Hotel service quality dimensions and attributes: An analysis of online hotel customer reviews

Abstract

This study examines service quality dimensions and attributes of the hotel industry in a famous beach resort destination of Phuket based on 25,267 online reviews from the TripAdvisor website collected for 56 hotels. Machine learning analysis using the KNIME analytics platform was employed to analyze four datasets, namely the total dataset, the couple dataset, the family dataset, and the friend dataset. The results reveal six dimensions of guest service quality in the hotel industry: leisure activities, tangibles and surroundings, reliability, responsiveness, service process, and food, with specific attributes identified in each dimension. The study was able to verify the robustness of HOLSERV Plus model as the dimensions developed by topic modelling of online reviews are found to correspond to the dimensions of HOLSERV framework, with some adaptation required. It is also confirmed by the current study that the same set of service quality dimensions and attributes is not applicable to all groups of customers, instead each group has its own unique requirements and expectations. In addition, service process is revealed in this study as the most sensitive dimension that determines customer dissatisfaction.

Keywords

Online reviews hotels service quality dimensions Phuket

Introduction

Service quality management is the key to successful hotel operations. Understanding factors that determine service quality and customer satisfaction with the visit experience enables hotels to effectively direct efforts and focus on relevant factors in order to enhance service quality, satisfaction, and customer loyalty (Padlee et al., 2019; Sharma and Srivastava, 2018). Although existing hotel service quality literature has traditionally used quantitative methods (e.g. Ali et al., 2021; Ahmad et al., 2019; Padlee et al., 2019), research using online reviews or unsolicited data is growing (e.g. Alrawadieh and Law, 2019; Padma and Ahn, 2020). This hotel service quality trend is imperative, as online reviews are vital to the customers’ decision making, and the customers are increasingly reliant on online reviews and e-word-of-mouth when selecting a hotel to stay (Yang et al., 2017).

Although previous hotel online review analysis studies have provided an extent of conception of quality factors that determine guest satisfaction, there are notable discrepancies in the literature that need to be addressed. Firstly, many hotel online review studies rely on traditional qualitative methods of thematic/content analysis (Dinçer and Alrawadieh, 2017; Kim et al., 2016; Padma and Ahn, 2020). These studies are generally limited to relatively fewer reviews and hence, their results may not permit generalization. Secondly, they tend to analyze the data using an induction approach with a lack of theoretical foundation in their studies. Thirdly, most online review studies analyze aggregate data of all types of customers and base conclusions regarding the service quality on hotel customers in general (Chittiprolu et al., 2021; Peres and Paladini, 2021). Even if these studies offer useful insights into service quality dimensions and attributes, it is still uncertain whether the conclusions from these studies can be applied to specific groups of customers who may have different service behaviors and requirements during their hotel visits. The lack of insights into service quality determinants of specific groups of customers who have distinct needs, characteristics, and perceptions, requires academic attention as this knowledge is imperative to hotel quality improvement (Aksu et al., 2021; Dabestani et al., 2016).

This current study attempts to address the above research gaps in a number of ways. Firstly, it aims to examine underlying hotel service quality dimensions and attributes specifically applied to a beach resort destination, guided by existing service quality frameworks in the literature and using online reviews of hotel customers from the TripAdvisor platform. Specifically, the study seeks to offer a comparison between the dimensions of HOLSERV, which is a well-established and reliable hotel service quality instrument that has been developed based on the SERVQUAL scale specifically for hotel service quality measurement (Feiz et al., 2011; Mei et al., 1999), and the topic modelling results. Secondly, it attempts to offer a comparative analysis of service quality dimensions and attributes between different types of tourists. The quality dimensions developed in this study are associated with the three largest groups of customers, namely the family segment, the couple segment, and the friend segment. This study offers insights into how these different segments view and assess hotel service quality differently.

In addition, this research provides a number of academic contributions. The first contribution is in an effort to close the methodological gaps. The current study uses a larger number of online review comments from a variety of sampled hotels, which offers more generalizability to the findings. In addition, a novel method of text mining analysis using machine learning implemented in the KNIME analytics platform is applied, in order to effectively manage the analysis of big data used in the study. A second research discrepancy is addressed in this study by employing a theoretical foundation, namely by relating the topic modeling analysis to HOLSERV. Furthermore, this study adds to the body of knowledge by providing a comparative analysis of service quality attributes based on the reviews offered both by customers in general and by three different subgroups that are unique in their service requirements. In particular, the study contributes to the service quality body of knowledge by advancing understanding in service quality from the lens of online reviews, and argues that general service quality dimensions and attributes are not applicable across all specific groups of customers.

Online reviews

The importance of online reviews or electronic word-of-mouth to hospitality marketers and mangers has been widely acknowledged by scholars (Alrawadieh and Law, 2019; Fang et al., 2016; Luo et al., 2021; Xie et al., 2016). This is due to the fact that online review platforms such as TripAdvisor offer a domain where customers can reflect their service evaluation by generating their own content with reviews, ratings and photos of their own experiences. Hence, positive and negative experiences are shared, and they are constantly and widely accessed by fellow consumers worldwide (Chittiprolu et al., 2021). Furthermore, information that is helpful to hotel customers in making purchasing decision including review content, review sentiment, and reviewer characteristics can be retrieved from online reviews (Lee et al., 2018).

In addition, electronic word-of-mouth has become an important source for customers to base their decisions on, due to the intangibility and high-risk nature of hospitality service consumption (Alrawadieh and Law, 2019). As pre-testing and pre-evaluating services before consumption are infeasible, hotel purchase decisions rely on a complex process (Chittiprolu et al., 2021; Yang et al., 2017). Moreover, taking a holiday abroad is a high-cost activity, requiring both financial commitment and time investment in the trip and its planning process (Promsivapallop and Kannaovakun, 2017). Furthermore, travel expenses are often incurred prior to the actual trip, and are difficult to recover if the expectations are not met. Tourists are therefore faced with financial risk, particularly when making travel arrangements to unfamiliar destinations abroad (Promsivapallop and Kannaovakun, 2018). Therefore, tourists opt to utilize electronic word-of-mouth to assist them in such decision making including the choice of an accommodation to stay at (Alrawadieh and Law, 2019; Chittiprolu et al., 2021). Furthermore, online reviews offer great help to customers in the decision process by reducing confusion and time in the selection process (Tan et al., 2018) and provide more influence on customers’ trust from word-of-mouth communication than direct communications from hotel companies (Chittiprolu et al., 2021). Due to these reasons, scholars have demonstrated the importance of online reviews, in their direct impact on the popularity of hospitality businesses such as restaurants (Tian et al., 2021; Gan et al., 2017) and hotels (Ban et al., 2019; Peres and Paladini, 2021).

Hotel service quality dimensions

Understanding the quality of service is crucial to hotel businesses as it helps hotel managers to improve service and business performance (Ali et al., 2021). Pullman et al. (2005) argue that analyzing guests’ feedback is the best way to understand their satisfaction and preferences of service offerings. According to Parasuraman et al. (1985), customer satisfaction and service quality can be evaluated as the gap between expected service and perceived service for a customer, and the most popular service quality measurement tool called SERVQUAL is based on this. The SERVQUAL model consists of five dimensions that determine service quality, representing tangibles, reliability, responsiveness, assurance, and empathy. Based on Boon et al. (2014), the five SERVQUAL dimensions can be explained as follows:

• Tangibles: Appearance of personnel, physical facilities, and equipment.

