Journal of Business Research 171 (2024) 114378

Contents lists available at ScienceDirect

Journal of Business Research

journal homepage: www.elsevier.com/locate/jbusres

Wisdom of crowds: SWOT analysis based on hybrid text mining methods

using online reviews
Jie Wu a, b, *, Narisa Zhao a, *, Tong Yang a, c
a
Institute of Systems Engineering, Dalian University of Technology, Dalian, China
b
Department of Industrial Systems Engineering and Management, National University of Singapore, Singapore
c
Department of Information Systems and Analytics, National University of Singapore, Singapore

A R T I C L E I N F O A B S T R A C T

Keywords: Consumer online reviews are prolific, authentic, and real-time, providing potential business opportunities.
SWOT analysis Traditional SWOT analysis is highly subjective, lacks an effective focus on the core competitive advantages and
Online reviews disadvantages, and fails to act on the dynamic business environment. To address these issues, we propose an
Sentiment analysis
innovative methodology that utilizes hybrid text mining methods to conduct SWOT analysis using online re­
TF-IDF
Dynamic perspective
views. After first identifying the service attributes from the literature, we calculate attribute performance via
Core competitive advantages/disadvantages sentiment analysis. Second, guided by salience theory, factor importance is calculated based on the TF-IDF al­
gorithm. Following the essential meaning of SWOT factors, we then construct the factor determination rules.
Finally, we model the changing trends of SWOT factors from a dynamic perspective, which can help managers
develop preventive strategies. Online reviews of five hotels crawled from Ctrip were applied to validate the
proposed method. The comparison with two state-of-the-art methods also demonstrated its effectiveness.

1. Introduction exert action and monitor changes throughout a long-term competitive

situation (Coman & Ronen, 2009; Vlados, 2019). The above shortcom­
The ancient Chinese militarist Sun Tzu claimed that “if you know the ings lead to the lack of practicable SWOT analysis steps in the current
enemy and know yourself, you need not fear the result of a hundred stage of business operations (GÜRel, 2017). As a result, many strategic
battles”. As with Sun Tzu’s observation of strategy in the Art of War, planning managers even no longer trust, and thus, have gradually
competitive analysis is an important part of strategy formulation and abandoned, SWOT (Fehringer et al., 2006). However, such an extreme
implementation in the business environment (Bergen & Peteraf, 2002). move requires a revolution in existing strategic management and
SWOT (Strength, Weakness, Opportunity, and Threat) is a competitive corporate governance, which entails additional costs.
analysis tool that has long been used to assess an organization’s internal With the rapid development of information technology, consumers
capabilities (strengths and weaknesses) and external context (threats can easily post online reviews on e-commerce platforms. The sentiment
and opportunities) (Phadermrod et al., 2019). It is widely used in expressed in online reviews reflects consumers’ satisfaction with the
different industries because of the simplicity of its process of execution focal product/service, which is the essence of competitiveness (Wu
and presentation of results. et al., 2023). SWOT is a technique precisely created to conduct
Despite its popularity, SWOT analysis has been criticized regarding competitive strategy analysis (Coman & Ronen, 2009). Additionally,
its practical usefulness. First, traditional SWOT analysis is typically used performing SWOT analysis from the consumers’ perspective, rather than
to identify SWOT factors manually, which inevitably leads to subjective just from the organization’s perspective, ensures that the capabilities
results (Cheng et al., 2021; Hill & Westbrook, 1997). Second, existing perceived by the organization can be recognized by their customers as
studies rarely consider the importance of SWOT factors, resulting in well (Phadermrod et al., 2019). Consequently, using consumer-
managers failing to focus on the product’s core competitive advantages generated content as input to conduct SWOT analysis makes sense.
and disadvantages (Coman & Ronen, 2009; Phadermrod et al., 2019), Compared to the traditional data collected through surveys or ques­
especially when faced with limited resources. Third, the SWOT analysis tionnaires, using online reviews for SWOT analysis can result in multiple
is typically a one-time event, and it lacks mechanisms with which to benefits. First, online reviews are massive and authentic (Wu & Zhao,

* Corresponding authors at: Institute of Systems Engineering, Dalian University of Technology, No. 2 Linggong Road, Dalian 116024, China.
E-mail addresses: [email protected] (J. Wu), [email protected] (N. Zhao), [email protected] (T. Yang).

https://doi.org/10.1016/j.jbusres.2023.114378
Received 9 May 2022; Received in revised form 25 October 2023; Accepted 3 November 2023
Available online 20 November 2023
0148-2963/© 2023 Elsevier Inc. All rights reserved.
J. Wu et al. Journal of Business Research 171 (2024) 114378

2023; Yang et al., 2023c), which reinforces the generalizability and in the determination of factors and usually descriptive in terms of high-
portability of SWOT analysis. Second, the low cost of acquiring such level perceptions (Hill & Westbrook, 1997), which can lead to biased
data (Tirunillai & Tellis, 2014) can help organizations reduce their and more subjective results (Smith, 2003).
reliance on traditional means (e.g., questionnaires and interviews) and Second, many studies lack the discussion of the importance of SWOT
minimize the accompanying costs. Third, the real-time nature of online factors, which leads to managers lacking the ability to focus on core
reviews (Tirunillai & Tellis, 2014; Wu et al., 2023) also facilitates the issues in the face of SWOT analysis results (Büyüközkan & Ilıcak, 2019;
prompt understanding of organizations regarding their competitive Coman & Ronen, 2009), especially in businesses with limited resources.
position and thus enables the timely adjustment of their strategic plan­ The goal of SWOT analysis is to leverage core advantages to maximize
ning and resource allocation. their value creation potential; thus the identification of the core issues to
Despite the above benefits, there are difficulties in conducting SWOT be analyzed needs to be made in an accurate and timely manner (Coman
analysis based on such user-generated content (UGC). Specifically, (1) & Ronen, 2009). For this reason, researchers have proposed various
how can the product attributes that are valued by consumers be objec­ methods to measure the importance of SWOT factors. For instance,
tively identified as different SWOT factors? Further, (2) how can UGC be Yüksel & Daǧdeviren (2007) integrated analytic network process into
used to assess the factor importance and thus identify the company’s SWOT construction, and a similar concept can also be found in Sevkli
core competitive advantages and disadvantages? Furthermore, (3) how et al.’s (2012) work. Moreover, many scholars have developed Analytic
can a dynamic SWOT construct be developed that captures the micro Hierarchy Process-based SWOT analysis to quantify factor importance
trends of different factors? Presumably, these challenges hinder the (Kangas et al., 2001; Lee & Lin, 2008). These studies typically determine
exploitation of business opportunities from UGC for SWOT analysis. the relative importance of different factors based on expert or manager
To overcome the difficulties and challenges discussed above, in this judgment; then factor importance is calculated through a constructed
study an innovative framework that uses mixed text mining methods comparison matrix. To some extent, such methods can help businesses
was proposed. Specifically, after identifying service attributes from the prioritize their core advantages and disadvantages; however, the ob­
literature, attribute performance was calculated via sentiment analysis. tained priorities and weights of factors are mainly assessed by experts or
Then, based on salience theory (Oh, 2001), the TF-IDF algorithm was managers, and thus are inevitably subjective. In addition, such ap­
adopted to measure attribute importance. Next, on the basis of the proaches do not consider factor importance from the perspective of the
essential connotation of SWOT itself, we propose elaborate rules for customer (Phadermrod et al., 2019), which is nearly the most important
SWOT factor identification, which addresses the first difficulty. Based on aspect of product development.
the measuring results of attribute importance, the importance ranking of Third, traditional SWOT does not bestow the capacity to respond to a
SWOT factors is determined, which further clarifies both the core changing environment in a timely manner (Koch, 2000; Vlados, 2019).
competitive advantages and disadvantages, and this overcomes the Typical SWOT analysis provides a snapshot view where product per­
second difficulty. Finally, a dynamic SWOT construct with two di­ formance is simply measured statically, which is inconsistent with the
mensions is developed to model the evolutionary trajectory of the fac­ dynamics of a changing business world. Therefore, it has been suggested
tors, which tackles the third difficulty. To verify the effectiveness of the that SWOT analysis should be conducted from a dynamic perspective.
proposed method, we applied it to online hotel reviews crawled from For example, Vlados (2019) proposed a framework for analyzing the
Ctrip.com and compared the results with those obtained through two competitive situation by combining SWOT analysis with the concept of
state-of-the-art methods. To our knowledge, this is the first compre­ evolution in the strategic planning of contemporary organizations.
hensive study to address all of the abovementioned issues using online However, such studies tend to collect data through questionnaires or
reviews, which precisely shapes the uniqueness of this study. managerial interviews (Coman & Ronen, 2009), which require time-
consuming iterations to calibrate the SWOT factors.
2. Related works
2.2. Online review and text mining methods
According to the research objectives, in this section we deliver a
literature review of SWOT analysis, online reviews and text mining As Web 2.0 applications evolve, e-commerce platforms have begun
methods. to support communication interactions among their consumers. Online
reviews on these platforms are growing significantly (Bi et al., 2019c).
2.1. SWOT analysis Unlike traditional data collected through surveys/questionnaires, online
reviews spontaneously generated by consumers are more authentic and
Among competitive analytical tools, SWOT analysis is an effective massive (Yang et al., 2023c). More importantly, such data are easy to
tool for helping organizations understand their internal and external obtain, entail a low cost, and are timely, and they are easier for man­
business environment by analyzing their resources across four areas, agers to monitor and manage than traditional data (Tirunillai & Tellis,
namely strengths, weaknesses, opportunities, and threats, while assist­ 2014; Yang et al., 2023b). However, the confusion and diversity of such
ing with the formulation of strategic plans and decisions (Hill & West­ data make it challenging for organizations to integrate unstructured and
brook, 1997; Sevkli et al., 2012). Strengths and weaknesses are the massive data into the assessment of useful business opportunities.
internal factors that support and hinder an organization from achieving With the rapid development of information technology, natural
its mission, respectively. Opportunities and threats, on the other hand, language processing techniques, including text mining, have been pro­
are external factors that enable or disable an organization to accomplish moted. Text mining, which is also known as text data mining or
its mission, respectively (Dyson, 2004). knowledge discovery in text databases, usually refers to the process of
Simplicity is the main advantage of SWOT analysis, which has led to extracting interesting and nontrivial patterns or knowledge from un­
its extensive use in mainstream business and academia ever since it was structured text documents (Tan, 1999). Some commonly used text
first developed in the 1960s (Ghazinoory et al., 2011). In a survey mining techniques include sentiment analysis (Wu & Zhao, 2023) and
conducted by the Competitive Intelligence Foundation (Fehringer et al., feature extraction (Yang et al., 2023b). The purpose of sentiment anal­
2006) among 520 competitive intelligence professionals, SWOT analysis ysis is to determine the sentiment tendency of a text by calculating its
was supported by 82.6% of the respondents and was identified as the sentiment score (Tan et al., 2020). Feature extraction entails the
second most commonly used strategic analysis tool in the business extraction of information from text, and this process can reduce the
world. dimensionality of the original data feature space to extract important
Despite its merits, however, SWOT has also been criticized for a information (Yang et al., 2023b).
number of problems. First, traditional SWOT analysis is very subjective The application of the above techniques can be found in many

