laboratory to identify areas for improvement, enhance service quality, and ensure operational efficiency. This not only helps retain customers but also builds a strong reputation, fostering trust and credibility. Therefore, this research aims to assess customer satisfaction with the services provided by a water laboratory.

The key contributions of this study include a detailed analysis of customer opinions on service quality within a water laboratory, highlighting specific areas in need of improvement. The study provides a comprehensive evaluation of customer feedback, pinpointing critical service aspects that require enhancement. It also offers insights into how satisfaction levels vary across different socio-demographic groups, guiding the customization of service improvements. Furthermore, the study introduces a predictive model based on artificial neural networks (ANNs), designed to anticipate factors influencing customer satisfaction and provides a structured framework for future service evaluations and enhancements. In addition, integrating principles of Explainable Artificial Intelligence (XAI) into the ANN model ensures that the predictive factors influencing customer satisfaction are transparent and easily interpretable, aiding in the formulation of targeted service improvements.

2. State of Art

2.1. Studies on Customer Satisfaction Evaluation

Research on employee satisfaction is more prevalent in the literature than studies on customer satisfaction, covering sectors such as healthcare [7,8,9,10], tourism [11,12], call centers [13,14], and banking [15,16]. Although customer satisfaction has been addressed in various studies, only the study by Kaynar et al. [17] has explored this in the context of water laboratories. In their study, the authors evaluated customer satisfaction levels in water laboratories accredited to TS EN ISO/IEC 17025 using a ten-question questionnaire with a five-point scale. The results showed an average score of 4.83 ± 0.07, equating to 96.6% satisfaction. The study concluded that customer satisfaction was higher for these accredited laboratories, leading to corrective actions to improve services [17].

Iswahyuningsih et al. [18] conducted a study examining the impact of service quality on customer satisfaction in an ISO/IEC 17025:2017 accredited calibration laboratory in West Sumatra, considering the moderating roles of organizational culture and price. The study found that service quality is crucial for customer satisfaction, which, in turn, can drive economic improvement. Data were collected from 93 respondents representing government agencies and private companies in West Sumatra through questionnaires. The data analysis revealed that service quality has the most significant impact on customer satisfaction, suggesting that organizations that enhance the quality of their services can significantly improve customer satisfaction. Additionally, organizational culture has a significant influence on customer satisfaction, with improvements in poor organizational culture leading to a major positive effect. The price of services also plays a significant role in customer satisfaction; transparent, reasonable pricing based on legal standards can build customer trust, further enhancing satisfaction. However, the study found that the moderating effects of organizational culture and price do not significantly alter the relationship between service quality and customer satisfaction.

Pillai et al. [19] discuss the importance of a laboratory quality management system for the effective operation of research, clinical, testing, or production/manufacturing laboratories. According to the authors, maintaining a robust laboratory quality management system is critical for managing a variety of laboratory activities, including basic and applied research, regulatory and clinical testing, and proficiency testing, as well as for ensuring records management and continuous improvement. This system provides a framework to identify and mitigate gaps and risks throughout the laboratory workflow, helping to prevent critical errors that could undermine the integrity and credibility of the institution. It ensures that results and data are reliable, accurate, timely, and reproducible. The authors explore different laboratory quality management system framework options suitable for various laboratory categories and discuss the necessary considerations for their implementation. They demonstrate that a laboratory quality management system that properly integrates all relevant requirements, promotes risk-based thinking, and addresses the entire laboratory workflow can ultimately lead to the generation of high-quality and relevant data. Such a system fosters greater confidence in the quality of data and results, encourages best practices, and cultivates a culture of responsibility and safety within the laboratory.

A study conducted by Magwaza and Pradhan [20] aimed to apply lean six sigma in a water testing laboratory to enhance efficiency, operate effectively in a lean environment, improve customer satisfaction, and reduce customer complaints. Lean six sigma offers a comprehensive approach to process improvement and is applied across various industries to boost operational efficiency and the quality of products or services. It is widely used in sectors such as manufacturing, services, healthcare, and finance, helping organizations optimize processes, reduce costs, and enhance customer satisfaction. The study sought to sustain and maintain accreditation, protect consumer health by preventing disease outbreaks, and facilitate continual improvement of the system. The findings reveal that brainstorming teams successfully identified bottlenecks and wastes contributing to customer dissatisfaction. These wastes were the root causes of the laboratory’s inefficiency and failure to meet service level agreements with customers. The application of lean six sigma tools was successful, as the accurate selection and implementation of appropriate tools led to noticeable improvements in laboratory performance.

Ningsih et al. [21] conducted a study aimed at identifying indicators and analyzing customer satisfaction performance at a local water company in Muara Enim Regency, South Sumatra Province, Indonesia. The evaluation of services was based on the timing of water distribution and the quality of the delivered clean water. Customer satisfaction in this study is defined as individuals’ feelings after comparing their experiences with their expectations. The study assessed customer satisfaction using criteria such as tangibles, reliability, responsiveness, assurance, and empathy. Three key strengths identified for improving customer satisfaction include the extensive water distribution network, high-quality staff service, and the experience of staff and technicians. These strengths can be leveraged to address potential weaknesses, such as low water discharge, poor water quality, infrequent streaming, and short duration of water flow. However, to ensure that the existing strategies are effective and meet the planned objectives, it is also crucial to focus on optimizing processes and minimizing threats.

Adams et al. [22] emphasize the workforce challenges in environmental laboratories, noting the necessity of effective management to lead laboratory staff and meet the stringent requirements of analyzing samples critical for public health and environmental protection. According to the authors, an environmental laboratory can utilize metrics such as customer satisfaction with treated water to monitor customer feedback over time. If the data indicate that customers are not entirely satisfied, the laboratory can review and refine its processes to enhance customer satisfaction.