• Reliability: Ability to deliver service consistently and accurately as promised

• Responsiveness: Willingness to provide prompt service and assistance

• Assurance: Ability to instigate trust and confidence through knowledge, competence and courtesy of employees

• Empathy: Caring for customers and providing personalized attention

Scholars have modified SERVQUAL to derive adjusted service quality models to fit specific service operations contexts. LODGSERV (Knutson et al., 1990), Lodging Quality Index (Getty and Getty, 2003), and HOLSERV (Mei et al., 1999) have been developed based on the SERVQUAL scale specifically for hotel service quality measurement. More recently, Lai and Hitchcock’s (2016) study on hotel service quality developed hotel service quality dimensions and attributes based on SERVQUAL, using survey data of hotel customers in Macao. The modified SERVQUAL dimensions were established to include Basics (Tangibles), Reliability, Assurance, Environment (giving the feelings of comfort, relaxation, and being welcome), Technology (in-room and hotel technologies), and Entertainment (recreation and entertainment facilities/activities). According to an online review study by Boon et al. (2014), HOLSERV was further extended to include five dimensions based on the online review results. It was named HOLSERV Plus and included the following dimensions and attributes:

• Room: Equipment, fixtures and fittings in the hotel room, services available in the room, cleanliness and user-friendliness

• Facilities: Facilities and services available in the hotel including as examples breakfast, restaurants and bars, pool, and fitness/spa facilities

• Surroundings: Location of the hotel, proximity to amenities, public transport, and attractions

• Employees: Appearance and behavior of staff, promptness, politeness, understanding, and neatness

• Reliability: The willingness of staff to help guests in specific situations and handle requests and complaints

Another approach to measure service quality was developed by Cronin and Taylor (1992), called SERVPERF. This is a performance-based model of customer assessment based on perceived performance of service. Thus, an advantage of the SERVPERF instrument is that it does not require measuring the expectations of the customer, but measures only the service performance. Similar to service quality evaluation, customer satisfaction is defined as the overall evaluation of service performance based on previous experiences (Luo and Qu, 2016) and on prior expectations of the performance. Moreover, these evaluations are based on the expectancy disconfirmation theory (Luo and Qu, 2016; Swan and Comb, 1976).

In addition, hotel service quality studies using online reviews have recently emerged in the literature. These studies have been conducted to examine service quality dimensions of hotels in different parts of the world such as Malaysia (Padma and Ahn, 2020), India (Chittiprolu et al., 2021), Brazil (Peres and Paladini, 2021), USA (Kim et al., 2016), China (He et al., 2017), and Vietnam (Thu, 2020). These studies used TripAdvisor as the data source platform as this platform has been claimed by researchers to provide richer and more reliable reviews than the other platforms (Chittiprolu et al., 2021; Taecharungroj and Mathayomchan, 2019). Although these studies did not specifically analyze the data based on any service quality framework, the results of these studies are generally in line with the five dimensions of HOLSERV Plus model of Mei et al. (1999), which are provided in Table 1. For example, Chittiprolu et al. (2021) identified room, food, physical signifiers (including ambience, architect, style, décor, etc.), staff behavior and professionalism, hotel amenities, service, and traditional services as key service quality dimensions of heritage hotels. Similarly, Padma and Ahn (2020) found hotel related attributes, room related attributes, staff related attributes, and travel related attributes to be key to hotel customer experience in their study of online reviews of hotels in Malaysia. These factors relate directly to room, facilities, employees and reliability of the HOLSERV model. In addition, hotel location has been discovered in the majority of these studies (such as Peres and Paladini, 2021; Suvachart, 2021; Thu, 2020, He et al., 2017; Kim et al., 2016) as a key service quality component. This dimension is connected with the surroundings dimension of the HOLSERV model.

Table 1.

Hotel service quality literature using online review analysis.

Authors	Data source	Location	Hotel service quality dimensions/attributes
Chittiprolu et al. (2021)	TripAdvisor	India	Room, food, physical signifiers (including ambience, architect, style, décor, etc.), staff behavior and professionalism, hotel amenities, service, traditional services
Peres and Paladini (2021)	TripAdvisor	Florianópolis-SC, Brazil	Location, room, staff, parking, ambience, reservation, cost-benefit, breakfast
Suvachart (2021)	Booking.com	Various countries	Accommodation amenities, location, service, servicescapes, budget, general satisfaction
Padma and Ahn (2020)	TripAdvisor	Malaysia	Hotel related attributes, room related attributes, staff related attributes, travel related attributes,
Padma and Ahn (2020)	TripAdvisor	Malaysia	Possible consequences
Thu (2020)	TripAdvisor	Vietnam	Location, room, services, staff, meal, value
Ban et al. (2019)	Google.com	Various countries	Access, food and beverage, purpose, tangibles, empathy
He et al. (2017)	TripAdvisor	China	Food, location, rooms, service, staff, lounge, check-in
Kim et al. (2016)	TripAdvisor	New York city, USA	Location, staff and their attitude, room size, service, breakfast, overall room, bed, view, neighborhood, hotel design/style
Zhou et al. (2014)	Agoda	Hangzhou, China	Physical setting (room, hotel, food), value, staff, location
Dong et al. (2014)	Daodao.com	Sanya, China	Hotel, location, room, service, dining and food, value, facility availabilities
Li et al. (2013)	Daodao.com	Beijing, China	Transportation convenience, food and beverage management, convenience to tourist destinations, value for money, bed, reception services, room size and decoration

The service quality dimensions above have been identified in the literature as key factors influencing customer satisfaction and dissatisfaction, and subsequently customer behavioral intention (Padlee et al., 2019). As argued by Chittiprolu et al. (2021) and Berezina et al. (2016), hotel customers tend to be more satisfied with intangible features of the hotel such as courtesy of staff, but more dissatisfied with tangible features of the hotel such as room and furnishing. In other words, dissatisfied customers generally complain about tangible and financial issues. Furthermore, it was further argued that service dimensions that make customers satisfied, could also contribute to dissatisfaction if they are not offered or there are problems in the delivery. In addition, the results of the study conducted by Padlee et al. (2019) offer additional insights showing that food quality, employee behavior, and room amenities are three key quality dimensions that influence hotel customer satisfaction, but customers did not find physical evidence a satisfaction determinant. Moreover, the relationships between service quality dimensions and hotel customer satisfaction were proven to be moderated by the hotel star rating by Nunkoo et al. (2020). In other words, this service quality and customer satisfaction relationship is found to change by the hotel rating category. While infrastructure and employee expertise appear to be important in determining guest satisfaction in lower rated hotels, safety and security as well as room quality play important roles in shaping three-star hotel customer satisfaction. On the other hand, waiting time and customer interaction are significant satisfaction determinants in high-end hotels. Therefore, the study by Nunkoo et al. (2020) confirms that the influence of service quality on customer satisfaction of a hotel may vary by its star rating.

Types of hotel customers and service quality perceptions

Similar to the results offered by Nunkoo et al. (2020) above, where the influence of service quality on customer satisfaction was found to be conditional upon the hotel differences, also the types of customers can play a role in shaping service their quality perceptions (Aksu et al., 2021; Tsaur et al., 2005). As argued by Dabestani et al. (2016), generalizing hotel customer service quality perceptions based on hotel customers in general should be made only tentatively, as hotel customers in the industry differ widely in their personality traits, needs, expectations, and perceptions. In addition, Tsaur et al. (2005) confirm differences in service quality perceptions among hotel customers from different cultural groups. For example, hotel customers from English heritage cultures were found to perceive higher levels of service quality than Asian and European groups in many aspects. These differences were also reflected in loyalty levels. The study also discovered that while strong loyalty intention tends to be from less masculine cultures, strong switch attitudes were more likely to be associated with more masculine cultural backgrounds. Moreover, Aksu et al. (2021) verify the differences in eco-service quality perceptions among two hotel customer segments, namely sensitive customers, and apathetic customers, who differ in socio-demographics and travel-choices.

In addition, based on an online hotel review study by Wang et al. (2020), it is evident to differentiate hotel selection criterion preferences and quality requirements among 5 groups of travellers including business, couples, families, friends and solo. The results of the study suggest that business travellers tend to determine hotel service quality based on guestrooms, lobby, and restaurants as they need to meet their business guests in the lobby and restaurants and work in their rooms. On the other hand, as couples tend to celebrate their special days and enjoy life, their hotel service quality is influenced more by bars and surroundings. Furthermore, while families tend to focus on breakfast and suites to meet their family needs, friend travellers put emphasis on cleanliness and location as they tend to visit multiple places for leisure. Moreover, as solo travellers tend to be young, they are expected to pay more attention to prices and location in their service quality requirements. Therefore, based on the existing literature, service quality perceptions and evaluation tend to be contingent upon hotel customer groups such as the friend segment, the family segment, and the couple segment in the current study, who differ in personality, needs and expectations.