2
J. Wu et al. Journal of Business Research 171 (2024) 114378

studies. For example, Fan et al. (2017) used a naive Bayesian to calculate performance between two competing companies. Overall, scholars have
the sentiment of online reviews while carrying out product sales pre­ made great efforts to improve SWOT analysis in recent decades, but the
diction, and the results showed a higher prediction accuracy than other newly proposed methods still face one or more of the problems
methods. Some scholars have also applied sentiment analysis to mentioned above.
competitiveness analysis (Jin et al., 2017; Liu et al., 2019). Among them,
Liu et al. (2019) effectively obtained product competitiveness informa­ 3. Methodology
tion by mining domain-specific sentiment from online reviews, and the
authors confirmed that their method better reflects the consumer This section is focused on developing a framework with mixed
perspective than traditional methods. In addition, there are also appli­ methods to bridge the gaps identified in the literature review, namely,
cations in product ranking (Bi et al., 2019a), satisfaction modeling (Bi the subjectivity of factor determination, the lack of focus on core
et al., 2019b), etc. This is sufficient to illustrate text mining as an competitive advantages and disadvantages, and the failure to adapt to
effective technique for extracting valuable business opportunities from a the dynamic business environment. As shown in Fig. 1, we first use
large number of online reviews. sentiment analysis to reveal attribute performance from online reviews.
In particular, a few scholars have begun to explore various ways of Then, to clarify the core competitive advantages and disadvantages, we
conducting SWOT construction or analysis through online review (or extract feature words from online reviews based on the TF-IDF algorithm
user-generated content, UGC) mining. For instance, aiming to improve (Yang et al., 2023b) and further calculate the importance of different
health care service quality during the COVID-19 pandemic, Shah et al. attributes. Next, we construct the factor determination rules based on
(2021) manually identified the SWOT factors of a targeted hospital after the connotation of SWOT factors and capture the changing trends of
mining salient topics and sentiments from online doctor reviews. Simi­ these factors from a dynamic perspective.
larly, there are several online review-based studies on SWOT analysis in
the literature (Mitra & Jenamani, 2020; Rajendran et al., 2022). Despite 3.1. Attribute performance calculation via sentiment analysis
their application of online reviews, these studies have primarily iden­
tified SWOT factors manually and have not proposed quantitative rules To calculate product attribute performance, the reviews are first
for factor identification, which essentially left the subjective issues of divided into sentences, and then sentiment scores are calculated via
traditional SWOT analysis unaddressed. In contrast, Cheng et al. (2021) PaddlePaddle, which is an open-source natural language processing
proposed quantitative and clear rules for identifying SWOT factors by (NLP) tool provided by Baidu (Ma et al., 2019). Utilizing a massive
mining UGC, which to some extent reduces the subjectivity of SWOT Chinese corpus, this interface generates a high-dimensional word vector,
construction. However, this study still failed to discuss the process of and then develops various advanced sentiment analysis models based on
measuring the importance of SWOT factors to further help businesses deep learning algorithms such as Convolutional Neural Network and
clarify their core advantages/disadvantages; it also lacked the ability to Long Short-Term Memory. There are a large number of applications for
capture the evolutionary trajectory of the SWOT factors of businesses in PaddlePaddle in both industry and academia (Wu & Yang, 2023; Yang
a dynamic competitive environment. Besides, Cheng et al.’s (2021) et al., 2023c). Hence in this study, we adopt PaddlePaddle to calculate
method may not effectively enable the observation of subtle threats or the sentiment of the extracted sentences.
opportunities because it does not consider the specific differences in The sentiment score ranges from 0 to 1, with 0 indicating extremely

Fig. 1. Methodology framework.

3
J. Wu et al. Journal of Business Research 171 (2024) 114378

poor and 1 indicating extremely good. To obtain attribute-level perfor­ through the consideration of semantic information in text, important
mance, a domain attribute lexicon including attributes and their related keywords can be extracted (Duari & Bhatnagar, 2019; Xiong et al.,
vocabulary needs to be constructed. Then the algorithm Jieba, which has 2021). For example, Xiong et al. (2021) developed Semantic Clustering
been shown to be effective in searching for the most likely combination TextRank (SCTR), which is a heavy clustering algorithm based on the
(e.g., the collocation of negative word + adjectives) based on word Bidirectional Encoder Representation from Transformers (BERT) model,
frequency, was used to segment sentences into words. After matching, to extract important feature words. Using these methods, feature/attri­
sentences of different attributes are obtained. bute importance can also be further quantified. However, such deep
The pseudocode is shown in Fig. 2. According to the output of Al­ learning-based algorithms generally require massive computing re­
gorithm 1, the performance of attribute j can be calculated by Equation sources (Yang et al., 2023c), as well as requiring human labor for
j
(1), where sentimentm is the sentiment score of sentence m of attribute j, annotation, which may prevent their use by small and medium-sized
and Nj is the sentence number of attribute j. To better understand the companies.
calculation of attribute performance, an example is illustrated in Web In our work, inspired by salience theory (Guido, 1998), which in­
Appendix A. dicates that people are used to allocating their attention to areas of the
perceptual field where salient issues are considered, important attri­
Nj
∑
sentimentjm butes are considered those aspects that are prominent and frequently
Perfj = m=1 (1) mentioned. Moreover, this section is aimed at proposing a method for
Nj assessing attribute importance from a consumer group perspective with
broad contextual applicability and ease of use by other researchers/
3.2. TF-IDF-based attribute importance calculation managers. Among various methods of measuring attribute importance,
the TF-IDF algorithm may be suitable for our research context since it
Essentially, the SWOT factor importance measurement assesses can automatically filter out irrelevant words while extracting the salient
attribute importance. Cutting through customers’ perspectives, previous aspects that people value. Its low computational complexity (Wu &
studies typically obtained attribute importance by collecting surveys in Yuan, 2018) and unsupervised operation (i.e., no annotation data
which customers are asked to rate their feelings about different attri­ required) facilitate its use by researchers with different levels of
butes (Lai & Hitchcock, 2015; Phadermrod et al., 2019). Despite its knowledge, as well as by managers lacking significant computing re­
simplicity, the process of designing, distributing, and collecting ques­ sources. Besides, as a statistical method based on word frequency and
tionnaires requires a large investment of human and financial resources, inverse document frequency, TF-IDF is highly interpretable (Zhao et al.,
and it may also face issues such as inflated ratings and biased sampling 2019) and can help researchers and practitioners understand the ante­
(Chen et al., 2022). Moreover, extant studies have also proposed that cedents and effects of features (attributes).

Fig. 2. Algorithm for calculating attribute performance.

