Understanding retail quality ofsporting goods stores: a text

mining approachLuke Lunhua Mao

University of New Mexico, Albuquerque, New Mexico, USA

Abstract

Purpose – Sporting goods retailing is a significant sector within the sport industry with the total revenue ofthis sector reaching $52.2 billion in 2018. Beset with formidable competition, sporting goods stores arecompelled to augment theirmerchandise with service and improve retail quality. The purpose of this study is toinvestigate retail quality of sporting goods stores (RQSGS).Design/methodology/approach – Based on 27,793 online reviews of 1481 stores in the United States, thisstudy used Leximancer 4.0, a text mining software, to identify critical retail quality dimensions associatedwithsporting goods stores, and further explored the most salient dimensions among different levels of ratings.Findings – Customer service and store aspects are the two higher-order dimensions of RQSGS; holisticexperience, manager and staff are three themes under customer service, and product, B&M store and online–offline integration are three themes under store aspects. Furthermore, extreme reviews focusmore on customerservice, whereas lukewarm reviews focus more on store aspects.Practical implications – Knowledgeable staff, managers and online–offline integration are instrumental increating superior retail quality. Sporting goods stores should enhance hedonic and social values for consumersin order to ward off online competitions.Originality/value – This study explored retail quality dimensions that are pertinent to sporting goodsretailing utilizing text mining methods. This study to certain extent cross-validated the existing retailingliterature that is developed on alternative methods.

Keywords Retail quality, Sporting goods stores, e-Commerce, Leximancer, Text mining

Paper type Research paper

IntroductionSporting goods stores, coded as 451110 and defined as “business establishments primarilyengaging in retailing new sporting goods” by the North American Industry ClassificationSystem (NAICS), is an important sector within the sport industry (U.S. Census, 2019;Eschenfelder and Li, 2007). Major players in this sector include Dick’s Sporting Goods Inc.,Foot Locker, Academy Sports and Outdoors and Bass Pro Shops Inc. They retail bicycles,camping equipment, exercise equipment, athletic apparel, athletic footwear, sporting gunsand the like. There were 40,122 sporting goods stores in the United States and the totalrevenue of this sector reached $52.2 billion in 2018 (Hyland, 2018). Roughly speaking, thisindustrial sector is composed of three categories of stores: a) sporting goods superstores, suchasDick’s SportingGoods andAcademy Sports andOutdoors, which typically have a retailingspace of over 35,000 square feet and emphasize high volume sales; b) traditional sportinggoods stores, such as Big 5 Sporting Goods, which range in size from 5,000 to 20,000 squarefeet and typically carry a varied assortment of merchandise; and c) specialty stores, whichtypically carry a wide assortment of one specific product category, such as shoes, golf oroutdoor equipment. Examples include Golf Galaxy, Finish Line and lululemon athletica.

Primarily driven by the increased sports participation, health-conscious consumers and astrong rise in demand for athletic footwear from consumers aged 10 to 19, sporting goodsretailing was projected to have an annual growth rate of 1.9% for the following five years

Sport retailquality

This project was supported by the College of Education at the University of New Mexico. The viewsexpressed are those of the author and do not necessarily reflect the views of the funders.

The current issue and full text archive of this journal is available on Emerald Insight at:

https://www.emerald.com/insight/1464-6668.htm

Received 6 March 2020Revised 28 April 2020

20 May 2020Accepted 20 May 2020

International Journal of SportsMarketing and Sponsorship

© Emerald Publishing Limited1464-6668

DOI 10.1108/IJSMS-03-2020-0029

(Hyland, 2018). However, sporting goods retailing is highly competitive (Dick’s SportingGoods, 2018; Gale, 2018). Sporting goods stores compete not only with adjacent stores in theirown industry but also with online retailers (e.g. Amazon), mass merchandisers (e.g. Costco,Walmart), department stores (e.g. Kohl’s), direct channels of sports brands (e.g. Callaway,Nike), along with others (Gale, 2018). In fact,Walmart, with $9.8 billion revenue from sportinggoods sales, was the biggest retailer in themarket in 2015 (Statista, 2018). Themost recent USCensus Bureau Consumer Spending Report suggested that sporting goods, hobby, musicalinstrument and book stores is the only category that has not seen a positive month-over-month increase in 2018, and the total sales decreased by 3.4% in the first 10 months of 2019(U.S. Census, 2018). Based on the business statuses listed on yelp.com, about 15–22% ofsporting goods stores might have been closed in the past five years.