Methods

Data collection

TripAdvisor was selected due to its popularity, wide use in previous studies, and data reliability. The platform has been reported by scholars such as Chittiprolu et al. (2021) and Taecharungroj and Mathayomchan (2019) to have systems to avoid false reviews, giving higher quality data for analysis when compared to other platforms. This contributes to reliability of data and results in the study.

Phuket was chosen as the location targeted by the study due to its rich online review data on TripAdvisor available for the study. As it is a world class tourist beach destination that attracted over 14 million tourists annually prior to the COVID-19 pandemic (C9 Hotelswork, 2019), the large scale of tourism invited different market segments and characteristics of hotel customers. According to the TripAdvisor website, the island has over 4,000 hotels in various service styles and categories on offer to its diverse guests. The extensive online review dataset on these hotels offers valuable comments and discussion on hotel service quality, enabling this study to fully investigate hotel service quality dimensions and attributes.

Online reviews of hotels in Phuket from TripAdvisor during 2018–2021 were identified as the population for the study. Only online review comments within a 4-year period from the data collection date were included in the study, in order to ensure that the study includes mainly current issues pertaining to hotel service quality. Mid-scale and upscale hotels were included in the study as they offer a variety of services and activities. Fifty-six (56) 3–5 star rated hotels on TripAdvisor were identified as the hotels for the study using selection criteria: being rated 3–5 stars in TripAdvisor, located at a beach, and having more than 1,000 online reviews in total. In the screening results, 24 hotels of 5-star luxury category, 5 hotels of 5-star midrange category, 15 hotels of 4-star category, and 11 hotels of 3-star category met these criteria and were included in the study. In total, the number of online reviews collected during 2018–2019 was 22,116 reviews, while the number of reviews collected during the pandemic period was only 3151. The reviews during the pandemic period were examined and initially analyzed. No specific COVID-19 related dimensions were identified from the initial analysis. Therefore, they were included into the datasets for further analysis together with all the data prior to the pandemic.

Only original online English language review comments without translation were included in the study to ensure the precision of the meaning of the comments. The review data were extracted for each hotel using a robotic process automation program, called KOFAX. Only reviews in the years 2018–2021 were extracted for the study. After screening for reviews only in English and cleaning of the data by removing reviews that showed only review scores without comments, there remained 25,267 reviews for further analysis, comprising 10,410 comments in 2018, 11,706 comments in 2019, and 2749 comments in 2020, and 402 comments in 2021. Then, items were created by converting each review to a string.

Data preparation

The open-source program KNIME analytics Platform 4.4.0 was utilized to prepare the data. This analytics platform has been widely used for text mining and sentiment analyses in previous online research on hospitality and tourism, such as the studies by Taecharungroj and Mathayomchan’s (2019), Chittiprolu et al. (2021), and Kalnaovakul and Promsivapallop (2021). The data preparation procedures developed by these authors were adopted to guide this research. First, all 25,267 reviews were converted from strings to documents, resulting in 25,267 documents for analysis. Punctuation in the documents was removed. Second, part of speech (POS) tagger and lemmatization used for tokenizing and converting each word to its root form (e.g. converting “is,” “am,” “are” to “be” or “studies” to “study”) were applied. Third, filters such as number (words representing numbers), N Char (reviews that have fewer than three characters) and case converter (converting all upper-case letters to lower case letters) were applied. The fourth step was to filter stop words (such as a, an, the, and by) with MySQL FullText features. Fifth, custom words (e.g. hotel names, and employee names) were removed. The final step involves getting all reviews stemmed with the use of Porter stemmer algorithm, which changes each word to its root form (e.g. conversion of “swimming” to “swim” or “drinking” to “drink” or “kids” to “kid” or “children” to “child”). Typos in the comments were considered during preprocessing and were found to be minor and of no concern to the analysis.

After the data were prepared in the previous steps, the document vector was produced by generating a Bag of Words from all documents. Term Frequency (TF: count frequency of each word in each document) and Inverse Document Frequency (IDF: total number of documents divided by the number of documents that the word appears in) were used to filter out unimportant common words and retain only meaningful words. Relatively less frequent words (with a threshold of 350 times of occurrence or around 1% of the most frequent word) in all documents were considered less significant and were removed from the data. TF/IDF technique is used to help filter out less significant words (low frequency words found in each document). This will allow only significant or meaningful words to be chosen for prediction (by LDA algorithm) for each dimension. Top 10 frequent words (or terms) represent each dimension. This practice helps improve the accuracy of principal component analysis (PCA), cluster analysis, and predictions made by machine learning (Kalnaovakul and Promsivapallop, 2021).

The total dataset was further separated into three sub-datasets based on the customer type: the couple group, the family group, and the friend group. This gave 4 datasets for analysis, namely (1) the total dataset (25,267 reviews), (2) the couple dataset (6,029 reviews), (3) the family dataset (5,665 reviews), and (4) the friend dataset (2,343 reviews).

Data analysis methods

Latent Dirichlet allocation (LDA)

In order to identify dimensionality of the data, PCA with K-means clustering, using the sum of squared errors to identify the number of dimensions, was used. The number of clusters or dimensions was determined by the elbow method, and factors of each dimension were extracted by using parallel latent Dirichlet allocation (LDA). Latent Dirichlet allocation is an efficient algorithm in determining the hidden structure of dimensions, in big data analysis of online reviews (Taecharungroj and Mathayomchan, 2019). Each dimension containing 10 terms was named based on an overall meaning of its terms. The analysis firstly looked to find the meaning of perceived service quality based on dimensions of HOLSERV Plus model. The naming of each dimension involved an initiation by the first researcher and confirmation by the second researcher.

Salience-valence analysis

To understand the sentiment of each dimension, whether it is positive or negative in the eyes of guest reviews, the number of reviews counted in each dimension was computed against the total number of reviews to derive salience of the dimension. In other words, the salience of a dimension is the proportion of number of reviews in each dimension divided by the total number of reviews. The high salience of a dimension indicates a large proportion of reviews on that specific dimension as compared the total number of reviews as a whole. The calculation is as follows

D i m e n s i o n S a l i e n c e = (N u m b e r o f R e v i e w s / T o t a l N u m b e r o f R e v i e w s)

The valence of a dimension is the difference between the observed and expected numbers of positive reviews in the dimension divided by the total number of reviews in the same dimension. In other words, while dimensional salience designates the proportion of each dimension, valence of the dimension denotes the extent of positivity or negativity of the dimension. A negative dimension valence implies negative sentiment in the dimension. It can be computed as follows

D i m e n s i o n V a l e n c e = (O b s e r v e d N o o f P o s i t i v e R e v i e w s - E x p e c t e d N o o f P o s i t i v e R e v i e w s) / (T o t a l r e v i e w s o f t h e d i m e n s i o n)

The same procedures were applied to each data set. The positive category was appointed to the reviews rated 4 and 5, whereas the reviews rated 1 to 3 were classified as negative. These procedures were adopted from Taecharungroj and Mathayomchan’s (2019) and Kalnaovakul and Promsivapallop (2021). Following the procedures suggested by Taecharungroj and Mathayomchan (2019), dimensional salience-valence (DSV) based on the parallel LDA was used to reveal the dimensionality of reviews. Then, lexical salience-valence (LSV) analysis was utilized to identify 10 factors (words) fundamental to each dimension. Words in each dimension are analyzed for whether they reflect guest experiences in positive or negative sentiment. The weights of salience are indicated by the sizes of the bubbles displayed in the relevant charts. A larger weight for a word gives a larger sized bubble in the graph. The larger sized bubble means a higher frequency of that word in the reviews. The term valence is calculated as the difference of the mean of the word found in positive reviews and the mean of the word found in negative reviews, divided by the sum of means of the word found in positive reviews and negative reviews, as follows.