4
J. Wu et al. Journal of Business Research 171 (2024) 114378

Regarding the TF-IDF algorithm, it mines important feature words Table 1

and calculates the TF-IDF value from text content. TF-IDF is a commonly Rules for identifying SWOT factors.
used keyword analysis technique (Yang et al., 2023b). Its main idea is Focal company Competing company Perff - SWOT Rule
that if a word frequently appears in one document and rarely appears in Perfc factor ID
△Perf Internal △Perf Internal
other documents, then the word is considered to be of high importance Focal
S/W Competitor
S/W
to the current document (Wu & Zhao, 2023; Yang et al., 2023a). The
>0 Strength >0 Strength 0 Strength R1
algorithm consists of two parts, namely, the term frequency (TF) and the
>
Strength ≤ 0 Threat R2
inverse document frequency (IDF). Mathematically, TF-IDF is defined as <0 Weakness > 0 Opportunity R3
follows: Weakness ≤ 0 Threat R4
<0 Weakness <0 Weakness ≤ 0 Weakness R5
TFIDF = TF(t, d)⋅IDF(t, D) (2) Weakness > 0 Opportunity R6
>0 Strength ≤ 0 Threat R7
where t presents the term, f is the frequency, d is the individual docu­ Strength > 0 Opportunity R8
ment, and D is the total number of documents. TF and IDF are calculated Note: (1) S/W indicates Strength/Weakness; (2) △PerfFocal = Perff-thresholdf,
as follows: thresholdf is the focal company’s average performance of all attributes; (3)
△PerfCompetitor = Perfc-thresholdc, thresholdc is the competing company’s average
ft
TF(t, d) = ∑ (3) performance of all attributes.
t′∈d ft′

D thresholdc < 0), according to the connotation of the SWOT concept that
IDF(t, D) = log (4) “the weaknesses of competitor can become the opportunities of the
DF(t)
organisation” (Phadermrod et al., 2019), it is deemed an opportunity for
where DF(t) is the total number of documents in the corpus that contain the focal company if its performance outperforms that of the competing
at least one term. company (i.e., Perff > Perfc), since such a situation can create opportu­
In conjunction with our work, we applied the algorithm TF-IDF to nities for the focal company to maintain or gain a competitive advantage
extract feature words and their TF-IDF values from online reviews after in the market; otherwise (i.e., Perff ≤ Perfc), it is a threat to the focal
completing word splitting, de-punctuation, and stop word removal. company because its “internal strength” attribute faces challenges from
Then, the feature words were matched with words in the attribute the competing company, and thus the focal company should improve
lexicon. Finally, according to Eq. (5), where Qj is the number of feature this attribute in a timely fashion to avoid possible losses. More rules can
words of attribute j, we accumulated the TF-IDF values to obtain the be found in Table 1.
attribute importance.
3.3.2. Dynamic SWOT construction
To capture changing trends, especially when they are subtle, SWOT
Qj
∑
Impj = TFIDF jp (5)
p=1
analysis needs to be conducted from a dynamic perspective. Conse­
quently, we first specify the time period T and divide the dataset into T
Focal Competitor
3.3. SWOT analysis subsets. Then, by calculating PerfTj , PerfTj , thresholdFocal
T and
thresholdCompetitor
T
Focal
for each time period, ΔPerfTj Competitor
and ΔPerfTj are
3.3.1. SWOT factor identification obtained. The SWOT factors for each time period can be identified ac­
As Dyson (2004) stated, strengths and weaknesses are internal fac­ cording to Table 1.
tors that support and hinder the achievement of an organization’s Fig. 3 presents the dynamic SWOT construct, with ΔPerfTj Focal
and
mission, while opportunities and threats are external factors that enable Competitor
ΔPerfTj as the vertical and horizontal coordinates, respectively.
or disable the organization to achieve its mission, respectively. This view
illustrates the fact that strengths and weaknesses are the results of The color of the directed lines reflects attribute importance, with darker
relative comparisons between different factors within the organization, colors indicating higher importance and lighter colors indicating lower
while opportunities and threats are the results of comparisons with importance. Managers can understand the evolutionary trajectory of
competitors in reference to the external environment. their products in the external competitive environment and thus develop
To determine the internal strengths and weaknesses of the focal targeted preventive strategies in a timely manner. Throughout this
company, a threshold value threshold needs to be obtained. By process, attribute importance needs to be fully considered, and strategic
comparing the attribute performance with the threshold, we can then planning and resource allocation can be appropriately oriented to the
determine the internal strengths and weaknesses. Referring to Pha­
dermrod et al.’s (2019) work, we set the threshold to the average per­
formance of all attributes. Then, the SWOT elements of the target
company are further identified by comparing the internal strengths and
weaknesses of the two companies, as well as identifying their specific
performance differences. All the rules for identifying SWOT factors can
be seen in Table 1.
Specifically, when an attribute serves as an internal strength of both
companies (i.e., ΔPerfFocal = Perff-thresholdf > 0 and ΔPerfCompetitor =
Perfc-thresholdc > 0), it is considered a strength of the focal company if its
performance outperforms the competing company (i.e., Perff > Perfc);
otherwise (i.e., Perff ≤ Perfc), according to the connotation of the SWOT
concept that “the strengths of competitor become the threats of the
organisation” (Phadermrod et al., 2019), it is deemed a threat to the
focal company because the focal company faces potential risks from the
competing company. When an attribute is deemed an internal strength
of the focal company (i.e., ΔPerfFocal = Perff-thresholdf > 0) and internal
weakness of the competing company (i.e., ΔPerfCompetitor = Perfc- Fig. 3. Dynamic SWOT concept.

5
J. Wu et al. Journal of Business Research 171 (2024) 114378

product’s core competitive advantages/disadvantages. comparison results of Hotel_A vs. Hotel_B, and the comparison results of
Hotel_A vs. Hotel_C and Hotel_D vs. Hotel_E can be found in Web Ap­
4. Case study pendix D.
From Fig. 6, we know that two five-star hotels (Hotel_A vs. Hotel_B)
4.1. Data collection share the same dominant attributes (i.e., Location, Surrounding, and
Transport). Presumably, such a situation might be because these two
The proposed methodology enables SWOT analysis to be carried out hotels are located in similar locations and have similar surroundings
in many scenarios, such as different brands and products/services. In according to Fig. 5(a). In contrast, their disadvantageous attributes are
this paper, we choose online review data from hotel industries to not exactly the same. Hotel_A’s Facility, Parking, and Sleep attributes
implement SWOT analysis. Specifically, Ctrip.com is a leading online perform worst, while Hotel_B performs badly in these attributes.
travel agency in China. Consumers who have experienced hospitality Regarding other attributes, the two hotels perform differently, which
services can freely post online comments on Ctrip, and Fig. 4 is a data might imply that they have different competitive opportunities or
sample. Many scholars (Hou et al., 2019; Wu & Zhao, 2023) have used threats. For example, Hotel_A’s Value, Cleanliness, and Room attributes
online reviews on Ctrip to conduct research. Accordingly, we also collect perform better than Hotel_B’s, which might imply that these attributes
online reviews from this platform. are Hotel_A’s competitive advantages; however, Hotel_A’s Service
To validate the applicability of the proposed framework in different attribute performs slightly worse than Hotel_B’s, which may be the
types of hotels, we considered a variety of hotels with different char­ competitive disadvantage of Hotel_A.
acteristics, including hotel star, location, etc. Specifically, two focal
hotels from different locations in Beijing were selected; one is a five-star 4.3. Results of attribute importance calculation
hotel (named Hotel_A), and the other is a three-star hotel (named
Hotel_D). Applying the method in Section 3.2, we obtained the attribute
Since this study constructs SWOT for competitive analysis, it is importance. Meanwhile, to gain insight into the characteristics of
critical to identify potential and reasonable competitors of the focal different attributes, the top 5 feature words of each attribute were listed,
hotel. Regarding Hotel_A, we considered several factors, such as loca­ as shown in Table 3.
tion, hotel star, parking availability, and year built, and identified On the one hand, attribute importance can help managers reveal how
another five-star hotel (named Hotel_B) as the competing hotel. Notably, much consumers value different service aspects as a whole. For example,
many studies have uniformly classified five-star and four-star hotels as from the table, we know that the attributes that consumers mentioned
luxury hotels (Wu & Zhao, 2023), as they usually provide quality ser­ frequently were Location and Service. According to salience theory (Oh,
vices. In this view, four-star hotels may also compete with five-star 2001), this reflects that consumers pay great attention to these attri­
hotels. Therefore considering factors such as location and parking butes, which is basically consistent with Chaves et al.’s (2012) findings.
availability, we identified a four-star hotel (named Hotel_C) as another In contrast, attributes such as Parking and Bathroom are less important
potential competitor of Hotel_A. More details about the competitor to consumers because they are mentioned less frequently.
identification can be found in Web Appendix B. Fig. 5 (a) illustrates the On the other hand, feature words under each attribute reflect the
locations of the above hotels, and both Hotel_B and Hotel_C are no >500 focus of consumers on different aspects of the attribute. Table 3 shows
m away from Hotel_A. that the feature words are mostly nouns or adjectives that describe how
Regarding Hotel_D, we also considered location, year built, price, customers feel about the specific attributes, based on which practi­
etc., to identify its potential competitor. Eventually, another three-star tioners could develop targeted service improvement strategies. Taking
hotel (named Hotel_E) was identified. Hotel_D and Hotel_E are adja­ the Service attribute as an example, from the table we know that cus­
cent to each other and share a nearby parking lot, as shown in Fig. 5 (b). tomers frequently mentioned “Check-in”, “Front desk” and “Enthusi­
More details of these two hotels can be found in Web Appendix B. astic”; hence, managers should value the service of their front desk and
In summary, we identified the potential competitors of Hotel_A and ensure enthusiastic and thoughtful service details to satisfy their cus­
Hotel_D separately, and subsequently conducted three groups of SWOT tomers. Another notable example is concerned with the Restaurant
analyses, namely (1) Hotel_A vs. Hotel_B, (2) Hotel_A vs. Hotel_C, and attribute; Table 3 shows that customers’ hot words about this attribute
(3) Hotel_D vs. Hotel_E. Eventually, we crawled online reviews of the include “Breakfast”, “Children”, “Fruit”, etc. Therefore, managers may
above five hotels from Ctrip. The data covered from January 2019 to need to consider paying further attention to their breakfast and
December 2022, and their descriptive statistics are shown in Table 2. providing child-friendly dishes. More details can be found in the table,
and we will further apply the results in SWOT construction later.
4.2. Results of attribute performance calculation
4.4. Identification of SWOT factors and result evaluation
Before calculating attribute performance, 12 hotel attributes were
identified from the literature (in Web Appendix C), including Location, 4.4.1. Identification of SWOT factors
Transport, Value, etc. Then, we obtained the attribute performance of As mentioned above, we take Hotel_A and Hotel_D as the focal hotels
the five hotels utilizing the algorithm described in Fig. 2. Seeking an separately and further conduct SWOT analysis by considering their
initial understanding of the competitive position of the two focal hotels competitors’ attribute performance. Subsequently, utilizing the pro­
(Hotel_A and Hotel_D) in the market, we first compare their attribute posed methodology, we conducted three sets of SWOT analyses in turn,
performance with that of their competitors. Fig. 6 presents the namely (1) Hotel_A vs. Hotel_B, (2) Hotel_A vs. Hotel_C, and (3) Hotel_D