The studies reviewed highlight the critical role of service quality, organizational culture, and effective management systems in enhancing customer satisfaction within laboratory settings, particularly those accredited to ISO/IEC 17025. Across various sectors, including water testing, calibration, and environmental laboratories, high service quality is consistently linked to greater customer satisfaction, which in turn drives operational efficiency and economic improvement. Lean six sigma and robust laboratory quality management systems emerge as effective tools for identifying inefficiencies, ensuring reliable results, and fostering a culture of continuous improvement. Overall, improving service quality, optimizing processes, and addressing customer feedback are pivotal strategies for laboratories to maintain accreditation, protect public health, and enhance customer satisfaction.

2.2. Applications of Artificial Neural Networks

Drawing parallels to biological learning, particularly in the human brain, ANNs seek to mimic human cognitive functions. This is accomplished through a network of interconnected units known as neurons. ANNs offer the advantage of addressing complex challenges efficiently by relying solely on examples, bypassing the necessity for a thorough understanding of the specific function or process involved [23,24]. The basic version of an ANN is the multilayer perceptron, featuring an input layer, one output layer, and possibly additional hidden layers, but does not include any internal recurrence. In contrast, recurrent neural networks are designed with cyclical connections and self-reinforcing elements. In the multilayer perceptron architecture, data flow forward through interconnected nodes, with each layer getting its input from the previous one. The sigmoid function is a popular activation function, modifying input features nonlinearly to produce the final outputs [23,24].

Backpropagation is the standard algorithm used to train multilayer perceptrons [23,24]. It involves two key phases: forward and backward propagation. In the forward phase, data flow through the hidden layers with random weights to produce the output. During the backward phase, the error, calculated as the difference between the actual and target outputs, is propagated backward through the network to adjust the weights. The backpropagation algorithm is controlled by two parameters: the momentum coefficient and the learning rate, with both parameters restricted to the range of 0 to 1. The momentum coefficient controls the direction of weight updates to maintain stability, while the learning rate controls the change applied to weights at each iteration [25].

Traditional ANNs, advanced methods that have evolved from them (e.g., deep learning, generative adversarial networks, or convolutional neural networks), and hybrid models integrating various computational intelligence techniques offer highly adaptable and flexible modeling capabilities applicable across many sectors. They find applications in financial markets for stock prediction [26,27], in medicine for disease identification and diagnosis [28,29], in industrial settings for enhancing production efficiency [30,31], in environmental assessment and monitoring [32,33,34], fault diagnosis [35,36], defect detection [37], optimization [38], and across various other domains.

3. Methodology

This section reviews the research’s architecture (Figure 1), encompassing the design, place of study, data collection methods, instruments applied, sample characteristics, and analytical techniques. It also considers the ethical aspects taken into account during this study.

3.1. Study Design

The objective of this study is to evaluate customer opinions regarding the quality of service provided by a water laboratory. Additionally, it aims to contribute to the improvement of service quality by identifying areas where customer feedback is less favorable. The research questions posed are as follows:

• What are the customer opinions regarding the quality of service provided by the water laboratory?
• In which areas do customers feel the service provided by the water laboratory could be improved?

The questionnaire included six areas (customer service, quality of service provided, support documentation, technical support, billing and payment, and online services and tools) to apply the methodology suggested by Fernandes et al. [39] for converting qualitative responses into quantitative data. The subsequent subsections provide further insights.

3.2. Place of the Study

The research was conducted at the Water Laboratory of Santiago do Cacém Municipality, situated in Alentejo Litoral, southern Portugal. Accredited under the ISO/IEC 17025 standard, this laboratory is recognized for its expertise in sampling and analytical testing. It provides comprehensive services, including collection, transport, and physicochemical and microbiological analysis of both treated and untreated water samples from sources like wells, boreholes, pools, and whirlpools, catering to both municipalities and private clients. The laboratory staff consists of 18 professionals, including a laboratory manager, a quality manager, an administrative worker, four sample technicians, five microbiology technicians, and six physicochemical technicians.

3.3. Data Collection

The selection of a survey utilizing questionnaires resulted from an in-depth analysis of different methodologies, with the choice strengthened by its simplicity and adaptability. Even though this approach may lack detailed context, it offers efficiency, standardization, and anonymity. Alternative methods like interviews, focus groups, observations, experiments, and case studies can yield more in-depth insights but are often resource-intensive, prone to bias, and limited in generalizability. Additionally, the structured format of the questionnaire simplifies the process of transforming participants’ qualitative responses into quantitative data [40,41,42,43].

To address the research questions and objectives defined earlier, a questionnaire was specifically designed. It is divided into three sections. The first section collects socio-demographic details like gender, age group, academic qualification, and employment status, enabling the analysis of how satisfaction levels vary across different socio-demographic groups. The second section includes various statements (

Table 1. Breakdown of the areas and statements featured in the second section of the questionnaire.

Customer service	S1	Staff friendlily and approachability.
	S2	Easiness of contact.
	S3	Response time to requests.
	S4	Service availability.
	S5	Service flexibility (to meet my needs).
	S6	Effectiveness in the clarification of questions.
	S7	Clarity in answers to questions.
Quality of service provided	S8	Explanation of the testing process.
	S9	Sampling.
	S10	Number and type of tests available.
	S11	Reliability of results.
	S12	Appropriateness of delivery times.
	S13	Delivery deadlines.
Support documentation	S14	Clarity in the presentation of the commercial proposals
	S15	Clarity in the presentation of analysis reports
	S16	User-friendliness of the format of analysis reports.
	S17	Helpfulness of guidelines provided for interpreting test results.
	S18	Inclusion of all relevant legal and regulatory information in the documentation.
Technical support	S19	Availability of technical support.
	S20	Expertise of the technical support team.
	S21	Solutions provided by technical support.
Billing and payment	S22	Clarity and transparency of billing information.
	S23	Ease of the payment process.
	S24	Flexibility of payment options offered.
Online services and tools	S25	Usability of the laboratory’s online portal.
	S26	Response time for appointment requests via email.
	S27	Effectiveness of the email scheduling process.