T e r m V a l e n c e = (x P O S - x N E G) / (x P O S + x N E G)

Therefore, negative valence implies that the word is found in negative reviews more than in positive reviews. Even though those words with negative valence might not be always involved in the context of negative customer experience, there might be significant concerns when a guest shares and reviews their experiences.

These keywords were further plotted on the lexical salience-valence analysis grid to determine the weight and positivity of each dimension and word. These were grouped into four quadrants based on Kalnaovakul and Promsivapallop (2021), which were (a) High Salience – High Valence quadrant, (b) High Salience – Low Valence quadrant, (c) Low Salience – Low Valence quadrant, and (d) Low Salience – High Valence quadrant. The mean of salience was calculated in each dataset to separate high salience and low salience along the y-axis. The mean of valence was used to separate high valence and low valence. The x-axis and y-axis bound the four quadrants of lexical salience-valence (LSV) plane. An LSV chart containing bubbles of words of relevant dimensions, with the sizes of the bubbles representing the standardized weights of salience, was generated for the analysis of each dataset.

Results

The results of the study are reported based on the analysis of four datasets. Overall, all four datasets have similar average review ratings in the range 4.51–4.55 on a 5-point scale. Furthermore, 89.75% of the comments are classified as positive comments and 10.25% of the comments are regarded as negative. The overall rating scores show that customers in all groups are generally highly satisfied with the experiences of their stay at the sampled hotels.

The total dataset

LDA result

According to the elbow method, six clusters or dimensions of the hotel service quality can be identified as the appropriate number of clusters in the total dataset. The six clusters (dimensions) of hotel service quality discovered by the LDA, including the terms in each dimension, are reported in Table 2. These six dimensions are named: leisure activities, tangibles and surroundings, reliability, responsiveness, service process, and food. Although there are a number overlapping terms among the six dimensions, there are distinctive themes discovered in each dimension.

• The leisure activities dimension is clustered around family activities, swimming pool, and kid’s club.

• The tangibles and surroundings dimension features terms that represent hotel facilities, amenities, beach, and the surrounding area.

• The reliability dimension includes terms that relate to the service outcome factor. It indicates the quality of room and service, including its convenient accessibility to hotel facilities and beach. These terms include examples such as room, clean, stay, staff, friendly, location, pool, and beach. Content analysis of the online reviews discovered that all terms imply the service outcome.

• The responsiveness dimension comprises terms that represent the quality service provided by staff such as staff, service, and amazing.

• The service process dimension involves terms that relate directly to the service procedures such as check (check-in), book and staff.

• The food dimension encompasses dining experience of the customers, including terms such as food, restaurant, service, staff, and friendly.

Table 2.

Dimensions and underlying terms of all datasets.

Dataset	Dimension	Word	Dimension	Word	Dimension	Word	Dimension	Word
Total dataset (6 dimensions)	Leisure activities	Kid	Tangibles and surroundings	Beach	Reliability	Room	Responsiveness	Staff
		Family		Pool		Staff		Service
		Pool		Room		Pool		Amazing
		Child		Restaurant		Beach		Time
		Staff		Food		Clean		Villa
		Holiday		Area		Friendly		Beautiful
		Time		Breakfast		Location		Experience
		Activity		day		Breakfast		day
		Club		bar		Stay		Room
		Love		Water		Helpful		Wonderful
	Service process	Room	Food	Food
		Staff		Service
		Check		Staff
		day		Place
		Book		Best
		Time		Recommend
		Service		Come
		Night		Amazing
		Need		Restaurant
		Come		Friendly
Couple dataset (3 dimensions)	Tangibles and surroundings	Pool	Service process	Room	Responsiveness	Staff
		Beach		day		Service
		Room		Time		Amazing
		Staff		Service		Food
		Restaurant		Staff		Time
		Breakfast		Pool		Friendly
		Food		Book		Place
		bar		Need		Room
		Clean		Check		Best
		Stay		Night		Recommend
Family dataset (4 dimensions)	Responsiveness	Staff	Tangibles and surroundings	Pool	Service process	Room	Leisure activities	Kid
		Service		Beach		Day		Family
		Food		Room		Staff		Staff
		Amazing		Restaurant		Book		Pool
		Friendly		Breakfast		Time		Holiday
		Room		Staff		Service		Love
		Best		Family		Check		Time
		Time		Walk		Night		Activity
		Restaurant		Food		Need		Amazing
		Helpful		Clean		Pool		Friendly
Friend dataset (4 dimensions)	Service process	Room	Responsiveness	Staff	Reliability	Pool	Food	Food
		day		Friendly		Beach		Service
		Staff		Service		Room		Place
		Check		Room		Staff		Best
		Time		Amazing		Breakfast		Restaurant
		Need		Come		Walk		Recommend
		Breakfast		Helpful		Restaurant		Staff
		Book		Place		Location		Amazing
		Night		Pool		View		Come
		Guest		Holiday		Friendly		Highly

DSV analysis

The next step after dimension identification was the salience and valence analysis of each dimension reported in Table 3 and visually illustrated in Figure 1. The predicted results by the model show accuracy of 93.57%, which is above the threshold level 65% for acceptable machine learning predictions (Kalnaovakul and Promsivapallop, 2021; Taecharungroj and Mathayomchan, 2019). F-measure (majority/minority class) and Cohen’s Kappa were 0.96/0.66, and 0.63 respectively. The average F-measure of positive-negative sentiment classification is around 0.60 and 0.80 respectively (Pang and Lee, 2004). Based on Madhukar (2022), Cohen’s Kappa coefficient indicates the reliability of agreement in the data collected for training purpose, regarding whether it is a correct representation of overall data, and its values are interpreted as: 0.81–1.00 for almost perfect agreement, 0.61–0.80 for substantial, 0.41–0.60 for moderate, 0.21–0.40 for fair, and 0.01–0.20 for none to slight (McHugh, 2012). Therefore, this classification prediction is fairly accurate.

Table 3.

Dimension salience-valence analysis of all datasets.

Dataset	Salience/Valence	Leisure activities	Tangibles and surroundings	Reliability	Responsiveness	Service process	Food
Total	Salience	7.39%	15.32%	30.77%	19.32%	11.52%	15.68%
Total	Valence	8.16%	−2.51%	7.28%	9.19%	−49.83%	9.62%
Couple	Salience		42.76%		41.07%	16.17%
Couple	Valence		6.27%		9.93%	−41.78%
Family	Salience	16.95%	38.23%		30.96%	13.86%
Family	Valence	8.98%	4.52%		9.67%	−45.05%
Friend	Salience			32.91%	37.22%	11.35%	18.52%
Friend	Valence			2.58%	6.82%	−38.73%	5.44%

Figure 1.

Dimension salience-valence chart (total dataset).

The DSVA reveals that reliability is the most salient or most frequently mentioned dimension (30.77%) and food is the dimension with the most positive valence (9.62%). Responsiveness has the second highest salience of 19.32% and second highest positive valence at 9.19%. It is important to note that service process has the most negative valence (−49.83%), along with tangibles and sourrondings illustrating relatively negative valence (−2.51%).

LSV analysis

Words in each dimension are analyzed for whether they reflect customer experiences in positive or negative sentiment. The weight of salience is indicated by size of bubble in Figure 2. A higher weight for a word gives a larger sized bubble in the graph, meaning higher frequency of the word in the reviews.

Figure 2.

Lexical salience-valence chart of the total dataset (Mean of Salience: 3.98, Mean of Valence: −0.05).

As displayed in Figure 2, the lexical salience-valence chart of the total dataset separates the terms into four quadrants: high salience – high valence, high salience – low valence, low salience – low valence, and low salience – high valence. In addition, the four most frequently used words in the reviews represented by the largest bubbles were room, pool, staff and beach.