Sporting goods retailing is highly competitive and moderately concentrated (Gale, 2018;Hyland, 2018). The IBISWorld report suggests that the four major players of the industry,including Bass Pro, Dick’s Sporting Goods, Academy Sports and Outdoors and RecreationalEquipment have a total market share of 48.7%. About 93% of sporting goods stores employfewer than 20 people and cater to specialty niches and local consumers. As a result, most storesstill rely on the revenues from their brick and mortar (B&M) stores even in the era of rapidgrowth of e-commerce andmultichannel retailing (Chiu et al., 2011). For instance, 87.6% of salesin 2017 at Dick’s Sporting Goods and 78.3% of Cabela’s sales in 2016 occurred in their offlinechannels (Cabela’s, 2016; Dick’s Sporting Goods, 2018). It is reasonable to assume that theproportion of offline sales for independent local stores and smaller chain stores will be greater.

To effectively compete in the sporting goods market, retailers have extensivelyaugmented their products with additional services. In addition to such services asinstallation, warranty, credit cards, return and exchange, sporting goods retailers mayutilize other more idiosyncratic tools (Homburg et al., 2002). For instance, specialty runningshoe stores may augment their products with gait analyses; tennis pro-shop with stringingservices; bike shops with repair services; yoga equipment with training classes; and generalstores with loyalty programs. The demarcation between products and services has beenblurred. Service qualitymatters to sporting goods stores asmuch as to any pure service firms,because it not only elicits positive behavioral consequences and engenders consumer loyalty(Zeithaml et al., 1996) but also enhances retailing differentiation (Davis-Sramek et al., 2009;Homburg et al., 2002; Jiang and Zhang, 2011).

Retail quality is considered more complex than quality of pure service as retail stores sellproducts, together with services implicitly or explicitly (Dabholkar et al., 1996). Furthermore,it is generally found that retail quality dimensions are often contextualized: qualitydimensions and perceived importance of these dimensions tend to vary across differentindustries and different types of stores (Mehta et al., 2000). The quality dimensions associatedwith high fashion retailing (Gagliano and Hathcote, 1994) likely differ from those withgrocery retailing (Martinelli and Balboni, 2012); the quality dimensions of an online store(Yang et al., 2003) likely differ from those of a conventional B&Mstore (Dabholkar et al., 1996);the quality dimensions of service-intensive stores likely differ from those of goods-intensivestores (Mehta et al., 2000). From this reasoning, although we suspect sporting goods storesmay have some unique characteristics, we do not have a prior information or belief about howthey may differ.

Situating in the wider context of retailing quality (Dabholkar et al., 1996; Simmers andKeith, 2015), the purpose of this study is therefore to empirically examine the qualitydimensions of sporting goods retailing. Similar to the conventional content analysis approach(Simmers and Keith, 2015; Yang and Fang, 2004), this study draws insights about retailquality based on consumer self-generated online reviews, specifically a unique real-worlddataset released by Yelp on August 1, 2018 for their round 12 dataset challenge competition.

IJSMS

Our specific research questions include:What are the retail quality dimensions as reflected byonline reviews? How do reviews differ by ratings?

Yelp is a publicly traded company publishing crowd-sourced reviews for local businessesthrough yelp.com and their yelp mobile app. By the end of 2017, Yelp has 148 million reviewson their platform. Yelp is one of few online review companies that have built sophisticatedmachine learningmodels to filter fake reviews (Kamerer, 2014) and their filtration system hasbeen found reasonably effective by third-party researchers from University of Illinois atChicago and Google (Mukherjee et al., 2013). Additionally, it is ruled by the CaliforniaSupreme Court that businesses cannot force Yelp to remove a review (Thanawala, 2018), thusreviews from dissatisfied consumers are kept intact. It can be argued that the reviews haveprovided a more comprehensive overview of a business. Therefore, this dataset has enoughcredibility to provide insights about sporting goods business.