) focused on the study’s areas (customer service, quality of service provided, support documentation, technical support, billing and payment, and online services and tools). This encourages participants to provide feedback, addressing the study’s goal of identifying areas for improvement. The final section includes a question assessing participants’ opinions on the overall quality of service provided by the laboratory (Figure 2), guiding the customization of service improvements. Additionally, the data collected from the second and third sections will serve as the basis for developing a predictive model using ANNs. This model is designed to anticipate factors influencing customer satisfaction and provides a structured framework for future service evaluations and enhancements. The second and third sections of the questionnaire use a six-level Likert scale (i.e., very dissatisfied, dissatisfied, slightly dissatisfied, slightly satisfied, satisfied, and very satisfied) for participants to express their opinions. Additionally, participants were asked to indicate the trend of their responses, whether it was an upward trend (from very dissatisfied to very satisfied) or a downward trend (from very satisfied to very dissatisfied), as shown in Figure 2. The choice of a six-level Likert scale was because it offers more options for intermediate responses without a neutral central option.

Following the recommendations given by Bell [44], a team of specialists conducted a detailed evaluation of the questionnaire, recommending changes that were included in an updated version. The panel consisted of five individuals selected for their professional backgrounds in customer service management, quality assurance and quality management systems, analytical chemistry, consumer association representative, and social science researcher. Collectively, the expert team brought deep expertise in customer service management, quality assurance and management systems, analytical chemistry, and social science research methodologies. The updated questionnaire’s validity and clarity were assessed using a separate subset of participants not in the main group. Cronbach’s alpha was used to assess the reliability of the instrument, resulting in a score of 0.85 for the questions in the second and third parts of the questionnaire.

The revised questionnaire was distributed in hard copy to each participant personally. All 421 questionnaires were returned, achieving a 100% response rate. Data collection took place from January 2021 to September 2023.

3.4. Participants

Out of the 421 questionnaires handed out, 9 (2.1%) were excluded due to incomplete responses in the former section. Consequently, the study included 412 participants ranging from 22 to 79 years old, with an average age of 43.8 ± 22.3 years.

Table 2. Breakdown of participants by gender, age group, academic qualification, and employment status.

Socio-Demographic Characteristics	Class	Frequency
		N	%
Gender	Female	153	37.1
	Male	259	62.9
Age (years old)	<25	22	5.3
	[25, 45]	114	27.7
	[46, 65]	168	40.8
	>65	108	26.2
Academic qualifications	Basic education	153	37.1
	Secondary education	144	35.0
	High education	89	21.6
	Post-Graduate education	26	6.3
Employment status	Student	11	2.7
	Employed	183	44.4
	Unemployed	41	10.0
	Retired	177	42.9

presents the breakdown of participants by gender, age group, educational qualification, and employment status.

Online platforms, such as websites and social media, were the introduction point for 44.2% of the customers. Another 13.3% of customers found the laboratory by visiting it, 18.9% were repeat clients, 16.3% learned about it through friends or family, and 7.3% became aware of it through advertisements.

3.5. Qualitative Data Processing

Responses in the second and third sections of the questionnaire are recorded using a six-level Likert scale (i.e., (1) very dissatisfied, (2) dissatisfied, (3) slightly dissatisfied, (4) slightly satisfied, (5) satisfied, and (6) very satisfied). To reflect the trend of participants’ answers, the scale was expanded to include 11 levels:

(6) very satisfied, (5) satisfied, (4) slightly satisfied, (3) slightly dissatisfied, (2) dissatisfied, (1) very dissatisfied, (2) dissatisfied, (3) slightly dissatisfied, (4) slightly satisfied, (5) satisfied, and (6) very satisfied.

The expanded Likert scale can be interpreted either from left to middle, showing a transition from very satisfied (6) to very dissatisfied (1), or from middle to right, showing a transition from very dissatisfied (1) to very satisfied (6). The former interpretation indicates a move toward a more negative opinion, while the latter indicates a move toward a more positive opinion.

Using the approach outlined by Fernandes et al. [39], qualitative data were converted into quantitative data. This technique depicts the responses for each main topic within a circle of radius $1/\sqrt{\pi }$, segmented into k slices (where k corresponds to the number of statements for the topic under analysis), with each axis mark representing a response option. This approach facilitates drawing parallels between knowledge representation and thermodynamics, analogous to the degradation of the energy process. To demonstrate the core principles of the proposed methodology, the First and Second Laws of Thermodynamics are referenced, considering that a system evolves over time. The First Law, also called the Law of Energy Conservation, states that the total energy of an isolated system remains constant. The Second Law pertains to entropy, which quantifies the disorder within a system and its evolution. In this context, a data item is assumed to be in an entropic state, where its energy can be divided and utilized for degradation, but not for destruction, encompassing the following:

• Exergy (EX), also termed usable energy or available work, denotes the fraction of energy that can be expended by a system;
• Vagueness (VA), which pertains to energy values that may or may not have been expended; and
• Anergy (AN) signifies an energy potential that has not been consumed, making it available, which includes all energy, excluding exergy and vagueness [45].

This methodology will be thoroughly explained and applied to a specific case in Section 4.3.

3.6. Artificial Neural Networks

ANN-based models were constructed using WEKA 3-8-6 software, keeping the default settings (e.g., learning rate = 0.3, momentum = 0.2) [46,47]. The backpropagation algorithm and logistic activation function [23,24,25] were used in the learning stage. Each experiment was conducted 25 times to achieve statistically significant results. For each simulation, data were split into a training set (67%) and a test set (33%). The training set was used for model development, while the test set was employed for generalization assessment.

3.7. Ethical Aspects

The participants were aware of the questionnaire’s objectives and engaged in the study willingly, free from any undue influence. The research was carried out in accordance with applicable regulations and institutional protocols, receiving approval from the quality manager of the Water Laboratory of Santiago do Cacém Municipality. Additionally, the individuals provided informed consent to take part in the study.