Fig. 4. Data Sample.

6
J. Wu et al. Journal of Business Research 171 (2024) 114378

Fig. 5. Hotel location illustration.

Table 2
Descriptive statistics.
Hotel Rating distribution Number of reviews Average rating

1 2 3 4 5

Hotel_A (Five-star) 1.78% 1.45% 4.43% 17.67% 74.68% 3661 4.71

Hotel_B (Five-star) 3.82% 2.32% 5.45% 14.56% 73.85% 2589 4.60
Hotel_C (Four-star) 0.58% 0.62% 2.50% 13.32% 82.98% 2755 4.84
Hotel_D (Three-star) 1.12% 0.56% 2.59% 7.76% 87.96% 889 4.84
Hotel_E (Three-star) 2.08% 2.08% 3.02% 5.63% 87.19% 1686 4.77

vs. Hotel_E. Restaurant, and Sleep quality attributes. Since both hotels perform badly
in these aspects, but Hotel_A is slightly better than Hotel_B, these are
4.4.1.1. Hotel_A vs. Hotel_B. Table 4 presents the identification results opportunities that Hotel_A can seize and achieve improvement. For
of SWOT factors, where Hotel_A (five-star) is the focal hotel, and Hotel_B example, regarding the Restaurant attributes, Hotel_A’s managers are
(five-star) is the competing hotel. From the table, we know that encouraged to pay more attention to their breakfast and provide child-
Hotel_A’s strengths include Location, Room, Cleanliness, Surrounding, friendly dishes to satisfy their customers, since in Section 4.4 we
and Value. Among them, Location, Room, and Cleanliness are Hotel_A’s found that consumers mention and are concerned about these aspects.
core strengths since consumers value these attributes very much. Furthermore, we also noticed that Service and Transport attributes are
Furthermore, the table also reveals potential improvement oppor­ Hotel_A’s threats. Regarding these two attributes, managers should
tunities for Hotel_A when competing with Hotel_B, such as Facility, prioritize the Service attribute, because this attribute performs badly (i.
e., only 0.742) and is more important than the Transport attribute.

7
J. Wu et al. Journal of Business Research 171 (2024) 114378

Fig. 6. Attribute performance comparison of Hotel_A vs. Hotel_B.

Table 3
Attribute importance and Top 5 feature words.
Attribute Top 5 feature words Imp

1 2 3 4 5

Location Not bad Location Experience Great Very close 2.537

Service Check-in Front desk Enthusiastic Service attitude Thoughtful 2.442
Facility Complete Luxury Design Air conditioner Old 1.232
Transport Convenient Traveling Business trip Time Subway 0.998
Restaurant Breakfast Children Fruit Afternoon tea Dinner 0.953
Room Comfortable Decoration Suite Room type Spacious 0.884
Cleanliness Clean Environment Hygiene Epidemic Smell 0.700
Surrounding Style Walking Nearby Mall Walking street 0.649
Value Value for money Free Upgrade Worth Price 0.470
Sleep quality Quiet Evening Next door Soundproof King-size bed 0.223
Bathroom Warmth Accessories Toilet Hot water Bathtub 0.195
Parking Park Parking lot Parking Spaces Garage Parking Fee 0.043

Table 4
Identification results of SWOT factors (Hotel_A vs. Hotel_B).
Attribute Hotel_A (focal hotel) Hotel_B (competing hotel) PerfA - PerfB SWOT factor Importance ranking

Perf ΔPerf S/W Perf ΔPerf S/W

Location 0.840 0.113 Strength 0.815 0.133 Strength 0.025 Strength 1

Room 0.743 0.016 Strength 0.738 0.056 Strength 0.005 Strength 2
Cleanliness 0.802 0.074 Strength 0.774 0.091 Strength 0.028 Strength 3
Surrounding 0.827 0.099 Strength 0.789 0.107 Strength 0.037 Strength 4
Value 0.806 0.078 Strength 0.707 0.025 Strength 0.099 Strength 5
Facility 0.608 − 0.119 Weakness 0.501 − 0.181 Weakness 0.107 Opportunity 1
Restaurant 0.681 − 0.046 Weakness 0.665 − 0.017 Weakness 0.016 Opportunity 2
Sleep quality 0.596 − 0.131 Weakness 0.582 − 0.100 Weakness 0.014 Opportunity 3
Bathroom 0.666 − 0.062 Weakness 0.519 − 0.164 Weakness 0.147 Opportunity 4
Parking 0.604 − 0.124 Weakness 0.510 − 0.172 Weakness 0.094 Opportunity 5
Service 0.742 0.014 Strength 0.745 0.063 Strength − 0.003 Threat 1
Transport 0.815 0.088 Strength 0.840 0.158 Strength − 0.025 Threat 2
threshold 0.727 0.682

Note: 1. “Perf” indicates attribute performance; 2. “ΔPerf” indicates the difference between attribute performance and average performance (threshold); 3. “PerfA –
PerfB” indicates the difference between the two hotels.

4.4.1.2. Hotel_A vs. Hotel_C. Table 5 shows the identification results of customers (e.g., adding acoustic glass, referring to Table 3 in Section
SWOT factors, where Hotel_A (five-star) is the focal hotel, and Hotel_C 4.4).
(four-star) is the competing hotel. From the table we know that when Further, we also found potential opportunity attributes of Hotel_A,
competing with Hotel_C, Hotel_A’s strength is its surrounding; Hotel_A’s such as Room, Bathroom, and Parking. Furthermore, multiple threats
weakness is its Sleep quality attribute; according to the map in Fig. 5 (a), that Hotel_A faces can be found in Table 5, such as Location, Service, and
this might be because Hotel_A is closer to the street and is relatively Facility. Considering the fact that the Location attribute of Hotel_A
noisier, which influence customers’ sleep quality. In this regard, Hotel_A performs well (i.e., 0.840) and is not changeable, Hotel_A should first
should take effective measures to improve the sleeping environment of pay more attention to its Service attribute; as per the hot feature words

8
J. Wu et al. Journal of Business Research 171 (2024) 114378

Table 5
Identification results of SWOT factors (Hotel_A vs. Hotel_C).
Attribute Hotel_A (focal hotel) Hotel_C (competing hotel) PerfA-PerfC SWOT factor Importance ranking

