Constructing a Decision Model for Health Club Members to Purchase Coaching Programs during the COVID-19 Epidemic

Abstract

The recent COVID-19 epidemic has affected the global sports industry to a certain extent, and health clubs are no exception. To avoid unsustainable operations, health clubs need to restructure their programs to suit members’ needs. Therefore, this study constructs a two-stage framework model to evaluate health club members’ purchase of coaching programs. The first stage is to construct a hierarchy of evaluation, using the modified Delphi method, to select suitable criteria and extended sub-criteria, and add and delete them through expert discussion. In the second stage, we use the pairwise comparison matrix to calculate the weight of each criterion and sub-criterion to influence each other. Next, we evaluate and compare physical, online and offline, and live-stream coaching programs, by using network hierarchy analysis to identify the best class purchase plan during the epidemic and provide relevant suggestions. The results of the study found that during the epidemic, the primary sales were for weight training among physical programs (0.314), and activity classes among online and offline programs (0.633) as well as live-stream coaching programs (0.280). These findings have implications for health clubs in deciding which mode they need to adopt for sustainable operations.

1. Introduction

The COVID-19 pandemic disrupted normal life to a great extent. This included an impact on global health, and educational, financial, commercial institutions, physical sports venues, fitness centers, sports events, broadcasts, etc., which were partially suspended [1]. Callaghan et al. [2] surveyed that the global wellness market exceeds US$1.5 trillion, with an annual growth rate of 5% to 10%. However, the uncertainty on the global growth forecast by the International Monetary Fund (IMF) in 2020 is expected to decline by 3%, and by 6.1% for advanced economies and cause huge economic losses [3,4]. In the early stages of the spread of COVID-19, Taiwan implemented its pandemic prevention policy, thereby keeping it under control within the region. In the long-awaited new season, the Chinese Professional Baseball League kicked off at the home stadium of the Rakuten Monkeys (Lamigo) in Taoyuan, becoming the world’s first professional baseball league to restart after the pandemic, and is attracting global attention [5].

Thompson [6] presented that the COVID-19 pandemic, which has transformed the global fitness market, caused health and fitness clubs to shut down, or at best, restructure their services. The American College of Sports Medicine ACSM [7] has announced the top 10 fitness trends, and ‘Exercise is Medicine’ has been on the list for three years, meaning that exercise is better than a cure in preventing disease during epidemics. In 2021, the No. 1 ranking in fitness trends was obtained by Online Training, while wearable technology was ranked first for two consecutive years in 2019 and 2020. Since the western world is not so mature in the concept of epidemic prevention, resulting in the rapid spread of the epidemic. The spread of COVID-19 is exacerbated by close contact, hence, gym personal training programs are a high-risk transmission channel. Therefore, from the perspective of epidemic prevention, the industry initiated the trend of online home training programs, regardless of their effectiveness.

During the COVID-19 pandemic, gyms in many countries or regions have been shut down, and instead, live-stream fitness classes by fitness trainers have become popular on social media [8]. In order to keep gyms, health clubs and fitness centers sustainable, this research proposed a two-stage framework for evaluating members’ decisions to purchase or subscribe to coaching programs. The first stage constructs a hierarchy of evaluation. The second stage calculates the weight between criteria and sub-criteria. The Analytic Network Process (ANP) is employed to evaluate and compare the most suitable solution. Finally, the feasibility of the evaluation model was assessed and verified by experienced trainers from World Gym, Genghis Khan, Igus, and other health club chains in central Taiwan.

The remainder of this article is structured as follows. Section 2 describes the Materials and Methods in three sub-sections. The third section, i.e., Results, discusses the findings of the study. Thereafter, the fourth section discuss the results and conclude the study, respectively.

2. Materials and Methods

The Materials and Methods have been described in three subsections: Section 2.1 on the “criteria for evaluating members to purchase classes”, Section 2.2 on “ANP analysis method”, and Section 2.3 on “model construction and research process”.