4. Results and Discussion

This section outlines the outcomes of a study on customer satisfaction with the quality of service provided by a water laboratory based on responses from a sample of 412 customers.

4.1. Frequency of Responses Analysis

Figure 3 presents the percentage breakdown of responses to statements S1 to S7 (Table 1) regarding customer service. The analysis reveals a strong prevalence of the most positive responses (very satisfied and satisfied) for all statements, with percentages ranging from 71.4% (S7, on clarity in answer to question) to 95.2% (S2, on easiness of contact). The percentage of responses slightly satisfied varied between 2.9% (S2) and 12.4% (S5, on service flexibility). Negative responses (slightly dissatisfied, dissatisfied, and very dissatisfied) were minimal, ranging from 1.5% (S1, on staff friendlily and approachability) to 5.6% (S5). However, statements S6 (on effectiveness in clarification of questions) and S7 stood out with higher percentages (13.8% and 20.6%, respectively). Only in statement S5 did 0.5% of the participants not select any option. These findings indicate a highly positive opinion of the lab’s customer service among participants. Nevertheless, regarding clarity and effectiveness in responding to customer inquiries, some participants have expressed less favorable opinions, suggesting an area for the lab to improve in the future.

Figure 4 presents the percentage breakdown of responses to statements S8 to S13 (Table 1) regarding the quality of service provided. The analysis reveals a strong prevalence of the most positive responses (very satisfied and satisfied) for all statements, with percentages ranging from 69.6% (S11, on reliability of results) to 90.8% (S10, on number and type of tests available). The percentage of responses slightly satisfied varied between 6.8% (S10) and 15.8% (S8, on explanation of the testing process). Negative responses (slightly dissatisfied, dissatisfied, and very dissatisfied) ranged from 2.4% (S10) to 10.4% (S13, on delivery deadlines). However, statement S11 stood out with a higher percentage (20.7%). In statements S8 and S11, 2.4% and 1.7% of the participants did not select any option, respectively. These findings indicate a highly positive opinion of the quality of the lab’s services among participants. However, concerning the reliability of results, some participants have stated less favorable opinions, suggesting an aspect for the lab to address going forward.

Figure 5 presents the percentage breakdown of responses to statements S14 to S18 (Table 1) regarding support documentation. The analysis reveals a prevalence of the most positive responses (very satisfied and satisfied) for all statements, with percentages ranging from 50.5% (S17, on the helpfulness of test result interpretation guidelines) to 84.7% (S14, on clarity of presentation of the commercial proposal). The percentage of responses slightly satisfied varied between 2.4% (S14) and 12.4% (S17). Negative responses (slightly dissatisfied, dissatisfied, and very dissatisfied) ranged from 5.1% (S14 and S16) to 27.9% (S17), with statements S15, S17, and S18 standing out with percentages higher than 15%. Except for statement S15, some participants did not provide any response in all other statements, with S17 and S18 standing out at 3.9% and 7.5%, respectively. These findings indicate a positive opinion of the lab’s support documentation among participants. However, regarding the clarity of analysis report presentations, the helpfulness of test result interpretation guidelines, and the inclusion of legal information, some participants have expressed less favorable opinions, revealing areas for the lab to address in the future.

Figure 6 presents the percentage breakdown of responses to statements S19 to S21 (Table 1) regarding technical support. The analysis reveals a prevalence of the most positive responses (very satisfied and satisfied) for all statements, with percentages ranging from 68.5% (S20, on the expertise of the technical support team) to 71.1% (S21, on solutions provided by technical support). The percentage of responses slightly satisfied varied between 16.0% (S21) and 18.4% (S20). Negative responses (slightly dissatisfied, dissatisfied, and very dissatisfied) ranged from 7.8% (S20 and S21) to 8.2% (S19, on availability of technical support). In all statements within this area, some participants did not select any response, with this percentage ranging from 4.9% in S19 to 5.3% in S20. These findings indicate a positive opinion of the lab’s technical support among participants.

Figure 7 presents the percentage breakdown of responses to statements S22 to S24 (Table 1) regarding billing and payment. The analysis reveals a prevalence of the most positive responses (very satisfied and satisfied) for all statements, with percentages ranging from 66.8% (S24, on flexibility of payment options) to 75.7% (S23, on ease of payment process). The percentage of responses slightly satisfied varied between 8.3% (S23) and 17.2% (S24). Negative responses (slightly dissatisfied, dissatisfied, and very dissatisfied) ranged from 15.1% (S22, on clarity and transparence of billing information) to 16.0% (S23 and S24). These findings indicate a positive opinion on billing and payment options, although due to the percentage of negative responses, the lab should review these issues in the future.

Figure 8 presents the percentage breakdown of responses to statements S25 to S27 (Table 1) regarding online services and tools. The analysis reveals a prevalence of the most positive responses (very satisfied and satisfied) for all statements, with percentages ranging from 60.2% (S27, on the effectiveness of the email scheduling process) to 72.1% (S25, on the usability of the lab’s online portal). The percentage of responses slightly satisfied varied between 2.2% (S25) and 3.4% (S27). Negative responses (slightly dissatisfied, dissatisfied, and very dissatisfied) ranged from 14.3% (S25) to 15.3% (S26, response time on appointment requests via email). In all statements within this area, some participants did not select any response, with this percentage ranging from 21.4% in S25 and S26 to 21.6% in S27. These findings indicate a positive opinion of the lab’s online services and tools among participants. Nevertheless, given the percentage of negative responses and the number of participants who did not select any option, the laboratory should improve this type of service and actively promote its use by creating incentives for its adoption.