Perf ΔPerf S/W Perf ΔPerf S/W

Surrounding 0.827 0.084 Strength 0.769 0.028 Strength 0.058 Strength 1

Sleep quality 0.596 − 0.146 Weakness 0.654 − 0.087 Weakness − 0.058 Weakness 1
Room 0.743 0.001 Strength 0.669 − 0.072 Weakness 0.075 Opportunity 1
Bathroom 0.666 − 0.077 Weakness 0.566 − 0.175 Weakness 0.100 Opportunity 2
Parking 0.604 − 0.139 Weakness 0.369 − 0.372 Weakness 0.235 Opportunity 3
Location 0.840 0.098 Strength 0.854 0.113 Strength − 0.013 Threat 1
Service 0.742 − 0.001 Strength 0.831 0.090 Strength − 0.089 Threat 2
Facility 0.608 − 0.134 Weakness 0.807 0.066 Strength − 0.199 Threat 3
Transport 0.815 0.073 Strength 0.861 0.120 Strength − 0.046 Threat 4
Restaurant 0.681 − 0.061 Weakness 0.831 0.090 Strength − 0.150 Threat 5
Cleanliness 0.802 0.059 Strength 0.838 0.097 Strength − 0.037 Threat 6
Value 0.806 0.063 Strength 0.843 0.102 Strength − 0.038 Threat 7
threshold 0.727 0.741

in Section 4.4, Hotel_A is suggested to attach importance to its front-desk 4.4.2. Result evaluation
service (especially for check-in), to ensure that the services provided are In order to validate the proposed method, it is necessary to compare
“enthusiastic” and “thoughtful”. Moreover, the Facility attribute is also a it with existing methods. Phadermrod et al. (2019) proposed an IPA-
major threat for Hotel_A, which might be because Hotel_C was renovated based SWOT analysis method using traditional survey data, which has
more recently and thus provided better facilities. In this respect, Hotel_A received great attention in the last two years and is also a highly cited
should update its facilities in a timely manner, to avoid such threats from paper until December 2022. Regarding online review-based SWOT
becoming a weakness. research, although some studies have conducted SWOT analysis by
mining online reviews, they usually identify SWOT elements manually
4.4.1.3. Hotel_D vs. Hotel_E. Table 6 presents the identification results (Shah et al., 2021; Srinivas & Rajendran, 2019). Cheng et al.’s (2021)
of SWOT factors of two three-star hotels, where Hotel_D is the focal work is one of the few studies that has defined quantitative rules to
hotel, and Hotel_E is the competing hotel. From the table, Hotel_D’s identify SWOT factors. Therefore, the methods of Phadermrod et al.
strengths include Location, Transport, Room, etc. Considering the un­ (2019) and Cheng et al. (2021) could be regarded as two state-of-the-art
changeable location and transport, it is recommended for Hotel_D to works related to our study.
prioritize Room and Cleanliness attributes, since they are of high Regarding the specific rules, Phadermrod et al.’s (2019) study first
importance. If managers would like to strengthen the performance of constructed Importance-Performance Analysis (IPA) and then identified
Room, as per Table 3 in Section 4.4, comfortable decoration is a good SWOT factors by comparing the quadrants where the attributes were
place to start as customers mentioned this aspect a lot; moreover, located. Similarly, Cheng et al. (2021) defined how to identify SWOT
“spacious” room is another hot topic of discussion, and thus Hotel_D may factors of a target product from the perspective of product generation
consider adjusting the layout of the room to achieve the “room looks evolution. In other words, when the consumer satisfaction of an attri­
big” effect. bute of both product generations (i.e., Generation 1 and Generation 2) is
Importantly, Table 6 also reveals critical opportunities of Hotel_D (e. positive, the attribute is a strength; when an attribute of both product
g., Service, Facility, Restaurant attributes). Among these opportunities, generations is negative, the attribute is a weakness. When identifying
Service is of the greatest importance; as can be seen from the table, the the Opportunity or Threat factors, for an attribute, if the focal company
difference in Service performance between the two hotels is small (0.804 performs positively and the competitor performs negatively, the attri­
vs. 0.794); hence it is a great chance for Hotel_D to strengthen its Service bute is an opportunity; conversely, the attribute is a threat. Detailed
performance to get a competitive position. Further, the Facility attribute descriptions of these two methods can be found in Web Appendix E.
is also a good opportunity for Hotel_D to outperform Hotel_E; specif­ Taking the competition of Hotel_A and Hotel_B as an example,
ically, Hotel_D is suggested to regularly check the facilities in the rooms Table 7 presents the comparison results of different methods. From the
for deterioration, especially the air conditioners, as this is mentioned table we know that the identification results of the three methods are the
frequently by consumers. same for multiple attributes, such as Location, Surrounding, and Room
attributes. However, the results are not completely consistent with each

Table 6
Identification results of SWOT factors (Hotel_D vs. Hotel_E).
Attribute Hotel_D (focal hotel) Hotel_E (competing hotel) PerfD-PerfE SWOT factor Importance ranking

Perf ΔPerf S/W Perf ΔPerf S/W

Location 0.817 0.007 Strength 0.813 0.042 Strength 0.004 Strength 1

Transport 0.889 0.080 Strength 0.870 0.099 Strength 0.019 Strength 2
Room 0.847 0.037 Strength 0.808 0.037 Strength 0.039 Strength 3
Cleanliness 0.864 0.055 Strength 0.827 0.056 Strength 0.037 Strength 4
Surrounding 0.897 0.087 Strength 0.854 0.083 Strength 0.042 Strength 5
Value 0.855 0.045 Strength 0.846 0.075 Strength 0.008 Strength 6
Parking 0.953 0.143 Strength 0.882 0.111 Strength 0.070 Strength 7
Service 0.804 − 0.005 Weakness 0.794 0.023 Strength 0.011 Opportunity 1
Facility 0.787 − 0.023 Weakness 0.722 − 0.049 Weakness 0.065 Opportunity 2
Restaurant 0.820 0.011 Strength 0.751 − 0.020 Weakness 0.069 Opportunity 3
Sleep quality 0.631 − 0.179 Weakness 0.607 − 0.164 Weakness 0.024 Opportunity 4
Bathroom 0.552 − 0.258 Weakness 0.478 − 0.293 Weakness 0.074 Opportunity 5
threshold 0.810 0.771

9
J. Wu et al. Journal of Business Research 171 (2024) 114378

Table 7
Comparison to existing methods (Hotel_A vs. Hotel_B).
Attribute PerfA (focal hotel) PerfB (competing hotel) Phadermrod et al. (2019) Cheng et al. (2021) This study

Location 0.840 0.815 Strength Strength Strength

Surrounding 0.827 0.789 Strength Strength Strength
Value 0.806 0.707 Strength Strength Strength
Room 0.743 0.738 Strength Strength Strength
Cleanliness 0.802 0.774 Strength Strength Strength
Service 0.742 0.745 Strength Strength Threat
Transport 0.815 0.840 Strength Strength Threat
Restaurant 0.681 0.665 Weakness Strength Opportunity
Parking 0.604 0.510 Weakness Strength Opportunity
Facility 0.608 0.501 Weakness Strength Opportunity
Bathroom 0.666 0.519 Weakness Strength Opportunity
Sleep quality 0.596 0.582 Weakness Strength Opportunity

other. Specifically, our method identifies Service and Transport attri­ attribute to be a strength when it is positive sentiment and a weakness
butes as Threat factors, but Phadermrod et al. (2019) classify these two when it is negative sentiment. As such, the method cannot cope well
attributes as Strength factors. Combined with the attribute performance, with skewed sentiment distributions (e.g., skewed positive distribution
Hotel_A’s Service attribute did not perform well and was slightly weaker or skewed negative distribution), which in turn prevents the effective
than that of Hotel_B (i.e., 0.742 vs. 0.745). In this regard, it is difficult to identification of subtle opportunities and threats. Therefore, from these
say that the service performance of Hotel_A is a solid strength over perspectives, our identification results might be more reasonable.
Hotel_B; because of the slightly poor service performance, to some
extent, this is a potential threat that Hotel_B poses to Hotel_A. Essen­ 4.5. SWOT analysis from a dynamic perspective
tially, Phadermrod et al. (2019) only consider the comparison of product
strengths and weaknesses when identifying SWOT elements, without Utilizing the proposed method in Section 3.3, we divided the data
considering the actual performance differences between product attri­ into three periods on a yearly basis (i.e., T1, T2, and T3), and the
butes. Therefore, it leads to the inability to effectively identify potential attribute performance of each period can be seen in Web Appendix F.
threats. A similar assertion also applies to other attributes. Furthermore, we identified the SWOT factors of Hotel_A and Hotel_D in
Regarding Cheng et al.’s (2021) results, from the table we know that different periods, and the detailed data can be found in Web Appendices
it classifies all attributes as Strength factors. Except for the existing issue G–I. Figs. 7–9 present the construct results of dynamic SWOT.
similar to Phadermrod et al.’s (2019) work, Cheng et al.’s (2021)
method also faces another problem, namely that they considered an

Fig. 7. Dynamic SWOT (Hotel_A vs. Hotel_B).