2.1. Criteria for Purchase Methods

2.1.1. Purchase Criteria in Analytic Hierarchy Process

Many experts and scholars have proposed the use of AHP [9] to construct purchasing models for various products and services, such as mobile phones [10], computer managed maintenance system software [11], enterprise resource planning systems [12], medical information systems [13,14], expert knowledge-based systems [15], residential property [16], two-wheelers post pandemic [17], fuel cell hydrogen vehicles purchase policy [18], and warships [19], as well as for evaluating an outsourcing logistics company [20] and other applications in different industries. The above-mentioned literature, based on AHP, combined fuzzy theory and modified Delphi method to construct a procurement model. In terms of evaluation criteria, benefits, opportunities, costs, and risks (BOCR), and strength, weakness, opportunity, threat (SWOT) are used as evaluation criteria for procurement evaluation, and expert opinions are extended to formulate some criteria and sub-criteria. Amendments are made, or criteria, such as quality, time, management, and technology, are added, to construct an expert procurement model.

2.1.2. Analytic Network Process and Decision-Making Methods

The AHP hierarchical structure is linear, wherein, the criteria, sub-criteria, and alternatives are structured hierarchically with a linear and unidirectional relationship between levels. ANP is a non-linear structure, in which the criteria, sub-criteria, and alternatives are better adapted to the complexity of the real world [21,22]. Kheybari et al. [23] reviewed a total of 465 research papers from 2000 to 2017, using ANP alone, or combined with different research methods. After induction and sorting, the articles in the field of business and financial management are categorized into the nine key application areas. Peng and Lin [24] used the modified Delphi method to analyze the elements of bicycle brand equity, and identified four criteria: perceived quality, brand loyalty, brand awareness, and brand association, and finally used the ANP method to assign the highest weight to brand loyalty and, accordingly, suggest business strategies to dealers. Lee et al. [25] used ANP to develop three criteria, i.e., product quality, brand awareness, and price, to explore the factors that consumers consider when choosing sports equipment, among which price is the most important. Wang and Tzeng [26] noted that consumers would consider the quality, functions, and the design of the brand when purchasing, and proposed Decision-making Trial and Evaluation Laboratory (DEMATEL) with ANP and VIKOR methods to evaluate brand value. Sánchez-Garrido et al. [27] combined the ELECTRE method with ANP to evaluate different building alternatives. Ozdemir and Basligil [28] combined Fuzzy and ANP to develop a FANP model to evaluate Turkish Airlines, concluding that its future operations should include purchasing, renting or chartering aircraft. Ayağ and Özdemİr [29] proposed fuzzy ANP to evaluate ERP software alternatives. There are seven evaluation criteria for purchasing an ERP system: cost, vendor support, flexibility, functionality, reliability, ease of use, and technology advance. Li et al. [30] explored how to increase college students’ satisfaction with the Chinese professional baseball audience, in which they also used ANP and Quality Function Deployment (QFD) to analyze three levels, labeling them in order: reasons for participating in baseball, exciting baseball action, and modification of game rules to attract the audience to the game.

2.1.3. Evaluation of Sport Criteria in Analytic Network Process

Saaty and Vargas [31] presented ANP and BOCR to use in economic, political, social, and technological assessment. Chen et al. [32] applied FANP and BOCR to build an online to offline service model for sporting goods retailers. Lee et al. [33] applied the ANP method, combined with the four main criteria of BOCR, in the supply and demand study of electronic manufacturers and suppliers, and found the ranking order of supply and demand performance from it. Liao and Chen [34] used ANP and BOCR as a decision tool to identify customers and suppliers to collaborate on product design to enhance corporate competitiveness in a case study of a highway system in Taiwan. Montesinos-Valera et al. [35] used ANP and BOCR to evaluate the maintenance, renewal, and improvement projects in Spanish Rail railway network project. Huang and Wey [36] presented ANP with BOCR and demographics, facility use and social welfare indicators, to help the Taipei City Government construct a land reuse strategy for middle and elementary schools. The evaluation strategies included: closing, integrating, or downsizing, and assisting the Taipei City Government to promote the policy of reusing the remaining school land. Memari et al. [37], utilizing the AHP and ANP models, as well as inputs from 11 experts aware of football and sports industry issues, propose six criteria, i.e., human resources index, managerial expertise, marketing, software, legal and economic infrastructure and 33 sub-criteria, to assess the operational capability of football clubs.