An overall analysis of these results underlines that the lab customers’ opinion regarding the quality of service provided is positive in all areas included in the study. Nevertheless, there was a percentage of less positive responses concerning certain aspects, namely clarity and effectiveness in responding to customer inquiries, the reliability of results, the clarity of analysis report presentations, the helpfulness of test result interpretation guidelines, the inclusion of legal information, billing and payment options, and the use of online services. These findings are consistent with the results of Kaynar et al. [17], which indicate that although customer satisfaction was generally high, there were areas with less positive responses that needed improvement. These areas include the pre-service information, the attitude and behavior of the staff, the knowledge of the staff, the delivery deadlines, and the pricing policy. Although in other sectors of activity, various studies related to customer satisfaction have concluded that it is crucial for customer loyalty and the success of companies, being highly influenced by the quality of the service provided [7,8,9,10,11,12,13,14,15,16,17,21,22].

In the last section of the questionnaire, participants were asked to express their overall opinion about the service provided by the lab and the progression trend of their response. Figure 9 presents the percentage breakdown of responses to the statement in the third section of the questionnaire, while Figure 10 shows the breakdown of responses according to the trend indicated by the participants. The analysis of Figure 9 indicates that the general opinion of customers regarding the service provided by the lab is predominantly positive. Specifically, the percentages of positive responses (very satisfied, satisfied, and slightly satisfied) were 17.0%, 64.8%, and 9.2%, respectively. Negative responses (slightly dissatisfied, dissatisfied, and very dissatisfied) were minimal, being chosen by only 4.4%, 2.7%, and 1.9% of participants, respectively. These findings are consistent with the results of Kaynar et al. [17], which indicate that all participants declared their willingness to recommend the laboratory, revealing a high degree of satisfaction with the services provided.

The analysis of Figure 10 shows that the majority of participants who selected the most positive/negative responses did not indicate any trend, revealing a consolidated opinion. The high percentage of customers indicating either extremely negative or extremely positive responses without specifying a future trend may reflect that the customer has a well-established opinion, such as being very dissatisfied (extremely negative) or very satisfied (extremely positive). In these cases, the customer might feel that any mention of future trends is unnecessary. Furthermore, the customer might believe that the trend is implicit in their response. If they are extremely satisfied, they might assume that the situation can only stay the same or worsen in the future. Conversely, if they are extremely dissatisfied, they might feel that the situation will either stay the same or improve. Therefore, they see no need to explicitly state a trend. Among participants who selected other options, the majority indicated an upward trend, ranging from 52.6% to 61.1%. The downward trend was indicated by a percentage of participants that ranged from 16.7% to 35.6%. These results suggest that customer opinion is generally very positive. Nevertheless, there is potential for improvement, particularly in the areas identified previously. By addressing these specific aspects, the laboratory could further enhance customer satisfaction. The fact that most participants showed an upward trend in their overall opinion may be related to the laboratory’s accreditation under the ISO/IEC 17025 standard. This accreditation ensures rigorous procedures for method validation, equipment calibration, and staff competency, which together reduce errors and improve the consistency of outcomes. Indeed, the efforts made by the laboratory to meet customers’ needs and expectations have proven effective. As shown in Figure 10, the majority of customers who expressed dissatisfaction or slight dissatisfaction have indicated a positive trend in the evolution of their opinions. This suggests that the laboratory’s proactive measures are successfully addressing concerns and improving overall customer satisfaction. These results are consistent with those of Kaynar et al. [17], who found that customer satisfaction with a water laboratory accredited under the same standard in Turkey was 96.6%. Additionally, studies in other sectors highlight the importance of accreditation and/or quality management systems in achieving high levels of customer satisfaction [18,19,20].

4.2. Effects of Socio-Demographic Variables on Overall Opinion of Customers

To analyze how socio-demographic factors influence customers’ general opinions regarding the service provided by the lab, the responses from the third section of the questionnaire were examined separately based on gender, age, academic qualifications, and employment status. Figure 11 illustrates the percentage breakdown of responses to the third section of the questionnaire, segmented by the socio-demographic variables mentioned above. With respect to the influence of gender on customers’ opinions, it is observed that the percentage breakdown of responses differed by less than 0.5%. When comparing response frequencies by age group, positive opinions differ by 4.0% (17.6% vs. 13.6%), 2.6% (65.7% vs. 63.5%), and 1.4% (10.2% vs. 8.8%) for very satisfied, satisfied, and slightly satisfied, respectively. Negative opinions show differences of 1.1% (4.8% vs. 3.7%), 2.7% (4.6% vs. 1.9%), and 3.7% (4.6% vs. 0.9%) for slightly dissatisfied, dissatisfied, and very dissatisfied, respectively. Additionally, there is a trend of more positive opinions from customers over 45 years old and more negative opinions from those under 45 years old. When comparing response frequencies by academic qualifications, positive opinions differ by 6.1% (17.7% vs. 11.6%), 5.8% (67.3% vs. 61.5%), and 2.1% (9.8% vs. 7.7%) for very satisfied, satisfied, and slightly satisfied, respectively. Negative opinions show differences of 5.1% (7.7% vs. 2.6%), 6.4% (7.7% vs. 1.3%), and 2.5% (3.8% vs. 1.3%) for slightly dissatisfied, dissatisfied, and very dissatisfied, respectively. Additionally, there is a trend of more positive opinions from customers with lower academic qualifications and more negative opinions from those with higher academic qualifications. Finally, regarding the impact of employment status on customers’ opinions, it is observed that the percentage breakdown of responses varied by less than 1.6% across all categories, apart from students. Given that this group represents only 2.7% of the total sample, it is not feasible to include them in this comparative analysis. However, it is noteworthy that all student respondents expressed positive opinions. The data analysis reveals differences in customers’ opinions about the service provided by the lab based on socio-demographic factors such as age and academic qualifications. These variations can be attributed to differences in expectations, evaluation standards, and individual experiences. The data indicate that certain age groups and levels of academic qualification have more critical opinions, reflecting the influence of factors such as life experience, educational demands, and professional expectations. For example, older customers and those with lower educational levels tend to express more positive opinions, possibly due to greater tolerance for minor service shortcomings and adjusted expectations. In contrast, younger customers and those with higher educational levels may have higher expectations and stricter standards, resulting in more critical evaluations. These findings align with Kotler et al. [2], who indicate that satisfaction depends on customers’ perceptions and expectations.