The 1^st quadrant, high salience-high valence, refers to the terms that are high in frequency of being mentioned and appearing in positive or higher than average reviews. The most notable terms in this quadrant include staff, beach, and food. Staff and beach have the largest bubbles, indicating the high levels of frequency of them being mentioned in the review. Of the six dimensions, staff is key to five dimensions including leisure activities, reliability, responsiveness, service process, and food dimensions. Beach is dominant in two dimensions including tangibles and surroundings, and reliability. Therefore, staff, beach, and food are most frequently mentioned and generally positive in the reviews, implying that these terms are crucial to guest satisfaction.

The 2^nd quadrant, high salience – low valence, represents the expressions in the reviews that are frequently used but tend to be mentioned more in negative category of the reviews. In other words, guests are generally less satisfied with these items and talked frequently about them in the reviews. These include room, pool, breakfast, and service, ranked in order of their bubble sizes. These terms appear as important factors in service process dimension and food dimension.

The 3^rd quadrant is characterized by low salience – low valence. This indicates the terms being mentioned less frequently and generally appearing in negative reviews. Keywords found in this category include kid, clean, check, book, club, child, area and water. Kid and clean are noted to display the largest sizes of the bubbles. These terms are parts of leisure activities, tangibles and surroundings, responsiveness, and service process.

The 4^th quadrant, low salience – high valence, includes terms that are less frequently used in the reviews but mostly appear in the positive reviews. Key terms include family, helpful, villa, bar, and activity. They are parts of all dimensions except service process.

The couple dataset

LDA result

The elbow method suggests three dimensions for the couple dataset. Details of the three dimensions are named consistently with the total dataset and reported in Table 2. These include tangibles and surroundings, service process, and responsiveness. Key terms found in each dimension are in line with the total dataset.

DSV analysis

The salience and valence analysis of each dimension of the couple dataset are reported in Table 3 and visually illustrated in Figure 3. The accuracy of model was 91.32%, F-measure (majority/minority class), and Cohen’s kappa were 0.61/0.95, and 0.56 respectively. For the couples, both tangibles and surroundings (42.76%) and responsiveness (41.07%) are the more salient or most frequently mentioned dimensions, more so than service process (16.17%). In addition, both tangibles and surroundings (6.27%) and responsiveness (9.93%) have positive valence, whereas service process (−41.78%) is the only dimension found with a negative valence.

Figure 3.

Dimensional salience-valence chart of the couple dataset.

LSV analysis

Following the same procedures as for the total dataset, LSVA was conducted with the couple dataset and is displayed in Figure 4. The 1^st quadrant, high salience-high valence, includes pool, staff, and beach as notable words with larger bubbles. These words are present in all three dimensions with pool and beach being significant in the tangibles and surroundings dimension, pool in the service process dimension and staff being key in the responsiveness dimension.

Figure 4.

Lexical salience-valence analysis of the couple dataset (Mean of Salience: 3.48, Mean of Valence: −0.09).

The 2^nd quadrant, high salience – low valence, comprises room and service as key terms to this quadrant, represented by sizable bubbles. Both terms are integral to the responsiveness dimension, with room being also vital to the service process dimension.

The 3^rd quadrant, low salience – low valence, is made up of many terms in the service process dimension such as day, night, book and check. Time and stay are also included in this dimension.

The 4^th quadrant, low salience – high valence, includes several terms of the responsiveness dimension, namely amazing, friendly, and recommend. Bar and clean, which are parts of the tangibles and surroundings dimension are also found in this quadrant.

The family dataset

LDA result

The elbow method indicates the reviews comprise four dimensions for the family dataset. The four dimensions are named consistently with the total dataset and reported in Table 2. These include responsiveness, tangibles and surroundings, service process, and leisure activities. Key terms found in each dimension are in line with those found in the total dataset.

DSV analysis

The salience and valence analysis of each dimension of the family dataset are reported in Table 3 and Figure 5. The accuracy of the model was 91.40%. F-measure (majority/minority class) 0.95/0.61, and Cohen’s kappa 0.56. For the families, tangibles and surroundings (38.23%) and responsiveness (30.96%) are more salient or more frequently mentioned dimensions than leisure activities (16.95%) and service process (13.86%). In addition, responsiveness (9.67%), leisure activities (8.98%), and tangibles and surroundings (4.25%) have positive valence, while service process (−45.05%) is the only negative valence dimension in this dataset. The result of service process having negative valence is consistent with the results of the two previous datasets.

Figure 5.

Dimensional salience-valence chart of the family dataset.

LSV analysis

As presented in Figure 6, pool, beach, room, and kid are the largest bubbles in the LSVA chart for the family dataset. In the first quadrant for high salience-high valence, only family (leisure activities) and pool (tangibles and surroundings, and leisure activities) clearly appear in this category.

Figure 6.

Lexical salience-valence analysis of the family dataset (Mean of Salience: 3.43, Mean of Valence: −0.03).

The 2^nd quadrant, high salience – low valence, comprises room, kid, food and staff. These terms appear in all four dimensions.

The 3^rd quadrant, low salience – low valence, has words represented by smaller bubbles such as book, clean, need, and day, all of which are in service process dimension. Other notable terms are walk and clean, which are parts of tangibles and surroundings dimension.

The 4^th quadrant, low salience – high valence, includes small bubbles with terms in responsiveness dimension such as helpful, friendly, and best, and in leisure activities dimension such as restaurant, activity, love, friendly, and amazing.

The friend dataset

LDA result

The elbow method indicates the reviews comprise four dimensions for the friend dataset. Details of the four dimensions are named consistently with the total dataset and reported in Table 2. These include service process, responsiveness, reliability, and food. Key terms found in each dimension are in line with those found in the total dataset.

DSV analysis

The salience and valence analysis of each dimension of the friend dataset are reported in Table 3 and Figure 7. The accuracy of the model was 93.38%. F-measure (majority/minority class) 0.96/0.55, and Cohen’s kappa 0.52. For the friend data, responsiveness (37.22%) and reliability (32.91%) are more salient or more frequently mentioned dimensions than food (18.52%) and service process (11.35%). In addition, responsiveness (6.82%), food (5.44%), and reliability (2.58%) have positive valences, while service process (−38.73%) is the only negative valence dimension in this dataset. The result that service process has a negative valence is consistent with the results of the three previous datasets.

Figure 7.

Dimensional salience-valence chart of the friend dataset.

LSV analysis

As presented in Figure 8, pool, beach, room, and staff are the largest bubbles in the LSVA chart for the friend dataset. In the first quadrant for high salience-high valence, only food (food), service, and friendly (responsiveness and reliability) clearly appear in this category.

Figure 8.

Lexical salience-valence analysis of the friend dataset (Mean of Salience: 3.20, Mean of Valence: −0.01).

The 2^nd quadrant, high salience – low valence, comprises room, pool, staff and beach. The term “staff” appears in all four dimensions.

The 3^rd quadrant, low salience – low valence, has words represented by smaller bubbles such as book, check, need, and day, all of which are in service process dimension.

The 4^th quadrant, low salience – high valence, includes small bubbles with terms in responsiveness dimension such as helpful and amazing, and in food dimension such as recommend, best, and amazing.

Discussion and conclusion

Online reviews have increasingly been recognized as an important reflection and source of customer satisfaction and service quality insights to the hospitality and tourism industry. As indicated by prior literature, online reviews permit an understanding of customer satisfaction and service quality in the hotel industry through data mining (Peres and Paladini, 2021; Thu, 2020). This current study provides insights into service quality dimensions and attributes based on online reviews from the TripAdvisor platform, such that determine guest satisfaction in the hotel industry of Phuket island, Thailand; and may be applied to other tropical beach destinations. The study has attempted to fill the literature gaps in several aspects.