Khan and Ali [38] sorted out the papers of scholars using AHP and ANP in different fields from 2001 to 2019. The hierarchical structure of AHP does not have feedback from low to high levels in linear top-down interactions, while the ANP is composed of many elements and groups of elements that are interconnected at the same time [39]. To make our research more comprehensive, we applied the modified Delphi method to let the industry, government, and academia jointly formulate the evaluation framework and criteria for the sustainable operation of gyms during the epidemic, and then used the feedback from the ANP criteria to evaluate the sales ranking of different fitness programs.

2.2. ANP Method

Each expert uses the consistency ratio (subjectively determined by each of them) to check the consistency of the network relationship between each level of interaction. The calculation process is as follows [21,22,40].

Step 1:

Collect completed questionnaires from experts

Step 2:

Establish a pairwise matrix for comparison (1 to 9 importance scale, 1 being equally important and 9 being one is extremely important) Saaty [21]

Step 3:

Check consistency

Saaty [21] proposed the Consistency Index (CI) to measure the consistency of two pairwise comparative matrices, and the formula for the CI is as follows (1)

$$CI=\frac{{\lambda }_{max}-n}{n-1}$$

(1)

Consistency Ratio (CR) is used to measure the consistency formula, as follows (2).

$$CR=\frac{CI}{RI}$$

(2)

where RI is Random Index (RI).

When CR ≤ 0.1, it is judged that the two pairs of matrix A are consistent, or the inconsistency is acceptable; otherwise, the two pairs of matrix A are adjusted until the quality is consistent [41].

Step 4:

Form super matrix [42]

(1)

Unweighted super matrix (unweighted super matrix) consists of extremely original comparison weights. The formula is as follows (3).

$$\begin{array}{cc}& \begin{array}{ccc}& & \begin{array}{c} C_1 C_2 \dots C_{4 }\\ \begin{array}{c}S{C}_{11}S{C}_{12}\dots S{C}_{1n1} S{C}_{21}S{C}_{22}\dots S{C}_{2n2} S{C}_{i1}S{C}_{i2}\dots S{C}_{ini}\end{array}\\ \end{array}\end{array}\\ =\begin{array}{ccc}\begin{array}{c}\\ C_1\\ \\ \begin{array}{c}\\ \end{array}\\ C_2\\ \\ \\ ⋮\\ \\ C_4\\ \end{array}& \begin{array}{c}\begin{array}{c}\begin{array}{c}\begin{array}{c}\begin{array}{c}\begin{array}{c}\\ \\ S{C}_{11}\end{array}\\ S{C}_{12}\\ ⋮\end{array}\\ \begin{array}{c}S{C}_{1n1}\end{array}\\ S{C}_{21}\end{array}\\ \begin{array}{c}S{C}_{22}\\ ⋮\end{array}\\ S{C}_{2n2}\end{array}\\ \begin{array}{c}⋮\\ \end{array}\\ S{C}_{i1}\end{array}\\ S{C}_{i2}\\ ⋮\\ S{C}_{ini}\end{array}& \end{array}& \begin{array}{ccc}& & \begin{array}{c}\\ \\ \left[\begin{array}{c}\begin{array}{c}\begin{array}{c}\begin{array}{c}\begin{array}{c}\begin{array}{c}\begin{array}{cccc}& \begin{array}{c}\\ w_{11}\\ \begin{array}{c}\\ \\ \\ \begin{array}{c}{w}_{21}\\ \\ \\ \\ ⋮\\ \\ \\ \\ w_{i1}\end{array}\end{array}\end{array}& & \end{array}\begin{array}{c}\\ w_{12}\\ \begin{array}{c}\\ \\ \\ \begin{array}{c}{w}_{22}\\ \\ \\ \\ ⋮\\ \\ \\ \\ w_{i2}\end{array}\end{array}\end{array}\begin{array}{ccc}& & \end{array}\begin{array}{c}\\ \cdots \\ \begin{array}{c}\\ \\ \\ \begin{array}{c}\dots \\ \\ \\ \\ \ddots \\ \\ \\ \\ \dots \end{array}\end{array}\end{array}\begin{array}{cccc}& & \begin{array}{c}\\ w_{1i}\\ \begin{array}{c}\\ \\ \\ \begin{array}{c}{w}_{2i}\\ \\ \\ \\ ⋮\\ \\ \\ \\ w_{ii}\end{array}\end{array}\end{array}& \end{array}\end{array}\end{array}\end{array}\end{array}\end{array}\end{array}\right]\end{array}\end{array}\end{array}$$