4.3. A Thermodynamic Approach to Data Processing

Figure 12 presents participant one’s responses to the second section of the questionnaire, along with the evolution trend of his/her responses. To transform qualitative data into quantitative and ensure the process is understandable, full details for support documentation area (statements S14 to S18) are given. Thus, considering that in statement S14 the participant selected very satisfied, this response should be assessed on a downward scale, i.e., from very satisfied to very dissatisfied, since the response can only remain the same or worsen in the future (

Table 3. Translating participant one’s responses across study areas (customer service, quality of service provided, support documentation, technical support, billing and payment, and online services and tools) to an expanded Likert scale.

Area	Statements	Expanded Likert Scale 11-Levels *
		Downward Trend						Upward Trend
		6	5	4	3	2	1		2	3	4	5	6	Vagueness
Customer service	S1	✕
	S2											✕	✕
	S3											✕	✕
	S4	✕
	S5									✕	✕
	S6			✕	✕
	S7			✕	✕
Quality of service provided	S8	✕
	S9											✕	✕
	S10											✕	✕
	S11													✕
	S12									✕	✕
	S13								✕	✕
Support documentation	S14	✕
	S15			✕	✕
	S16											✕	✕
	S17						✕
	S18													✕
Technical support	S19	✕
	S20											✕	✕
	S21											✕	✕
Billing and payment	S22											✕	✕
	S23											✕	✕
	S24			✕	✕
Online services and tools	S25													✕
	S26					✕	✕
	S27					✕	✕

* (1) very dissatisfied, (2) dissatisfied, (3) slightly dissatisfied, (4) slightly satisfied, (5) satisfied, and (6) very satisfied.

). Regarding statement S15, the participant selected slightly satisfied and indicated a downward trend (Figure 12). Therefore, this response should be assessed on a downward scale, considering slightly satisfied as the best-case scenario (BCS) and slightly dissatisfied as the worst-case scenario (WCS).

Concerning statement S16, the participant selected satisfied and indicated an upward trend (Figure 12). Therefore, this response should be assessed on an upward scale, i.e., from very dissatisfied to very satisfied, considering very satisfied as the BCS and satisfied as the WCS (Table 3). With respect to statement S17, the participant selected very dissatisfied. This response should be assessed on the upward scale because the response can only remain the same or improve in the future. Finally, for statement S18, no options were selected, indicating a vague situation, where all scenarios are possible. Here, even though the specific values of exergy, vagueness, and anergy remain unknown, it is known that their values lie between 0 and 1.

Figure 13 shows the visual depiction of participant one’s responses to the second section of the questionnaire under both BCS and WCS. Blue regions symbolize exergy, indicating high-energy states or usable energy; gray regions indicate vagueness, implying uncertainty or indeterminate energy states; and white regions represent anergy, where energy is not available for work [48,49,50].

Table 4. Calculating exergy, vagueness, and anergy for Support Documentation (SD–5) in the best-case scenario, for both scales, i.e., from very satisfied (6) to very dissatisfied (1), and from very dissatisfied (1) to very satisfied (6).

Statement	SD–5–Scale (6) → (1)	SD–5–Scale (1) → (6)
S14	${exergy}_{S_{14}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_0^{{\displaystyle \frac{1}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}={\displaystyle \frac{1}{5}}\pi \left({\left({\displaystyle \frac{1}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}\right)}^2-0\right)=0.006$	$-$
	${vagueness}_{S_{14}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{{\displaystyle \frac{1}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}^{{\displaystyle \frac{1}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}=0$	$-$
	${anergy}_{S_{14}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{{\displaystyle \frac{1}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}^{\sqrt{{\displaystyle \frac{1}{\pi }}}}=0.194$	$-$
S15	${exergy}_{S_{15}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_0^{{\displaystyle \frac{3}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}=0.05$	$-$
	${vagueness}_{S_{15}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{{\displaystyle \frac{3}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}^{{\displaystyle \frac{3}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}=0$	$-$
	${anergy}_{S_{15}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{{\displaystyle \frac{3}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}^{\sqrt{{\displaystyle \frac{1}{\pi }}}}=0.15$	$-$
S16	$-$	${exergy}_{S_{16}}={-\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{{\displaystyle \frac{1}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}^0=0.006$
	$-$	${vagueness}_{S_{16}}=-{\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{{\displaystyle \frac{1}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}^{{\displaystyle \frac{1}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}=0$
	$-$	${anergy}_{S_{16}}=-{\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{\sqrt{{\displaystyle \frac{1}{\pi }}}}^{{\displaystyle \frac{1}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}=0.194$
S17	$-$	${exergy}_{S_{17}}=-{\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{\sqrt{{\displaystyle \frac{1}{\pi }}}}^0=0.2$
	$-$	${vagueness}_{S_{17}}={-\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{\sqrt{{\displaystyle \frac{1}{\pi }}}}^{\sqrt{{\displaystyle \frac{1}{\pi }}}}=0$
	$-$	${anergy}_{S_{17}}=-{\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{\sqrt{{\displaystyle \frac{1}{\pi }}}}^{\sqrt{{\displaystyle \frac{1}{\pi }}}}=0$
S18	${exergy}_{S_{18}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_0^0=0$	$-$
	${vagueness}_{S_{18}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_0^0=0$	$-$
	${anergy}_{S_{18}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_0^{\sqrt{{\displaystyle \frac{1}{\pi }}}}=0.2$	$-$

and

Table 5. Calculating exergy, vagueness, and anergy for support documentation (SD–5) in the worst-case scenario for both scales, i.e., from very satisfied (6) to very dissatisfied (1), and from very dissatisfied (1) to very satisfied (6).