First, the study has identified six general dimensions of guest service quality in the hotel industry, namely leisure activities, tangibles and surroundings, reliability, responsiveness, service process, and food, with identification of attributes in each dimension, which are in line with the HOLSERV Plus model (Boon et al., 2014) and also with previous hotel online review studies (Chittiprolu et al., 2021; Padma and Ahn, 2020; Suvachart, 2021). As summarized in Table 4, service quality dimensions discovered from the topic modelling of online reviews in this study are found to relate to the service quality dimensions offered by the HOLSERV Plus model. The tangibles and surroundings dimension in this study correlates with the room dimension and the surroundings dimension in the HOLSERV Plus model. The findings in the current study suggest combining the room dimension and surroundings dimensions of HOLSERV Plus into one unique tangibles and surroundings dimension, as they were found to appear together and relate to each other in the online reviews.

Table 4.

Comparison between HOLSERV and dimensions based on topic modeling of each dataset.

HOLSERV plus	Service quality dimension (dataset)
HOLSERV plus	Total	Couple	Family	Friend
Room	Tangibles & surroundings	Tangibles & surroundings	Tangibles & surroundings
Surroundings	Tangibles & surroundings	Tangibles & surroundings	Tangibles & surroundings
Employee	Responsiveness	Responsiveness	Responsiveness	Responsiveness
Reliability	Reliability			Reliability
Reliability	Service process	Service process	Service process	Service process
Facility	Food			Food
Facility	Leisure activities		Leisure activities

Furthermore, the employee dimension of the HOLSERV Plus was found to correspond to the responsiveness dimension of the current study, as the responsiveness was discovered to comprise mainly quality of service provided by staff. The reliability dimension of the HOLSERV Plus, on the other hand, can be associated with two dimensions of service quality discovered in the online reviews by this study, including reliability and service process dimensions. In other words, the current study offers service process as an additional and unique component of reliability of the HOLSERV Plus model. In addition, facility, which is the last dimension of HOLSERV Plus, in this study would be comprised of two related dimensions, namely food dimension and leisure activities dimension. Therefore, this study is able to confirm the robustness of HOLSERV Plus model, although with some modifications required. The HOLSERV Plus model has proven to be applicable to hotel service quality in the online review context.

The findings further offer in-depth understanding in the salience and the valence of each dimension. Based on the summary of results in Table 5, reliability, responsiveness, and service process have been mentioned most frequently among all dimensions, indicating their highest salience. On the other hand, it has become clear that the service process dimension has the most negative valence or tends to appear more toward dissatisfaction in the reviews. The results are in line with previous research such as He et al. (2017) and Li et al. (2013), who found elements in service process as key factors that contributed to customer dissatisfaction in the hotel online reviews.

Table 5.

Summary of dimensions of the four datasets.

Dimensions	Total dataset	Couple dataset	Family dataset	Friend dataset
Leisure activities	S = 7.39%, V = 8.16%	N/A	S = 16.95%, V = 8.98%	N/A
Tangibles and surroundings	S = 15.32%, V = −2.51%	S = 42.76%, V = 6.27%	S = 38.23%, V = 4.52%	N/A
Reliability	S = 30.77%, V = 7.28%	N/A	N/A	S = 32.91%, V = 2.58%
Responsiveness	S = 19.32%, V = 9.19%	S = 41.07%, V = 9.93%	S = 30.96%, V = 9.67%	S = 37.22%, V = 6.82%
Service process	S = 11.52, V = −49.83%	S = 16.17%, V = −41.78%	S = 13.86%, V = −45.05%	S = 11.35%, V = −38.73%
Food	S = 15.68%, V = 9.62%	N/A	N/A	S = 18.52%, V = 5.44%

Remarks: S; Salience, V; Valence.

At the attribute level, the findings show that room has the most negative valence. These results are confirmed across all four datasets. In addition to room at the attribute level, three additional attributes including pool, staff, and beach are found to be among the most frequently mentioned terms in the review. In terms of valence, room appears to be the single term with the most negative valence, which influenced dissatisfaction and complaints. On the other hand, staff and beach are two terms that influence satisfaction in reviews the most. The findings provide support to previous research, such as Anagostopoulou et al. (2019) who found staff attentiveness and professionalism, and hotel location, as important factors that contribute to hotel customer satisfaction.

Second, the findings of the current research are differentiated from previous hotel online studies as they provide evidence to support the view that customers are diverse in their views of service quality and satisfaction. Therefore, conclusions on service quality and guest satisfaction determinants in the hotel industry are not applicable to all groups of customers. Service quality and satisfaction components in the current research are found to vary across different groups of customers. In other words, different groups of hotel customers tend to put emphasis on different sets of service quality factors. Although the general online reviews of all groups of customers display six components of service quality and several specific attributes being key in the reviews, a closer examination into each group of customers provides unique results. While tangibles and surroundings, responsiveness, and service process dimensions are only three service quality dimensions to the couple segment, leisure activities dimension was an additional dimension to the family group in addition to the three dimensions identified in the couple dataset. In addition, it is important to note that leisure activities dimension is found to be a key component of the family segment only. On the other hand, the friend segment is the only customer group that has food as an important hotel service quality dimension, alongside with responsiveness, reliability, and service process dimensions. Therefore, this study confirms differences in service quality determinations across customer segments.

In addition, while room, pool, beach, and staff appear to be dominant in determining customer satisfaction in online reviews based on their salience levels, the valence results of these terms are not consistent in all three customer segments. Although room was discovered to have very negative valence across the three customer groups, the magnitude of valence tends to vary among pool, beach and staff by customer group. These three terms appear to be more positive in the reviews of the couple segment. Therefore, the couple segment tends to be more satisfied with pool, beach, and staff than the family and the friend segments. This result adds to the study provided by Almeida and Pelissari (2019) who discovered that the couples value the quality of room and services provided by staff, by indicating beach and pool as two additional key service quality factors especially in beach destination like Phuket in this study. In addition, it is discovered that the couples are easier to please with these service quality attributes as compared to the family and the friend segments. Moreover, kid and family terms were more dominant in the family segment but were not shown in the results of the other two segments. This implies that the family group value leisure activities and services that support their children and family more than the couple and friend groups. The results of this study clearly illustrate the differences in expectations, preferences and values of each customer group.

Theoretical implications

The results of the study have offered theoretical implications in two important aspects. First, the current study has expanded the prior existing general hotel service quality models such as LODGSERV (Knutson et al., 1990), Lodging Quality Index (Getty and Getty, 2003), and HOLSERV (Mei et al., 1999) and most particularly HOLSERV Plus (Boon et al., 2014) by providing a specific hotel service quality model applicable to beach destinations like Phuket. Specifically, it has confirmed the robustness of HOLSERV Plus model that can be applied to the online review context with some modifications offered in the results of this study.

Secondly, it has confirmed differences in service quality perceptions and determinations across different groups of hotel customers. Among the six dimensions of service quality identified in the study, the two components responsiveness and service process are fundamental to all groups of customers. However, three additional dimensions namely reliability, food, and leisure activities are perceived to be not equally important across the three groups of customers under study. Therefore, the findings provide confirmation that the same set of service quality dimensions and attributes may not be equally applied to different groups of customers as each group has its own unique requirements and expectations. For example, leisure activities for kids and families become an important aspect of service quality and satisfaction determinant of the family segment. On the other hand, food dimension was relatively more important to the friend segment than to the other groups of customers.

In addition, while reliability, tangibles and surroundings, and responsiveness were found in this study to be vital service quality determinants to customers, as indicated by the relatively high positive saliences, service process was revealed as the most sensitive dimension that determines customer dissatisfaction. This is indicated by the negative valences consistently across all datasets. Moreover, the salience of the service process dimension is comparatively lowest among the different groups of customers. The dimension or factor influencing customer satisfaction can be related to the “Herzberg two-factor theory” which classified two factors: hygiene and motivational factors (Chittiprolu et al., 2021; Herzberg, 1966). Hygiene factors do not create satisfaction, but lacking these hygiene factors creates customer dissatisfaction. On the other hand, motivational factors create exceeding satisfaction when it is delivered properly but will not create dissatisfaction if they are absent. The service process dimension found in this study can be considered a motivational factor as it can lift up service quality level in the final judgement of the overall service quality to reach exceeding satisfaction level.