(3)

Weighted super matrix (weighted super matrix) multiplies the corresponding weights using the same criteria as in the unweighted matrix. The formula is as follows (4).

$$w_n^{\prime }=\left[ \begin{array}{c} 0 0 0\\ w21 w22 0\\ 0 w32 w33\end{array}\right]$$

(4)

(2)

Limit super matrix (limit super matrix) multiplies the weighted matrix many times until each column has the same number. The formula is as follows (5).

$$w_{ANP}=\underset{\mathrm{g}\to \infty }{\lim }{\left(w_n^{\prime }\right)}^{2g+1}$$

(5)

Step 5:

Evaluate the best solution

(1)

Calculate the weights of each alternative under each criterion: Calculate the weight of the elements at each level, and then calculate the weight of all levels.

(2)

Select the best option: Based on the weights of each option, the most suitable option for the final goal is decided.

(3)

Construct an evaluation model for member purchase.

2.3. Model Construction and Research Process

A two-stage framework was employed for evaluating members’ purchase of coaching programs. In the first stage, the evaluation structure was constructed; the experts decided the evaluation structure and criteria by using the modified Delphi method; in the second stage, the weights of each criterion and sub-criterion were calculated; in the third stage, the final evaluation plan was formulated and compared by using the network hierarchy analysis method. Thereafter, the experts provided the experience of application and practice. Next, the ANP analysis was applied to assess and compare physical, online and offline, and live-stream coaching programs. The research process is shown in Figure 1, and the model is constructed as follows.

Stage 1:

Constructing the hierarchy of evaluation

This study used the modified Delphi method proposed by Murry and Hammons [43] to collect literature on the use of the modified Delphi method, and BOCR and ANP in evaluation decisions, and summarized them into a structured questionnaire for purchasing coaching programs [44,45]. There are five steps in this phase.

Step 1.1:

Defining the evaluation criteria

The experts used the Likert Scale to assess the importance of each criterion and sub-criterion (5-very important, 1-very unimportant) and completed the questionnaire.

Step 1.2:

Selecting expert panel members

Murry and Hammons [40] suggest that the appropriate number of experts for the Delphi method should be more than 10 and less than 30, primarily because a group of experts larger than 30 is likely to have different opinions, making it difficult to obtain valid conclusions.

Step 1.3:

Compiling expert interview questionnaire

In this study, we first considered the willingness and feasibility of each expert to complete the questionnaire during the epidemic. We used online electronic questionnaires and Line communication software to assist in understanding the experts’ requests.

Step 1.4:

Detecting the consistency of experts’ statistical opinions

Faherty [46] proposed a quartile statistical method to detect experts’ opinions on questions with a quartile difference of less than 0.60 to achieve consensus among experts; if the overall quartile difference is more than 1.00, then there is no consensus among experts. If the average importance of the topic is 4 to 5 and the total percentage of all experts is more than 75%, then the topic has reached consensus among the experts. If the standard deviation is used to assess the dispersion of opinions among experts, and if the importance of each question is between 4 and 5 and the percentage of experts’ scores is below 75%, the question will be revised or deleted. If the average of all experts’ opinions was 3.5 points and 75% of the total experts’ scores were above that, then the overall opinion was consistent with the index. Finally, the criteria for purchasing coaching programs for club members during the epidemic were confirmed.