10
J. Wu et al. Journal of Business Research 171 (2024) 114378

Fig. 8. Dynamic SWOT (Hotel_A vs. Hotel_C).

Fig. 9. Dynamic SWOT (Hotel_D vs. Hotel_E).

11
J. Wu et al. Journal of Business Research 171 (2024) 114378

4.5.1. Hotel_A (focal hotel) vs. Hotel_B (competing hotel) been improved (i.e., 0.858 → 0.833 → 0.854), however, the perfor­
Fig. 7 shows the dynamic SWOT analysis of Hotel_A vs. Hotel_B. As mance of Hotel_E has been strengthened (i.e., 0.809 → 0.799 → 0.822).
stated in Section 3.3, the arrow direction in the figure indicates the In this regard, if Hotel_D wants to maintain its competitive strength, then
temporal order. The color reflects the factor importance; a darker color it must keep improving its service performance; otherwise, the im­
indicates higher importance, and a lighter color indicates lower provements of the competing hotel would also weaken Hotel_D’s leading
importance. The shape of an attribute point reflects the type of SWOT aspects.
element to which it belongs. In general, when strengths turn into weaknesses or opportunities
From the figure, we know that almost all attributes are changing in turn into threats, making effective improvements often comes at a sig­
the competitive situation, and some attributes even show a shift in the nificant cost. This is where dynamic SWOT analysis comes in. With
SWOT factor category. For example, Service was Hotel_A’s Strength dynamic SWOT analysis, on the one hand, managers can understand
factor at T1 and T2 but began to change to a threat at T3; as per the which attributes are deteriorating. When an attribute tends to turn into a
attribute performance of Hotel_A and Hotel_B, we know that this situ­ weakness or threat, preventive measures can be developed and imple­
ation is because the competing hotel’s performance has been improving mented to reverse the situation. On the other hand, dynamic SWOT
(i.e., 0.607 → 0.769 → 0.772) while the focal hotel’s performance has analysis also allows managers to understand which attributes are
decreased (i.e., 0.756 → 0.747 → 0.716); therefore, the manager of this changing to a favorable position. For those attributes that are about to
five-star hotel should continue to improve its service performance to turn into opportunities, organizations can seize the opportunity to turn
avoid falling into “not to advance is to go back”. In addition, some at­ these potential capabilities into actual output as soon as possible.
tributes’ categories did not change, but their evolving trends still require Especially for those important attributes, internal resources can be
managers’ attention. For instance, Facility and Bathroom attributes are appropriately tilted toward them to ensure efficient conversion of
Hotel_A’s opportunities at all periods; however, Hotel_A did not seize investments.
these opportunities and turn them into solid strengths; a consequence of
wasting such opportunities is that they would turn into threats or 5. Conclusion and discussion
weaknesses (e.g., Restaurant and Sleep quality). In this process,
considering the importance of different factors contributes to more In this study, an innovative framework using mixed methods for
efficient resource allocation and strategy development. Specifically, conducting SWOT analysis by leveraging online reviews is proposed to
although there are several opportunity attributes (e.g., Bathroom, Fa­ address the problems of traditional SWOT analysis. First, quantitative
cility, and Parking), the core opportunities lie in the Facility attribute rules for identifying SWOT factors are developed, and the comparison
since it is with the darker color in Fig. 7; thus, managers can prioritize results with two advanced methods (i.e., traditional SWOT analysis and
this area. UGC-based SWOT analysis) are used to further illustrate the superiority
of the proposed method. Therefore, the proposed methodology can help
4.5.2. Hotel_A (focal hotel) vs. Hotel_C (competing hotel) managers or analysts effectively understand their products’ market
Fig. 8 presents the dynamic SWOT analysis of Hotel_A vs. Hotel_C. It competitive situation (especially when the performance differences be­
can be seen that although these two hotels are in similar geographical tween two competing products are subtle), with a lower expenditure of
locations, Hotel_A does not perform as well as Hotel_C in terms of both cost and time. Further, we also propose an effective and objective
Location and Transport attributes (i.e., Threat factor at different time TF-IDF-based approach for measuring the importance of SWOT factors.
points). As per expectation disconfirmation theory (Wu & Zhao, 2023), Different from existing studies (Büyüközkan & Ilıcak, 2019; Sevkli et al.,
customer expectation influences products’ usage experience, thus 2012) that assess factor importance based on individual expertise of
impacting customer satisfaction. The target consumers of five-star hotels experts or managers, this study adopts the perspective of customers to
(e.g., Hotel_A) probably have higher expectations than those of four-star help managers prioritize the core advantages and disadvantages of
hotels (e.g., Hotel_C), which leads to Hotel_A’s weaker performance in products, and thus efficiently allocate their limited resources. Further­
Location and Transport attributes. Similar speculations may also apply more, a dynamic SWOT construct was developed to model the evolving
to Service and Value attributes, since these two attributes are Threat trajectory of the SWOT factors. For instance, in the competition with
factors of Hotel_A at all time points. Overall, many of Hotel_A’s attri­ Hotel_B, Hotel_A’s Service attribute gradually changed from a Strength
butes do not compete well with Hotel_C, and many are threats or factor into a Threat factor. This makes it possible for SWOT to effectively
weaknesses. On the one hand, such a situation may be due to the cus­ respond to the changing market environment, thus enabling businesses
tomers’ higher expectations. On the other hand, it could also be caused to develop timely initiatives.
by the fact that Hotel_A is older compared to Hotel_C; as per the online
details in Ctrip, the opening/renovation dates of the two hotels were 5.1. Representativeness of online reviews
2008 and 2019, respectively. Taking certain tangible attributes as ex­
amples, Facility and Restaurant attributes are threats to Hotel_A at all In this study, a method for conducting SWOT analysis using online
time points, indicating that these aspects of Hotel_A performed slightly reviews is proposed. Whether such data are sufficiently representative
worse than Hotel_C. Therefore, managers of Hotel_A should update their may come under question. Therefore, we present the following
facilities timely, and especially their target consumers may have higher discussion.
expectations for five-star hotels. With the rapid development of information technology, an
increasing number of consumers are purchasing goods through online
4.5.3. Hotel_D (focal hotel) vs. Hotel_E (competing hotel) platforms. According to the statistics, there are now 5.29 billion internet
Fig. 9 presents the dynamic SWOT analysis of two three-star hotels, users worldwide (Internet Live Stats, 2014). Some studies indicate that
where Hotel_D is the focal hotel, and Hotel_E is the competing hotel. >60% of consumers have posted online reviews about a product/service
Similar to previous dynamic SWOT analysis, almost all attributes’ SWOT (Bi et al., 2019c). Given the breadth and influence of such online re­
factor categories are changing with different time points. Notably, views, many online review-based business analytics studies have sought
multiple attributes of Hotel_D are opportunities at different time points to identify valuable business opportunities (Bi et al., 2020; Costa Filho
when competing with Hotel_E, yet Hotel_D has not converted them into et al., 2023; Garner et al., 2022; Xu, 2020), such as competitiveness
solid competitive strengths such as Restaurant and Bathroom attributes. analysis (Gao et al., 2018), brand analysis (Bi et al., 2019c), and product
Moreover, several attributes of Hotel_D have changed from Strength sales forecasting (Fan et al., 2017). Regarding the research fields, the
factor to Opportunity factor such as Room; combined with attribute extant online review-based business analytics studies appear in the
performance change we know that while Hotel_D’s performance has not manufacturing industry (Yang et al., 2023b), tourism and hospitality