Therefore, in addition to comfort in guestroom, beach accessibility, and pool facilities, services provided by staff and service procedures including room servicing, and check-in process, are considered important determinants of service quality and guest satisfaction in the current model offered in this study. The outcome of the study further highlights specific attributes that are influential to hotel guest satisfaction in the model, including room, pool, beach, and staff.

Practical implications

From the hotel operators’ point of view, hotel managers must ensure the management of all key relevant components of service quality dimensions and attributes in order to maximize customer satisfaction and positive online word of mouth. Firstly, as suggested in the findings of this study, hotel managers are advised to ensure that two service quality dimensions, namely service process dimension and responsiveness dimension, are strictly maintained as basic requirements across all groups of customers. The service process dimension, which involves terms that relate directly to the service procedures such as check (check-in), booking, and staff, consistently had the most negative valence across all groups. Improvements in these attributes are immediately required. Therefore, it is advised that hotels need to establish or revise their service operation standards and train their staff regularly to ensure the service process and procedures are consistently followed and appropriately delivered to the guests. Furthermore, it is vital for service providers to monitor service process operations, including all guest contact points of guest experience during their stay, to ensure quality services are consistently delivered to the customers. These guest contact areas should include all key attributes identified in this study including swimming pool, room, staff, and beach accessibility. Furthermore, as indicated in the findings, hotel managers are required to urgently direct efforts to ensure the quality of facilities and attributes in the guest rooms as it is most frequently mentioned and appeared more towards negative reviews.

Furthermore, as responsiveness dimension was found as a source of customer satisfaction in all customer groups, hotel managers are advised to ensure their staff always provide caring and responsive services to their guests. The whole process of staff recruitment, staff training, and staff motivation must be taken into consideration. Managers should focus their efforts to recruit the right service-minded staff, develop them with training programs to deliver responsive services, and keep them motivated to maintain high quality service attitude.

With regards to the specific attributes, hotel managers are urged to ensure and enhance the quality of tangible appearance and good maintenance of guest rooms and swimming pool. In addition, hotels should also seek to create convenient access to the beach. Even if the hotel is not located close to the beach, it is advised to provide convenient shuttle service for the guests to gain easy access to the beach.

In addition, service operators should recognize the need to focus service and facility design to meet the requirements of a specific group of customers. As the results suggest, dissimilar groups of customers value different sets of service quality dimensions and attributes based on their needs and expectations. Therefore, hotel managers are advised to carefully study the needs of their target market. Based on the results of this study, hotels that target the couple, the family, or the friend segments may design their services around the key dimensions and attributes and closely monitor the relevant service quality determinants. For example, hotel managers are advised to focus extra efforts on improving the quality of foodservice and the reliability of service, if the friend segment is their target market, as these are two particular service quality areas that determine the service quality perceptions of the friend segment. On the other hand, hotels are required to pay particular attention in offering leisure activities to children as well as easy access to the surroundings and nearby tourist attractions for the family segment. In addition, it is imperative to ensure the quality of tangibles and surroundings: especially guestroom, swimming pool, and beach access for the couple market.

Another important implication to hotel managers is the need to use online reviews as an important tool for service quality management. As online reviews have increasingly become a significant influential source of information for customer decision-making, hotel managers should aim to minimize the negative aspects of reviews by enhancing guest experiences during their stay. The results of the current study offer details that need to be monitored in general and in each specific group as fundamental service quality determinants to the customers. In order to enhance current guest satisfaction and thus positive reviews, improvements in service process, room quality, swimming pool quality, and staff quality through training, should be pursued. Improving these aspects of service will enhance the quality view of customers, which eventually boosts positive online review comments. Furthermore, online feedback needs to be regularly monitored by hotel managers to keep up with the customer preferences and service quality feedback.

Limitations and future research

Although the present study has offered valuable information on hotel service quality, there are several limitations that should be borne in mind when interpreting the results. The current study has focused specifically only on one location, namely Phuket as the place of study. Although Phuket has been recognized as a world-class tourist holiday beach resort with many diverse hotels, it may not permit generalizations to other tourist destinations that are different in characteristics. For example, city destinations or culture-based destinations may feature different kinds of customers who may have different sets of needs and expectations. As a result, hotels located in such destinations may experience dissimilar service quality dimensions and attributes. In addition, the current study offers three groups of customers for analysis. There might be other groups of customers who may display unique requirements and thus unique sets of service quality dimensions and attributes. Therefore, in order to increase the generalizability and the scope of study, future research should seek to cover wider geographic areas including a variety of destination categories and different customer segments. Another important limitation of the current study is the lack of deeply qualitative analysis into specific service quality dimensions and attributes. Thus, based on the findings of the current study, future research is advised to further extend analysis to find in-depth meaning of each specific factor by applying content analysis to the online comments. Furthermore, the results may also be verified with in-depth interviews with hotel managers.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Faculty of Hospitality and Tourism, Prince of Songkla University.

ORCID iDs

Kanapot Kalnaovakul

Pornpisanu Promsivapallop

Author Biographies

Kanapot Kalnaovakul is a lecturer in Hospitality Management at the Faculty of Hospitality and Tourism, Prince of Songkla University, Phuket Campus, where he obtained his MBA in Hospitality and Tourism Management. His research interests include hospitality service quality management and destination management.

Dr. Pornpisanu Promsivapallop is currently an Assistant Professor in Hospitality Management and Dean at the Faculty of Hospitality and Tourism, Prince of Songkla University, Phuket Campus. He obtained his Ph.D. in Hospitality Management from University of Surrey, U.K. His research interests include gastronomy tourism, tourist behaviour, destination marketing and management, and hospitality service management.

References

Ahmad

Papastathopoulos

(2019) Measuring service quality and customer satisfaction of the small-and medium-sized hotels (SMSHs) industry: lessons from United Arab Emirates (UAE). Tourism Review 74(3): 349–370.

Aksu

Albayrak

Caber

(2021) Hotel customer segmentation according to eco-service quality perception: the case of Russian tourists. Journal of Hospitality and Tourism Insights 5: 501–514. DOI: 10.1108/JHTI-10-2020-0188.

Ali

B J

Gardi

Jabbar Othman

, et al. (2021) Hotel service quality: The impact of service quality on customer satisfaction in hospitality. International Journal of Engineering, Business and Management 5(3): 14–28.

Almeida

GSD

Pelissari

(2019) Customer satisfaction based on the attributes of accommodation services. Revista Brasileira de Pesquisa em Turismo 13: 32–53.

Alrawadieh

Law

(2019) Determinants of hotel guests’ satisfaction from the perspective of online hotel reviewers. International Journal of Culture, Tourism and Hospitality Research 13(1): 84–97.

Anagostopoulou

Buhalis

Kountouri

, et al. (2019) The impact of online reputation on hotel profitability. International Journal of Contemporary Hospitality Management 32(1): 20–39.

Ban

Choi

, et al. (2019) Investigating key attributes in experience and satisfaction of hotel customer using online review data. Sustainability 11(23): 6570.

Berezina

Bilgihan

Cobanoglu

, et al. (2016) Understanding satisfied and dissatisfied hotel customers: text mining of online hotel reviews. Journal of Hospitality Marketing & Management 25(1): 1–24.

Boon

Bonera

Bigi

(2014) Measuring hotel service quality from online consumer reviews: a proposed method. In: Xiang

Tussyadiah

(eds) Information and Communication Technologies in Tourism. Springer, pp. 367–378.

10.

C9 Hotelwors (2020) Robust Arrivals in 2nd Half and India Growth Rebalance Phuket Tourism Market. Retrieved 13 November 2020 from: https://www.c9hotelworks.com/wp-content/uploads/2020/02/phuket-market-update-2020-01.pdf

11.