Step 1.5:

Deciding the ANP model, criteria, and sub-criteria

After the experts reached a certain consensus on the structure of the questionnaire, the third revised Delphi questionnaire was forwarded to the experts who had implemented the revised Delphi method twice earlier. This time the experts mainly decided on the interdependence of the criteria and sub-criteria and determined the ANP evaluation framework for purchasing coaching programs.

Stage 2:

Comparing physical, online and offline, and live-stream coaching programs

Step 2.1:

Create pairwise comparison matrix between criteria: to evaluate the criteria and sub-criteria of the coaching programs purchased by members of the fitness industry during the epidemic, and to compare them within the same level of the nominal scale.

Step 2.2:

Calculate the eigenvalues and eigenvector values: After standardizing the feature vector corresponding to the maximum feature value (λmax), it is the relative weight of the criteria and sub-criteria for purchasing coaching program evaluation.

Step 2.3:

Check for consistency

The Consistency Index (CI) of Equations (1) and (2) is used to measure the consistency of the decision maker’s pre and post judgment.

Saaty [21] proposed the ANP, which has the ability to improve the analytical hierarchy process (AHP), because the AHP requires strict requirements for inverse contrast, independence, predictability, and homogeneity among matters. Saaty [21,22] incorporates Benefits, Opportunities, Costs, and Risks (BOCR) into the key ANP criteria.

During the epidemic, the Ministry of Health and Welfare of Taiwan [47] promulgated the following policies for disease prevention among the public: restriction on large-scale events (100 people indoors and 500 people outdoors), with implementation of the real-name system, regular disinfection and security checks, measurement of body temperature, wearing of masks throughout the whole process, keeping a social distance (1.5 m indoors and 1 m outdoors), prohibition of food and drink except for water, and imposition of penalties on those who repeatedly disobeyed the advice. In the third level of alert, new regulations were issued for gyms: 25% of the maximum number of people in the gym, suspension of services (of swimming pool, shower room, sauna, steam room, oven, social hall), coaching restrictions (avoiding body contact, wearing goggles or masks throughout, fast screening once every 7 days, no sports massage or stretching classes), prohibition of team competitions, disinfection of equipment before and after use, and social distance. The distance between the equipment and the team should be disinfected and socialized.

Saaty [21,22] analyzed the ANP as a web-like network structure between all criteria and sub-criteria, wherein the criteria are interdependent and provide feedback to each other. The first level is the research topic, the second level consists of the main criterion, and the third level comprises the sub-criteria, extended from the main criterion. The sub-criteria not only have certain interdependence with the main criterion, but also with each other within the group; they even affect the sub-criteria of other groups, as a network would.

Step 2.4:

Calculate the weight of each criterion and sub-criterion

At the third level of the sub-criteria evaluation system, we multiply the weight of the third level criteria downward by the weight of the second level primary criteria downward by the weight of the third level criteria to calculate the overall criteria weight.

Step 2.5:

Select the best program purchase plan

After determining the weight of all levels, we will arrange the weight of each plan one by one to calculate the best coaching program plan for members to purchase.

3. Results

In the early days of the outbreak in Taiwan, the Taiwanese government had appropriate anti-epidemic policies. The epidemic was well controlled, and people’s daily lives were not impacted much [48]. However, indoor activities were not allowed, such as: school classes, gyms, catering, movies, indoor singing, etc., which resulted in altering people’s consumption habits, from physical to online consumption. Consequently, many brick-and-mortar stores have closed [49]. Even in the post-epidemic era, people find it difficult to revert to their old ways. Gyms are amongst the establishments that were most affected by the epidemic. For the sake of sustainable operations and survival, some gyms developed online programs during the most severe period of the epidemic in Taiwan, when it was declared to have entered the third level of alert. In the post-epidemic era, with gradual reopening, people can return to physical gyms. However, there are still many regulations, one of which is to wear a mask while exercising and operating the gym equipment. Most consumers may feel uncomfortable or troubled with this regulation, and choose to return to online programs.