Statement	SD–5–Scale (6) → (1)	SD–5–Scale (1) → (6)
S14	${exergy}_{S_{14}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_0^{{\displaystyle \frac{1}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}={\displaystyle \frac{1}{5}}\pi \left({\left({\displaystyle \frac{1}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}\right)}^2-0\right)=0.006$	$-$
	${vagueness}_{S_{14}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{{\displaystyle \frac{1}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}^{{\displaystyle \frac{1}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}=0$	$-$
	${anergy}_{S_{14}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{{\displaystyle \frac{1}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}^{\sqrt{{\displaystyle \frac{1}{\pi }}}}=0.194$	$-$
S15	${exergy}_{S_{15}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_0^{{\displaystyle \frac{3}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}=0.05$	$-$
	${vagueness}_{S_{15}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{{\displaystyle \frac{3}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}^{{\displaystyle \frac{4}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}=0.039$	$-$
	${anergy}_{S_{15}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{{\displaystyle \frac{4}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}^{\sqrt{{\displaystyle \frac{1}{\pi }}}}=0.111$	$-$
S16	$-$	${exergy}_{S_{16}}={-\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{{\displaystyle \frac{1}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}^0=0.006$
	$-$	${vagueness}_{S_{16}}=-{\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{{\displaystyle \frac{2}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}^{{\displaystyle \frac{1}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}=0.016$
	$-$	${anergy}_{S_{16}}=-{\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{\sqrt{{\displaystyle \frac{1}{\pi }}}}^{{\displaystyle \frac{2}{6}}\sqrt{{\displaystyle \frac{1}{\pi }}}}=0.178$
S17	$-$	${exergy}_{S_{17}}=-{\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{\sqrt{{\displaystyle \frac{1}{\pi }}}}^0=0.2$
	$-$	${vagueness}_{S_{17}}={-\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{\sqrt{{\displaystyle \frac{1}{\pi }}}}^{\sqrt{{\displaystyle \frac{1}{\pi }}}}=0$
	$-$	${anergy}_{S_{17}}=-{\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{\sqrt{{\displaystyle \frac{1}{\pi }}}}^{\sqrt{{\displaystyle \frac{1}{\pi }}}}=0$
S18	${exergy}_{S_{18}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_0^0=0$	$-$
	${vagueness}_{S_{18}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_0^{\sqrt{{\displaystyle \frac{1}{\pi }}}}=0.2$	$-$
	${anergy}_{S_{18}}={\left.{\displaystyle \frac{1}{5}}\pi r^2\right]}_{\sqrt{{\displaystyle \frac{1}{\pi }}}}^{\sqrt{{\displaystyle \frac{1}{\pi }}}}=0$	$-$

summarize the evaluation of the regions depicted in Figure 13 for both the BCS and the WCS across both scales, i.e., from very satisfied to very dissatisfied, and vice versa.

By replicating the previous calculations, the values of exergy, vagueness, and anergy for the remaining areas (i.e., customer service, quality of service provided, technical support, billing and payment, and online services and tools) for all participants can be determined.

Table 6. Values of EXergy (EX), VAgueness (VA), and ANergy (AN) concerning participant one, for all areas included in the study (i.e., customer service (CS–7), quality of service provided (QSP–6), support documentation (SD–5 technical support (TS–3), billing and payment (BP–3), and online services and tools (OST–3)) in the best-case scenario for both scales, i.e., from very satisfied to very dissatisfied, and from very dissatisfied to very satisfied.

	Scale (6) (5) (4) (3) (2) (1) *				Scale (1) (2) (3) (4) (5) (6) *
	EX	VA	AN		EX	VA	AN
CS–76-1	0.080	0	0.492	CS–71-6	0.044	0	0.385
QSP–66-1	0.005	0	0.329	QSP–61-6	0.126	0	0.542
SD–56-1	0.056	0	0.544	SD–51-6	0.206	0	0.194
TS–36-1	0.009	0	0.324	TS–31-6	0.018	0	0.648
BP–36-1	0.083	0	0.250	BP–31-6	0.018	0	0.648
OST–36-1	0.462	0	0.538	OST–31-6	0	0	0

* (1) very dissatisfied, (2) dissatisfied, (3) slightly dissatisfied, (4) slightly satisfied, (5) satisfied, and (6) very satisfied.

displays these values for all areas considered in the study for participant one. For example, the exergy regarding the support documentation area in the BCS, on the scale from very satisfied to very dissatisfied, is determined using the values listed in Table 4, as follows:

$${exergy}_{6-1}={exergy}_{{6-1}_{S_{14}}}+{exergy}_{{6-1}_{S_{15}}}+{exergy}_{{6-1}_{S_{18}}}=0.006+0.05+0=0.056$$

For the scale from very dissatisfied to very satisfied, it is:

$${exergy}_{1-6}={exergy}_{{1-6}_{S_{16}}}+{exergy}_{{1-6}_{S_{17}}}=0.006+0.2=0.206$$

Similarly, using the same methodology, the values for different forms of energy were calculated for all participants in the WCS.

Table 7. Values of EXergy (EX), VAgueness (VA), and ANergy (AN) concerning participant one, for all areas included in the study (i.e., customer service (CS–7), quality of service provided (QSP–6), support documentation (SD–5 technical support (TS–3), billing and payment (BP–3), and online services and tools (OST–3)) in the worst-case scenario for both scales, i.e., from very satisfied to very dissatisfied, and from very dissatisfied to very satisfied.