12
J. Wu et al. Journal of Business Research 171 (2024) 114378

industry (Wu & Zhao, 2023), etc. In this regard, online reviews should be single platform, while data from multiple platforms can reflect the
representative enough for SWOT analysis. competitive position more comprehensively. Second, this study does not
On the other hand, we use the hotel industry as our research object, attempt to replace traditional survey or interview-based methods. Some
and different types of hotels are considered to test our framework. offline products may not have (enough) online reviews, thus making it
Regarding the proposed methodology, each step is independent of the difficult to directly apply the proposed method. Hence, more work (e.g.,
data field (e.g., PaddlePaddle-based sentiment analysis and TF-IDF-based utilizing surveys or interviews to collect customer comments) needs to
importance calculation), which basically guarantees the framework’s be done, based on which SWOT analysis could be conducted. Further­
generalizability. Moreover, online reviews reflect consumer feelings and more, in future work, it is encouraged to explore how to synthesize in­
acceptance of products/services (Yang et al., 2023b), which is true for ternal company data (e.g., resource and investment data) and online
different types of products/services. Thus, online hotel reviews can be reviews to conduct SWOT analysis using the proposed methodology,
speculated to be sufficiently representative for conducting SWOT which would be more comprehensive in helping companies carry out
analysis. competitive analysis. Besides, given its benefits and context suitableness
in our study, the TF-IDF algorithm is adopted to measure attribute
5.2. Implications importance, but it only considers lexical information and captures no
semantics of online reviews. In this regard, a promising research
The theoretical significance of this study mainly lies in the perspective might be the development of an improved TF-IDF that in­
advancement of SWOT construction and analysis methods. First, tradi­ corporates semantical information (e.g., integrating word embeddings
tional SWOT factor identification studies typically rely on individuals (e. and the TF-IDF algorithm).
g., experts or managers), which inevitably suffers from subjectivity and
biased results (Hill & Westbrook, 1997; Smith, 2003). While some CRediT authorship contribution statement
scholars have developed automatic rules to identify SWOT elements (e.
g., Cheng et al., 2021) by exploring UGC, they ignore the specific per­ Jie Wu: Writing – original draft, Methodology, Conceptualization,
formance differences between two competing products, which may Data curation, Investigation. Narisa Zhao: Supervision, Conceptuali­
cause an inability to effectively observe subtle strategic opportunities zation, Methodology, Writing – review & editing. Tong Yang: Software,
and threats. In this paper, we address this issue by constructing quan­ Data curation, Formal analysis, Writing – original draft.
titative rules for SWOT factor identification, which provides future
research with a more effective and efficient SWOT analysis method
Declaration of Competing Interest
through the use of online reviews. Second, previous studies have mainly
focused on constructing SWOT, and little attention had been given to
The authors declare that they have no known competing financial
factor importance (Büyüközkan & Ilıcak, 2019; Coman & Ronen, 2009).
interests or personal relationships that could have appeared to influence
By utilizing the TF-IDF method to measure factor importance, the pro­
the work reported in this paper.
posed method enables SWOT analysis researchers to clarify the core
competitive advantages and disadvantages, especially in regard to
Data availability
resource constraints. Third, existing SWOT studies are typically con­
ducted from a static perspective (Mitra & Jenamani, 2020; Shah et al.,
Data will be made available on request.
2021), and thus cannot efficiently respond to the dynamic market
environment. In this regard, the proposed methodology expands the
Acknowledgements
SWOT literature by modeling the evolutionary path of SWOT elements.
In summary, due to the improvement in methodology, the original
The authors acknowledge the Center for Big Data & Intelligent
SWOT principle is still valid in the era of big data, which avoids the cost
Decision-Making of Dalian University of Technology for providing
and risk of changing strategic planning tools for organizations. To our
computing resources. Jie Wu and Tong Yang would like to thank China
knowledge, this should be the first comprehensive work that addresses
Scholarship Council for its scholarship support [202106060108,
all the abovementioned issues that have been identified in traditional
202106060118].
SWOT analysis.
Meanwhile, our study also presents some practical implications.
Appendix A. Supplementary data
First, the proposed framework provides an effective method for con­
ducting competitive analysis using online reviews. By tapping into a
Supplementary data to this article can be found online at https://doi.
more authentic customer voice, this approach enables managers to un­
org/10.1016/j.jbusres.2023.114378.
derstand the competitive position of their products and to identify their
core competitive advantages and disadvantages at a lower cost. Second,
References
dynamic SWOT analysis can help managers keep track of changes in the
market performance of their products. For attributes with deteriorating Bergen, M., & Peteraf, M. A. (2002). Competitor identification and competitor analysis: A
trends, preventive measures can be taken in time. For example, from broad-based managerial approach. Managerial & Decision Economics, 23(4–5),
Fig. 7 we know that Hotel_A’s attributes such as Service, Cleanliness, 157–169. https://doi.org/10.1002/mde.1059
Bi, J.-W., Liu, Y., & Fan, Z.-P. (2019a). Representing sentiment analysis results of online
and Room are continuously deteriorating (i.e., changing from a strength reviews using interval type-2 fuzzy numbers and its application to product ranking.
or opportunity to a threat), which requires timely intervention and Information Sciences, 504, 293–307. https://doi.org/10.1016/j.ins.2019.07.025
improvement from managers. A consideration of the importance of Bi, J.-W., Liu, Y., Fan, Z.-P., & Cambria, E. (2019b). Modelling customer satisfaction from
online reviews using ensemble neural network and effect-based Kano model.
factors (e.g., prioritizing improvements in the Service attribute) leads to
International Journal of Production Research, 57(22), 7068–7088. https://doi.org/
more effective execution of strategic analysis results. Third, we validate 10.1080/00207543.2019.1574989
the proposed methodology using online reviews of different types of Bi, J.-W., Liu, Y., Fan, Z.-P., & Zhang, J. (2019c). Wisdom of crowds: Conducting
hotels. The proposed framework, however, can also be beneficial for importance-performance analysis (IPA) through online reviews. Tourism
Management, 70, 460–478. https://doi.org/10.1016/j.tourman.2018.09.010
other products/services (e.g., restaurants and destinations). Bi, J.-W., Liu, Y., Fan, Z.-P., & Zhang, J. (2020). Exploring asymmetric effects of attribute
performance on customer satisfaction in the hotel industry. Tourism Management, 77,
5.3. Limitations and future work Article 104006. https://doi.org/10.1016/j.tourman.2019.104006
Büyüközkan, G., & Ilıcak, Ö. (2019). Integrated SWOT analysis with multiple preference
relations: Selection of strategic factors for social media. Kybernetes, 48(3), 451–470.
This study is not without limitations. First, we only used data from a https://doi.org/10.1108/K-12-2017-0512