Therefore, this study takes a gym in Taiwan as an example. In response to the epidemic situation, the industry, government, and academia jointly designed a program for sustainable operation. The program includes: Weight Training (WT), Functional Training (FT), Stretching and Relaxation (SR), and Activity Classes (AC).

Table 1. Four alternative coaching programs.

Alternative	Coaching Program Features
Weight Training (WT)	1. Fixed equipment (fork type) 2. Semi-fixed equipment (hanging bar type) 3. Free weights (barbells or dumbbells)
Functional Training (FT)	1. Total Body Resistance Exercise 2. Boxing 3. Kettlebell
Stretching and Relaxation (SR)	1. Stretching (muscle/fascia) 2. Shiatsu
Activity Classes (AC)	One event every six months, the programs have 1. Weight loss competition 2. Kettlebell 3. Muscular energy circulation

summarizes the features of the four alternative programs. In terms of revenue and profits, apart from membership fees, the health club mainly relies on instructor programs as a source of income, while a substantial amount of money is invested in new trainer training and on-the-job education training, and each new trainer is required to obtain a company-approved and international license within six months, as well as the company’s internal assessment license.

This study proposes a two-stage, sustainable management curriculum evaluation model. In the first stage, using the modified Delphi method, the industry, government, and academia should be invited to decide the evaluation criteria, sub-criteria and ANP evaluation framework during the epidemic. In the second stage, ANP is used to invite gym coaches to evaluate four types of programs in three modes: physical, online and offline, and live-stream coaching programs. Finally, the differences of these four programs in these three different teaching modes are compared. The evaluation process is as follows.

Stage 1:

Establishment of evaluation structured hierarchy

Step 1.1:

Establish structure criteria and sub-criteria using the Delphi method

Saaty’s [21] BOCR model with ANP method is used to examine various political, social, and technical aspects of the application of the BOCR model. This study uses the BOCR and the market factor suggested by experts to be included as the evaluation items for criteria to investigate the member purchase programs, which include weight training (Genghis Khan only programs), stretching and relaxation (muscle stretching, massage), functional training (boxing; kettlebell; vitality, performance, reconditioning; total body resistance exercise, etc.), and activity programs (video teaching, weight loss, occasional competition, etc.) corresponding to coaching programs [46], as shown in

Table 2. Definitions and literature of criteria and sub-criteria.

Main Criteria	Definition	References	Sub-Criteria	References
Benefit (B)	Benefit gained from the decision. (Immediacy)	[21,31,32,33,34,35,36,37]	Large chain big box (B1) Bargaining space (B2) Contract (B3)	[50,51]
Opportunity (O)	Potential benefits to be gained from the decisions made. (Futuristic)	[21,31,32,33,34,35,36,37]	Marketing approach(O1) Coaching system and Coaching Salary System (O2) Professional image (O3)	[51]
Cost (C)	After the decision, the trouble caused from it. (Immediacy)	[21,31,32,33,34,35,36,37]	Service (C1) Professional competence (C2) Motion correction (C3)	[1,51]
Risk (R)	The potential distress caused by the decision. (Futuristic)	[21,31,32,33,34,35,36,37]	Actual present situation (R1) Negative media coverage (R2) Personnel changes (R3)	[51]
Market (M)	With the change of society, the concept and habit of consumers will change, and the supply and demand will change accordingly.	[51]	Influence of epidemic (M1) Peer competition (M2) Government policies (M3) Consumer perception (M4)	[51]

.

Step 1.2:

Expert Panel Source

In this study, industrial experts were defined as those who had been personal trainers at World Fitness, Genghis Khan, and Igus chain fitness clubs during, and prior to the pandemic, with a total cumulative coaching experience of more than three years or in a supervisory position. Additionally, there were five experts, two with an academic (background (Professors of National Taiwan University of Sport), and three from the Taichung City Government.

Table 3. Background information on experts.