Chittiprolu

Samala

Bellamkonda

(2021) Heritage hotels and customer experience: a text mining analysis of online reviews. International Journal of Culture, Tourism and Hospitality Research 15(2): 131–156.

12.

Cronin Jr

Taylor

(1992) Measuring service quality: a reexamination and extension. Journal of Marketing 56(3): 55–68.

13.

Dabestani

Shahin

Saljoughian

, et al. (2016) Importance-performance analysis of service quality dimensions for the customer groups segmented by DEA: the case of four-star hotels. International Journal of Quality & Reliability Management 33(2): 160–177.

14.

Dinçer

Alrawadieh

(2017) Negative word of mouse in the hotel industry: a content analysis of online reviews on luxury hotels in Jordan. Journal of Hospitality Marketing & Management 26(8): 785–804.

15.

Dong

Zhang

(2014) Classification of customer satisfaction attributes: an application of online hotel review analysis. In: Paper presented at 13^th Conference on E-Business, e-Services and e-Society, Sanya, China, November 28–30, vol. 445, pp. 238–250.

16.

Fang

Kucukusta

, et al. (2016) Analysis of the perceived value of online tourism reviews: influence of readability and reviewer characteristics. Tourism Management 52: 498–506.

17.

Feiz

Taherian

Zarei

(2011) Service quality and customer satisfaction in hotel industry (case study: Mashhad hotels). Journal of Business Administration Researches 3(6): 123–149.

18.

Gan

Ferns

, et al. (2017) A text mining and multidimensional sentiment analysis of online restaurant reviews. Journal of Quality Assurance in Hospitality & Tourism 18(4): 465–492.

19.

Getty

(2003) Lodging quality index (LQI): assessing customers’ perceptions of quality delivery. International Journal of Contemporary Hospitality Management 15(2): 94–104.

20.

Tian

Tao

, et al. (2017) Application of social media analytics: a case of analyzing online hotel reviews. Online Information Review 41(7): 921–935.

21.

Herzberg

(1966) Motivate Employees?”. Cleveland, OH: World Publishing Company, p. 88.

22.

Kalnaovakul

Promsivapallop

(2021) Dimensions of night market visit experience of international tourists: an analysis of google reviews of night markets in Phuket, Thailand. Asia-Pacific Journal of Social Science Review 21(3): 57–73.

23.

Kim

Heo

(2016) Analysis of satisfiers and dissatisfiers in online hotel reviews on social media. International Journal of Contemporary Hospitality Management 28(9): 1915–1936.

24.

Knutson

Stevens

Wullaert

, et al. (1990) LODGSERV: a service quality index for the lodging industry. Hospitality Research Journal 14(2): 277–284.

25.

Lai

IKW

Hitchcock

(2016) A comparison of service quality attributes for stand-alone and resort-based luxury hotels in Macau: 3-Dimensional importance-performance analysis. Tourism Management 55: 139–159.

26.

Lee

(2018) Assessing the helpfulness of online hotel reviews: a classification-based approach. Telematics and Informatics 35(2): 436–445.

27.

Law

(2013) Determinants of customer satisfaction in the hotel industry: an application of online review analysis. Asia Pacific Journal of Tourism Research 18(7): 784–802.

28.

Luo

, et al. (2021) Understanding service attributes of robot hotels: a sentiment analysis of customer online reviews. International Journal of Hospitality Management 98: 103032. DOI: 10.1016/j.ijhm.2021.103032.

29.

Luo

(2016) Guest-defined hotel service quality and its impacts on guest loyalty. Journal of Quality Assurance in Hospitality & Tourism 17(3): 311–332.

30.

Madhukar

(2022) Understanding Cohen’s Kappa score with hands-on implementation. Analytics India Magazine. Available at: https://analyticsindiamag.com/understanding-cohens-kappa-score-with-hands-on-implementation (Accessed 20 June 2022).

31.

McHugh

(2012) Interrater reliability: the kappa statistic. Biochemia Medica 22(3): 276–282.

32.

Mei

AWO

Dean

White

(1999) Analysing service quality in the hospitality industry. Managing Service Quality: An International Journal 9(2): 136–143.

33.

Nunkoo

Teeroovengadum

Ringle

, et al. (2020) Service quality and customer satisfaction: The moderating effects of hotel star rating. International Journal of Hospitality Management 91: 102414.

34.

Padlee

Thaw

Zulkiffli

SNA

(2019) The relationship between service quality, customer satisfaction and behavioural intentions. Tourism and Hospitality Management 25(1): 121–139.

35.

Padma

Ahn

(2020) Guest satisfaction & dissatisfaction in luxury hotels: an application of big data. International Journal of Hospitality Management 84: 102318.

36.

Pang

Lee

(2004) A sentimental education: Sentiment analysis using subjectivity summarization based on minimum cuts. In Proceedings of Meeting of the Association for Computational Linguistics (ACL-2004).

37.

Parasuraman

Zeithaml

Berry

(1985) A conceptual model of service quality and its implications for future research. Journal of Marketing 49(4): 41–50.

38.

Peres

Paladini

(2021) Exploring the attributes of hotel service quality in Florianópolis-SC, Brazil: an analysis of TripAdvisor reviews. Cogent Business & Management 8(1): 1926211.

39.

Promsivapallop

Kannaovakun

(2017) A comparative assessment of destination image, travel risk perceptions and travel intention by young travellers across three ASEAN countries: a study of German students. Asia Pacific Journal of Tourism Research 22(6): 634–650.

40.

Promsivapallop

Kannaovakun

(2018) Travel risk dimensions, personal-related factors, and intention to visit a destination: a study of young educated German adults. Asia Pacific Journal of Tourism Research 23(7): 639–655.

41.

Pullman

McGuire

Cleveland

(2005) Let me count the words: Quantifying open-ended interactions with guests. Cornell Hotel and Restaurant Administration Quarterly 46(3): 323–343.

42.

Sharma

Srivastava

(2018) Relationship between service quality and customer satisfaction in hotel industry. TRJ Tourism Research Journal 2(1): 42–49.

43.

Suvachart

(2021) Content analysis for promoting accommodation services business. KKBS Journal of Business Administration and Accountancy 5(1): 1–20.

44.

Swan

Comb

(1976) Product performance and consumer satisfaction: a newconcept. Journal of Marketing 40(4): 3–33.

45.

Taecharungroj

Mathayomchan

(2019) Analysing TripAdvisor reviews of tourist attractions in Phuket, Thailand. Tourism Management 75: 550–568.

46.

Tan

Liu

, et al. (2018) Evaluation nudge: effect of evaluation mode of online customer reviews on consumers’ preferences. Tourism Management 65: 29–40.

47.

Thu

HNT

(2020) Measuring guest satisfaction from online reviews: Evidence in Vietnam. Cogent Social Sciences 6(1): 1801117.

48.

Tian

McIntosh

(2021) What factors affect consumers’ dining sentiments and their ratings: evidence from restaurant online review data. Food Quality and Preference 88: 104060.

49.

Tsaur

Lin

(2005) Cultural differences of service quality and behavioral intention in tourist hotels. Journal of Hospitality & Leisure Marketing 13(1): 41–63.

50.

Wang

Peng

, et al. (2020) The differences in hotel selection among various types of travellers: a comparative analysis with a useful bounded rationality behavioural decision support model. Tourism Management 76: 103961.

51.

Xie

Chen

(2016) Online consumer review factors affecting offline hotel popularity: evidence from Tripadvisor. Journal of Travel & Tourism Marketing 33(2): 211–223.

52.

Yang

Shin

Joun

, et al. (2017) Exploring the comparative importance of online hotel reviews’ heuristic attributes in review helpfulness: a conjoint analysis approach. Journal of Travel & Tourism Marketing 34(7): 963–985.

53.

Zhou

Pearce

, et al. (2014) Refreshing hotel satisfaction studies by reconfiguring customer review data. International Journal of Hospitality Management 38: 1–10.