13
J. Wu et al. Journal of Business Research 171 (2024) 114378

Chaves, M. S., Gomes, R., & Pedron, C. (2012). Analysing reviews in the web 2.0: Small Expert Systems with Applications, 39(1), 14–24. https://doi.org/10.1016/j.
and medium hotels in Portugal. Tourism Management, 33(5), 1286–1287. https://doi. eswa.2011.06.047
org/10.1016/j.tourman.2011.11.007 Shah, A. M., Yan, X., Tariq, S., & Shah, S. A. A. (2021). Tracking patients healthcare
Chen, J., Becken, S., & Stantic, B. (2022). Assessing destination satisfaction by social experiences during the COVID-19 outbreak: Topic modeling and sentiment analysis
media: An innovative approach using importance-performance analysis. Annals of of doctor reviews. Journal of Engineering Research (Kuwait), 9(3), 219–239. https://
Tourism Research, 93, Article 103371. https://doi.org/10.1016/j. doi.org/10.36909/jer.v9i3A.8703
annals.2022.103371 Smith, B. (2003). The effectiveness of marketing strategy making processes: A critical
Cheng, L. C., Chen, K., Lee, M. C., & Li, K. M. (2021). User-Defined SWOT analysis – A literature review and a research agenda. Journal of Targeting, Measurement and
change mining perspective on user-generated content. Information Processing and Analysis for Marketing, 11(3), 273–290. https://doi.org/10.1057/palgrave.
Management, 58(5), Article 102613. https://doi.org/10.1016/j.ipm.2021.102613 jt.5740083
Coman, A., & Ronen, B. (2009). Focused SWOT: Diagnosing critical strengths and Srinivas, S., & Rajendran, S. (2019). Topic-based knowledge mining of online student
weaknesses. International Journal of Production Research, 47(20), 5677–5689. https:// reviews for strategic planning in universities. Computers and Industrial Engineering,
doi.org/10.1080/00207540802146130 128, 974–984. https://doi.org/10.1016/j.cie.2018.06.034
Costa Filho, M., Nogueira Rafael, D., Barros, S. G., & L., & Mesquita, E.. (2023). Mind the Tan, A.-H. (1999). Text mining: The state of the art and the challenges. Proceedings of the
fake reviews! Protecting consumers from deception through persuasion knowledge Pakdd 1999 Workshop on Knowledge Disocovery from Advanced Databases, 8, 65–70.
acquisition. Journal of Business Research, 156, Article 113538. https://doi.org/ Tan, X., Cai, Y., Xu, J., Leung, H.-F., Chen, W., & Li, Q. (2020). Improving aspect-based
10.1016/j.jbusres.2022.113538 sentiment analysis via aligning aspect embedding. Neurocomputing, 383, 336–347.
Duari, S., & Bhatnagar, V. (2019). sCAKE: Semantic connectivity aware keyword https://doi.org/10.1016/j.neucom.2019.12.035
extraction. Information Sciences, 477, 100–117. https://doi.org/10.1016/j. Tirunillai, S., & Tellis, G. J. (2014). Mining marketing meaning from online chatter:
ins.2018.10.034 Strategic brand analysis of big data using latent dirichlet allocation. Journal of
Dyson, R. G. (2004). Strategic development and SWOT analysis at the University of Marketing Research, 51(4), 463–479. https://doi.org/10.1509/jmr.12.0106
Warwick. European Journal of Operational Research, 152(3), 631–640. https://doi. Vlados, C. (2019). On a correlative and evolutionary SWOT analysis. Journal of Strategy
org/10.1016/s0377-2217(03)00062-6 and Management, 12(3), 347–363. https://doi.org/10.1108/jsma-02-2019-0026
Fan, Z.-P.-P., Che, Y.-J.-J., & Chen, Z.-Y.-Y. (2017). Product sales forecasting using online Wu, H., & Yuan, N. (2018). An Improved TF-IDF algorithm based on word frequency
reviews and historical sales data: A method combining the Bass model and sentiment distribution information and category distribution information. In Proceedings of the
analysis. Journal of Business Research, 74, 90–100. https://doi.org/10.1016/j. 3rd International Conference on Intelligent Information Processing (pp. 211–215).
jbusres.2017.01.010 Wu, J., & Yang, T. (2023). Service attributes for sustainable rural tourism from online
Fehringer, D., Hohhof, B., & Johnson, T. (2006). State of the art: Competitive intelligence. comments: Tourist satisfaction perspective. Journal of Destination Marketing &
Competitive Intelligence Foundation. Accessed on: May 4, 2022. https://static.canal Management, 30(2), Article 100822. https://doi.org/10.1016/j.jdmm.2023.100822
blog.com/storagev1/vtech.canalblog.com/docs/f_060608_stateofart_sum.pdf. Wu, J., Zhang, J., & Zhao, N. (2023). How to boost e‑commerce for poverty alleviation?
Gao, S., Tang, O., Wang, H., & Yin, P. (2018). Identifying competitors through A perspective on competitiveness analysis using online reviews. Electronic Commerce
comparative relation mining of online reviews in the restaurant industry. Research, doi:10.1007/s10660-023-09708-2.
International Journal of Hospitality Management, 71, 19–32. https://doi.org/10.1016/ Wu, J., & Zhao, N. (2023). What consumer complaints should hoteliers prioritize?
j.ijhm.2017.09.004 Analysis of online reviews under different market segments. Journal of Hospitality
Garner, B., Thornton, C., Pawluk, A. L., Cortez, R. M., Johnston, W., & Ayala, C. (2022). Marketing & Management, 32(1), 1–28, doi:10.1080/19368623.2022.2119187.
Utilizing text-mining to explore consumer happiness within tourism destinations. Xiong, A., Liu, D., Tian, H., Liu, Z., Yu, P., & Kadoch, M. (2021). News keyword
Journal of Business Research, 139, 1366–1377. https://doi.org/10.1016/j. extraction algorithm based on semantic clustering and word graph model. Tsinghua
jbusres.2021.08.025 Science and Technology, 26(6), 886–893. https://doi.org/10.26599/
Ghazinoory, S., Abdi, M., & Azadegan-Mehr, M. (2011). SWOT methodology: A state-of- TST.2020.9010051
the-art review for the past, a framework for the future. Journal of Business Economics Xu, X. (2020). Examining an asymmetric effect between online customer reviews
and Management, 12(1), 24–48. https://doi.org/10.3846/16111699.2011.555358 emphasis and overall satisfaction determinants. Journal of Business Research, 106,
Guido, G. (1998). The dichotic theory of salience: A framework for assessing attention and 196–210. https://doi.org/10.1016/j.jbusres.2018.07.022
memory. ACR European Advances. Yang, T., Dang, Y., & Wu, J. (2023a). Dynamic perceived quality analysis using social
GÜRel, E. (2017). SWOT analysis: A theoretical review. Journal of International Social media data at macro-SWOT factor identification and micro-levels. Industrial
Research, 10(51), 994–1006. https://doi.org/10.17719/jisr.2017.1832 Management & Data Systems, 123(5), 1465–1495, doi:10.1108/IMDS-08-2022-0478.
Hill, T., & Westbrook, R. (1997). SWOT analysis: It’s time for a product recall. Long Range Yang, T., Dang, Y., & Wu, J. (2023b). How to prioritize perceived quality attributes from
Planning, 30(1), 46–52. https://doi.org/10.1016/s0024-6301(96)00095-7 consumers’ perspective? Analysis through social media data. Electronic Commerce
Hou, Z., Cui, F., Meng, Y., Lian, T., & Yu, C. (2019). Opinion mining from online travel Research, doi:10.1007/s10660-022-09652-7.
reviews: A comparative analysis of Chinese major OTAs using semantic association Yang, T., Wu, J., & Zhang, J. (2023c). Knowing how satisfied/dissatisfied is far from
analysis. Tourism Management, 74, 276–289. https://doi.org/10.1016/j. enough: A comprehensive customer satisfaction analysis framework based on hybrid
tourman.2019.03.009 text mining techniques. International Journal of Contemporary Hospitality Management,
Internet Live Stats (2014). Internet Users. Accessed on: May 3, 2022. https://www.intern doi:10.1108/IJCHM-10-2022-1319.
etlivestats.com/internet-users/. Yüksel, I., & Daǧdeviren, M. (2007). Using the analytic network process (ANP) in a
Jin, J., Ji, P., & Yan, S. (2017). Comparison of series products from customer online SWOT analysis - A case study for a textile firm. Information Sciences, 177(16),
concerns for competitive intelligence. Journal of Ambient Intelligence and Humanized 3364–3382. https://doi.org/10.1016/j.ins.2007.01.001
Computing, 10(3), 937–952. https://doi.org/10.1007/s12652-017-0635-9 Zhao, J., Guan, Z., & Sun, H. (2019). Riker: Mining rich keyword representations for
Kangas, J., Pesonen, M., Kurttila, M., & Kajanus, M. (2001). A’WOT: Integrating the AHP interpretable product question answering. Proceedings of the ACM SIGKDD
with SWOT analysis. In Proceedings of the International Symposium on the Analytic International Conference on Knowledge Discovery and Data Mining, 1389–1398. https://
Hierarchy Process, 2–4. doi:10.13033/isahp.y2001.012. doi.org/10.1145/3292500.3330985
Koch, A. J. (2000). SWOT does not need to be recalled: It needs to be enhanced. B Quest,
1(1), 1–14.
Jie Wu is a doctoral student at the Institute of Systems Engineering at Dalian University of
Lai, I. K. W., & Hitchcock, M. (2015). Importance–performance analysis in tourism: A
Technology. He is currently also a visiting student in the Department of Industrial Systems
framework for researchers. Tourism Management, 48, 242–267. https://doi.org/
Engineering and Management at National University of Singapore funded by China
10.1016/j.tourman.2014.11.008
Scholarship Council. His research interests include text mining and competitive intelli­
Lee, K., & Lin, S. (2008). A fuzzy quantified SWOT procedure for environmental
gence analysis. His research has been published in International Journal of Hospitality
evaluation of an international distribution center. Information Sciences, 178(2),
Management, Journal of Hospitality Marketing & Management, International Journal of
531–549. https://doi.org/10.1016/j.ins.2007.09.002
Contemporary Hospitality Management, among others.
Liu, Y., Jiang, C., & Zhao, H. (2019). Assessing product competitive advantages from the
perspective of customers by mining user-generated content on social media. Decision
Support Systems, 123, Article 113079. https://doi.org/10.1016/j.dss.2019.113079 Narisa Zhao is a professor at the Institute of Systems Engineering at Dalian University of
Ma, Y., Yu, D., Wu, T., & Wang, H. (2019). PaddlePaddle: An open-source deep learning Technology. He graduated in System engineering at Dalian University of Technology and
platform from industrial practice. Frontiers of Data and Domputing, 1(1), 105–115. holds a Ph.D. in System Engineering. His research is about data mining, sentiment analysis
https://doi.org/10.11871/jfdc.issn.2096.742X.2019.01.011 and complex network. His research has been published in International Journal of Hos­
Mitra, S., & Jenamani, M. (2020). OBIM: A computational model to estimate brand pitality Management, Journal of Hospitality Marketing & Management, IEEE Wireless
image from online consumer review. Journal of Business Research, 114(May 2019), Communications Letters, among others.
213–226. doi:10.1016/j.jbusres.2020.04.003.
Oh, H. (2001). Revisiting importance–performance analysis. Tourism Management, 22(6),
Tong Yang is a doctoral student at the Institute of Systems Engineering at Dalian Uni­
617–627. https://doi.org/10.1016/s0261-5177(01)00036-x
versity of Technology. He is currently also a visiting student in the Department of Infor­
Phadermrod, B., Crowder, R. M., & Wills, G. B. (2019). Importance-performance analysis
mation Systems and Analytics at National University of Singapore funded by China
based SWOT analysis. International Journal of Information Management, 44, 194–203.
Scholarship Council. His research interests include data mining and business intelligence.
https://doi.org/10.1016/j.ijinfomgt.2016.03.009
His research has been published in International Journal of Hospitality Management, In­
Rajendran, S., Srinivas, S., & Pagel, E. (2022). Mining voice of customers and employees
ternational Journal of Contemporary Hospitality Management, Journal of Destination
in insurance companies from online reviews: A text analytics approach.
Marketing & Management, among others.
Benchmarking, 30(1), 1–22. https://doi.org/10.1108/BIJ-12-2020-0650
Sevkli, M., Oztekin, A., Uysal, O., Torlak, G., Turkyilmaz, A., & Delen, D. (2012).
Development of a fuzzy ANP based SWOT analysis for the airline industry in Turkey.

14