Experts	Personal Qualifications	Numbers
Industrial	Three members with minimum of 3 years of total coaching experience (between epidemic and non-epidemic) Two members who previously held the position of Head of Coaching	5
Academic	Two members from academic institutions (from the National Taiwan University of Sport)	2
Government	Three members of the Sports Bureau, Taichung City Government	3
Total		10

displays the distribution of the 10 available experts from the industrial, academic, and governmental sectors.

Step 1.3:

Conducting expert questionnaire interviews

If time permitted, the questionnaires were completed immediately, and the rest were collected by completing remedial education.

Step 1.4:

Set statistical criteria for consistency of expert opinion

Faherty [46] proposed the use of quartiles, and the quartile difference indicates a certain level of agreement among experts, with a high correlation of less than or equal to 0.60 and a moderate correlation of 0.60 to 1.00. At the same time, the Likert scale was completed with the average of its importance greater than 3.50, highly and moderately relevant questions, otherwise, the question was deleted. Figure 2 presents the results of ANP structure.

Stage 2:

Calculation of criteria and sub-criteria weights

According to steps 2.1–2.4, we collected the entire expert questionnaires, compared them two by two to form a comparison matrix, and used Super Decisions ANP version 2.10.0 to calculate the weights between the primary and secondary criteria respectively.

Table 4. Matrix and weights of criteria for relative comparison.

Goal	B	O	C	R	M	Weights
B	1	1.319	1.528	3.010	1.646	0.298
O		1	1.159	2.283	1.249	0.226
C			1	1.970	1.077	0.195
R				1	1.828	0.099
M					1	0.181

displays the results of the relative comparison of online and offline coaching programs criteria.

Table 5. Relative matrix and weights of sub-criteria.

Benefit (B)	B1	B2	B3	Weights
B1	1	2.842	2.912	0.590
B2		1	1.025	0.207
B3			1	0.203

presents the benefit sub-criteria under the main criteria as an example to calculate the weight of the relative comparison. The five criteria and sub-criteria were compared with eigenvectors with CI ≤ 0.1 and CR ≤ 0.1, and are completely consistent [9].

Table 6. Weighting of the four programs under the sub-criteria.

	B1	B2	B3	O1	O2	O3	C1	C2	C3	R1	R2	R3	M1	M2	M3	M4
WT	0.289	0.153	0.151	0.16	0.245	0.341	0.214	0.284	0.256	0.198	0.406	0.345	0.317	0.377	0.344	0.263
FT	0.324	0.172	0.201	0.245	0.104	0.226	0.329	0.113	0.182	0.256	0.141	0.161	0.150	0.218	0.152	0.155
SR	0.113	0.286	0.288	0.226	0.001	0.203	0.247	0.34	0.383	0.309	0.252	0.245	0.208	0.155	0.213	0.207
AC	0.274	0.389	0.361	0.369	0.65	0.23	0.213	0.263	0.179	0.237	0.201	0.249	0.325	0.250	0.291	0.375

shows the weights of the four programs of weight training, myofascial relaxation, functional training, and activity classes under the sub-criteria. Therefore, CI ≤ 0.1 and CR ≤ 0.1 shows the experts’ judgment as completely consistent.

Combining Table 5 and Table 6, the advantage value S of the four coaching programs is calculated as in Equation (6).

$$\begin{array}{cc}\mathrm{S} =& \\ \left[\begin{array}{c} 0.289 0.153 0.151 0.160 0.245 0.341 0.214 0.284 0.256 0.198 0.406 0.345 0.317 0.377 0.344 0.263 \\ 0.324 0.172 0.201 0.245 0.104 0.226 0.329 0.113 0.182 0.256 0.141 0.161 0.150 0.218 0.152 0.155\\ 0.113 0.286 0.288 0.223 0.001 0.203 0.247 0.340 0.383 0.309 0.252 0.245 0.208 0.213 0.213 0.207\\ 0.274 0.389 0.361 0.369 0.650 0.230 0.213 0.263 0.179 0.237 0.201 0.249 0.325 0.291 0.291 0.375\end{array}\right]& \left[\begin{array}{c}0.590\\ 0.207\\ 0.203\\ 0.626\\ 0.374\\ 0.398\\ 0.308\\ 0.294\\ 0.344\\ 0.403\\ 0.253\\ 0.193\\ 0.224\\ 0.129\\ 0.424\end{array}\right]\end{array}$$

(6)

From Step 2.5, under online and offline, the purchased coaching programs during the epidemic were: weight training (0.268), functional training (0.098), stretching and relaxation (0.001), and activity classes (0.633). Repeat all steps of Phase 2 to calculate the weights of the four physical and online courses respectively.