This research makes several contributions. First, to our best knowledge, this research isthe first attempt to examine retail quality of sporting goods stores. Although service qualityis a prominent research area in sport management, the existing literature has primarilyfocused on spectator sports and fitness clubs. Sporting goods retailing sector, as a majorsector of sports industry, warrants an examination. This study provides an early descriptionof retail quality in the context of sporting goods retailing. Second, this study to certain extentcross-validates the existing retailing literature that is developed on alternative methods.Third, this research extends service quality literature by considering the interaction betweenonline–offline channels as the online channels significantly shape consumers’ perceptions ofoffline service encounters. By doing so, this research updates the theory of retail servicequality. Fourth, this research utilizes the text miningmethodology and a large data set of self-generated online reviews. Specifically, a probabilistic topicmodelingmethod, as implementedby Leximancer 4.0, is used to uncover the underlying structure of the reviews, which providesa more objective calibration to the existing literature. Finally, our substantive results helpowners of sporting goods stores understand the important dimensions within sportsretailing.

Relevant literatureService qualityService quality is one of the most researched topics and the literature is rich. The focus of thisline of research has been primarily on the definition of quality pertaining to services incomparison with quality of physical goods. As service is thought intangible, inseparable,heterogeneous and perishable, it cannot be easily seen, felt or touched; thus the experience ofit is elusive and evaluation of it subjective (Kotler and Keller, 2006; Parasuraman et al., 1985).As a result, the prevailing paradigms of service quality acknowledge that service quality is anoverall judgment about service superiority. Service quality essentially is perceived quality(Oliver, 1980).

Most leading service quality scholars agree that service quality as a perception stems fromsome types of comparisons (Gr€onroos, 1984; Parasuraman et al., 1985; Zeithaml et al., 1990).The prevailing SERVQUAL paradigm, as developed by Parasuraman et al. (1985), followsOliver (1980) disconfirmation paradigm and contends service quality is a comparisonbetween what consumers feel the service providers should offer (i.e. customer expectations)and their perceptions of how the service deliverer performs (Parasuraman et al., 1985). Theinitial version of SERVQUAL includes 10 dimensions: reliability, responsiveness,competence, access, courtesy, communication, credibility, security, understanding/knowingthe customer and tangible (Parasuraman et al., 1985). In comparison, the revised version ofSERVQUAL includes five dimensions of service quality. They are tangibles, reliability,

Sport retailquality

responsiveness, assurance and empathy (Parasuraman et al., 1988). These dimensions aregeneric in nature and are suggested to be applicable to a wide variety of services.

Alternativemeasures of service quality exist, and they depart from the SERVQUAL eitheron the target to be compared or if a comparison is needed. For instance, Teas (1993) considersservice quality as a comparison between perception and ideals; the SERVPERF paradigmdeveloped by Cronin and Taylor (1992, 1994) contends that service quality is based onperceptions alone. Furthermore, in sport management literature, Chelladurai and Chang(2000) proposed a comprehensive evaluation model. They consider multiple standards andcontend that the standard to be used depends on the type of service being evaluated.

Inspired by the SERVQUAL and other service quality paradigms, service quality remainsa prominent research area of sport management since 1990s. In addition to Chelladurai andChang (2000) conceptual model, most empirical studies have focused on recreational sportscenters (Kim and Kim, 1995; Ko and Pastore, 2005; Lam et al., 2005), fitness centers (Foroughiet al., 2019; Papadimitriou and Karteroliotis, 2000), spectator sports (McDonald et al., 1995;Yoshida, 2017) and sport educational program (Mao and Zhang, 2012). Sancihak and Seuna-Chana (2013) examined the impact of sponsorship on perceived service quality of sportinggoods stores and they directly applied Dabholkar et al. (1996) scale in measuring sportinggoods retailing quality. Their focus was to examine the impact of sponsorship rather than toexamine the service quality. To our knowledge, no study has focused on sporting goodsstores.

Retail qualityBecause retail stores sell products, together with services implicitly or explicitly, the retailservice quality is more complex. Consumers experience, which is the origin of their qualityperception, is determined not only by in-store experiences but also experiences related to theproducts. Therefore, as reviewed by Dabholkar et al. (1996) and many others, earlier studieson retail service quality by directly applying SERVQUAL found that the five dimensions ofSERVQUAL were inadequate to capture the full gamut of quality in retailing. With anunderstanding of the gap in the literature, Dabholkar et al. (1996) developed a retail servicequality scale, which would later be cited as RSQS by other researchers.