	Scale (6) (5) (4) (3) (2) (1) *				Scale (1) (2) (3) (4) (5) (6) *
	EX	VA	AN		EX	VA	AN
CS–76-1	0.080	0.056	0.436	CS–71-6	0.044	0.052	0.333
QSP–66-1	0.005	0.167	0.162	QSP–61-6	0.126	0.103	0.439
SD–56-1	0.056	0.239	0.305	SD–51-6	0.206	0.016	0.178
TS–36-1	0.009	0.000	0.324	TS–31-6	0.018	0.056	0.592
BP–36-1	0.083	0.065	0.185	BP–31-6	0.018	0.056	0.592
OST–36-1	0.462	0.538	0	OST–31-6	0	0	0

* (1) very dissatisfied, (2) dissatisfied, (3) slightly dissatisfied, (4) slightly satisfied, (5) satisfied, and (6) very satisfied.

displays these values for all areas considered in the study for participant one.

The values of different forms of energy associated with each of the six specific areas considered computed for all participants in both BCS and WCS were used to build a database for training ANNs to predict customers’ overall opinions regarding the lab’s service (Figure 14). The input variables were calculated based on data gathered in the second section of the questionnaire, taking into account the participants’ responses and the trends ticked. The outputs (i.e., overall customer satisfaction considering the best-case scenario and overall customer satisfaction considering the worst-case scenario) were derived from data collected in the third section of the questionnaire. The effectiveness of the ANN model was evaluated using the confusion matrix displayed in

Table 8. Confusion matrix of the ANN model for predicting customers’ overall opinions on the lab’s services in both scenarios.

		Predict	Training						Test
	Target		Very dissatisfied	Dissatisfied	Slightly dissatisfied	Slightly satisfied	Satisfied	Very satisfied	Very dissatisfied	Dissatisfied	Slightly dissatisfied	Slightly satisfied	Satisfied	Very satisfied
Best-case scenario	Very dissatisfied		4	0	0	0	0	0	2	0	0	0	0	0
	Dissatisfied		0	4	1	0	0	0	0	2	0	0	0	0
	Slightly dissatisfied		0	1	7	1	0	0	0	0	3	1	0	0
	Slightly satisfied		0	0	1	17	2	0	0	0	1	7	1	0
	Satisfied		0	0	0	7	75	8	0	0	0	4	36	5
	Very satisfied		0	0	0	0	14	133	0	0	0	0	7	68
Worst-case scenario	Very dissatisfied		7	1	0	0	0	0	3	0	0	0	0	0
	Dissatisfied		0	6	1	0	0	0	0	3	1	0	0	0
	Slightly dissatisfied		0	1	11	2	0	0	0	0	7	1	0	0
	Slightly satisfied		0	0	4	77	3	0	0	0	3	36	3	0
	Satisfied		0	0	0	9	105	7	0	0	0	4	52	4
	Very satisfied		0	0	0	0	3	38	0	0	0	0	2	18

.

The accuracy of the ANN model in the BCS is 87.3% (240 out of 275 instances correctly identified) and 86.1% (118 out of 137 instances correctly identified) for the training and test sets, respectively. In the WCS, these metrics are 88.7% (244 out of 275 instances correctly identified) and 86.9% (119 out of 137 instances correctly identified) for the training and test sets, respectively. Thus, the ANN model demonstrates good effectiveness in predicting customers’ overall opinions regarding the lab’s service, attaining accuracy levels close to 90%. This model provides insights into customer satisfaction, identifies areas for service improvement, and helps prioritize actions based on simulated scenarios. By predicting multiple scenarios, the model offers several advantages for the water lab, including proactive management through forecasting various potential conditions and challenges and enhancing the customer experience by anticipating needs and addressing concerns more effectively. Developed based on historical data, including customer feedback and performance metrics, the model predicts how changes in factors such as service quality or response times impact customer satisfaction. It identifies key variables with the most significant impact, highlights performance gaps, and helps prioritize actions based on their potential effects. Additionally, the model forecasts potential challenges and guides resource allocation, ensuring proactive management. With continuous updates and refinement, the ANN model supports ongoing improvement by adapting to evolving customer needs and preferences.

5. Conclusions

This study evaluates customer opinions regarding the quality of service provided by a water laboratory in six areas, i.e., customer service, quality of service provided, support documentation, technical support, billing and payment, and online services and tools. The findings underline that the lab customers’ opinions on the quality of service are positive in all areas included in the study. Nevertheless, there was a percentage of less positive responses concerning certain aspects, namely clarity and effectiveness in responding to customer inquiries, the reliability of results, the clarity of analysis report presentations, the helpfulness of test result interpretation guidelines, the inclusion of legal information, billing and payment options, and the use of online services. Moreover, the results show that the overall customer opinion is generally very positive. However, there is potential for improvement, particularly in the areas identified above. By addressing these specific aspects, the laboratory could further enhance customer satisfaction. The findings also reveal differences in customers’ opinions based on socio-demographic such as age and academic qualifications. The data indicate that certain factors, such as age groups and levels of academic qualification, have more critical opinions, reflecting factors such as life experience, educational demands, and professional expectations. For example, older customers and those with lower educational levels tend to express more positive opinions, possibly due to greater tolerance for minor service shortcomings and adjusted expectations. In contrast, younger customers and those with higher educational levels may have higher expectations and stricter standards, resulting in more critical evaluations. These trends highlight the importance of considering demographic characteristics when evaluating customer satisfaction and personalizing services to better meet the needs and expectations of different groups. Understanding these variations enables a more targeted and effective approach to continuously improving the service offered by the lab. Additionally, a predictive model grounded on artificial neural networks was developed to forecast customers’ overall opinions about the lab’s service. This model proved effective, with accuracies approaching 90%, enabling proactive management by anticipating various potential conditions and scenarios. The model proposed is a customizable framework that adapts to the specific needs of any organization. However, it had limitations, such as a regionally confined sample, specifically the customers of the Water Laboratory of Santiago do Cacém Municipality. Expanding the sample size to include customers from other water laboratories across the country would provide more representative insights and enable decisions to be made with national relevance. Additionally, the model can be enhanced by integrating the feedback from other stakeholders, such as staff and suppliers. To ensure the model’s effectiveness in different organizational contexts, data collection instruments may need to be modified, with study areas adjusted according to the type of organization or activity.