Table 7. The weight of the four fitness coach courses under the physical, online and offline and live-stream mode.

	Programs	WT	FT	SR	AC
Type
Physical		0.314	0.208	0.261	0.218
Online and Offline		0.268	0.098	0.001	0.633
Live-stream		0.273	0.202	0.245	0.280

shows the results.

According to the analysis results in Table 7, in the entity mode, the coach’s favorite course to sell was weight training. In the online and offline mode, the most popular courses sold by coaches were activity classes. However, the weights of functional training and stretching and relaxation were almost 0, and coaches do not sell these two courses. In the live-stream mode, the member’s favorite course activity classes demonstrate the same results as the online and offline modes.

The following interpretations can be drawn from the above results: (1) The restriction of wearing a mask in the gym for class or exercise lowers the willingness to go to the gym for physical classes and exercise due to feelings of shortness of breath and discomfort. The industry develops online or offline courses, and reduces prices to allow members to change their exercise and class habits during the epidemic. (2) If online and offline courses are considered, members will feel inconvenience if some courses are taught online and some courses are taught in physical gyms. From the results, it can be seen that the two courses of functional training, and stretching and relaxation require a high degree of gym equipment or more than two people. Therefore, members tend to buy activity classes. (3) From Table 1, from the course definition of activity classes, it is not necessary to go to the physical class, and members can receive the desired fitness knowledge or content from the online class. Therefore, in the online, offline and live-stream mode (as in Table 7), the coach mainly sells activity classes, which is easier for members to accept.

Based on the above results, coaches are selling fitness courses. Unless the pandemic becomes severe, they will mainly sell physical courses, and the sales will focus primarily on equipment that is used in gyms. In addition to this, members may purchase new fitness courses when they resume going to the physical gym. In online courses, because the camera cannot completely capture the movements of the coaches or the fitness postures and movements of the members, the coaches cannot immediately adjust the movements or postures when the members need assistance. Therefore, the sale of online courses is based on the exercises that members can manage by themselves, without the support of a trainer.

4. Conclusions

We discussed the results with 10 experts to understand the best buying plans for members. Drawing from that, we derive the main contributions of this study, which are:

(1)

During the epidemic COVID-19, the global sports and gym markets faced operational problems. In the early stages of the epidemic, due to insufficient vaccines, Taiwan adopted the second- and third-level alerts to effectively control the epidemic. However, physical fitness-related industries were regulated and prohibited from operating. For sustainable operation, gyms, and fitness centers developed online courses. While not as remunerative as before the epidemic, this idea can prevent gyms or health clubs from shutting down. This study provides a global reference based on the experience of sustainable operation of gyms in Taiwan.

(2)

In the post-epidemic era, the vaccination rate in Taiwan was 92.42% [52]. However, masks are still required for indoor sports. Though gym operators developed online and offline fitness courses, the market did not respond well, except for activity classes (0.633). The lukewarm market response, and the weight of the courses of functional training, and stretching and relaxation being almost 0, resulted in coaches not being able to sell these two courses. This experience provides an important reference for designing online and offline fitness courses in the future.

(3)

The original gym operator designed online and offline fitness courses. The physical course part was designed in the gym. The main course design concept is to hope that members will return to the gym and use the equipment in the fitness room. However, the results from the analysis were not as expected.

For future research, the online and offline courses can be set in the members’ homes in the physical courses, and the operating equipment can be changed to elastic ropes, bottles, small dumbbells, and even household items that are readily available at home. Innovative activities or fitness courses and more favorable prices can meet the sports needs of consumers and create sustainable business for gym operators.