The RSQS can be considered as a modification and extension to the SERVQUAL.Dabholkar et al. (1996) did not attempt to conceptually differentiate retail service quality fromthe generic service quality. Nor were they bothered by if retailing is a service or more than aservice. Rather, efforts were made to incorporate the fact that retailers sell both products andservices; and identify additional dimensions of service quality that are unique to retailing.

The seminal hierarchical structure of retail quality identified by Dabholkar et al. (1996)includes five dimensions: physical aspects, reliability, personal interaction, problem solvingand policy. The physical aspects in turn have two subdimensions: appearance andconvenience, encompassing items relating to appearance of the facility, store layouts,convenience of shopping, etc. The reliability dimension resembles the reliability dimension inthe SERVQUAL; but the RSQS breaks reliability down to “keeping promises” and “doing itright” and considers merchandise availability as a component of “doing it right”. Thepersonal interaction dimension also has two subdimensions: inspiring confidence andcourteous/helpful. As explained by Dabholkar et al. (1996), this dimension subsumes theassurance and responsiveness dimensions of the SERVQUAL. Both problem solving andpolicy do not have subdimensions. Problem solving is specifically reserved for handling ofreturns, exchanges and complaints, which is common in retailing. Policy refers to the servicequality influenced by store policies. The policy dimension in the seminal article is a curiousone as it does not include return policy, warranty policy or other store policies. Instead, thisdimension includes merchandise quality, parking availability and so on.

IJSMS

In summary, in comparison with the pure SERVQUAL services, the peculiaritiesassociated with retailing as identified by the RSQS include: a) store layouts, convenience ofshopping, easiness of finding merchandise (physical aspects); b) merchandise availability(reliability); c) handling of returns, exchanges and complaints (problem solving); and d)merchandise quality, operating hours, parking, credit cards service (policy).

Both the SERVQUAL and the RSQS have inspired much applied research in retailing.Most subsequent investigations of service quality in retailing (Siu and Cheung, 2001), onlineretailing (Trocchia and Janda, 2003) or a subsector of retailing, such as apparel specialtystores (Gagliano and Hathcote, 1994) and grocery stores (Jain and Aggarwal, 2018; Siu andChow, 2004) would utilize the SERVQUAL /RSQS or a modified variant of SERVQUAL/RSQS. Mehta et al. (2000) differentiated two types of retailing, service-intensive retailing andgoods-intensive retailing, and found that SERVQUAL worked better for the former andRSQS excelled for the later.

With the phenomenal growth of e-commerce and online shopping, many recent studies inthe retailing literature have focused on Internet retail (Janda, 2002; Nguyen et al., 2018;Trocchia and Janda, 2003; Yang and Fang, 2004; Yang et al., 2003) or aspects of service qualitythat are pertinent to online retailing, such as logistics and online order fulfillment (Murfieldet al., 2017; Zhang et al., 2019). In the same fashion as RSQS’s extension of SERVQUAL toaccommodate characteristics of retailing, these later studies expanded the literature byincorporating additional dimensions associated with information systems, which is a criticalcomponent of online retailing.

Another salient development in this literature is to examine the retail quality from theperspective of multi-channel retailing as many retails nowadays have both online and offlinechannels (Lin, 2012). For instance, Acquila-Natale and Iglesias-Pradas (2020) developed acomprehensive multi-channel retailing quality scale by integrating quality dimensions thatare pertinent to both physical stores and online shopping. They identified in-store experience,reliability and fulfillment, service provision policies and customer service are the fourretailing quality dimensions.

Content analysis, text mining and leximancerTo uncover retailing quality dimensions, content analysis approach has been utilized bysome researchers (Aggarwal et al., 2009; Simmers and Keith, 2015; Yang and Fang, 2004;Yang et al., 2004; Yang et al., 2003). For instance, Yang and his colleagues have conducted aseries of content analytical studies on online retailing. Based on a total of 1078 onlinereviews, with an average length of 65 words, of nine online companies, Yang et al. (2003)explored service quality dimensions associated with online retailing. They classified thesereviews into positive and negative reviews and compared quality attributes across thesereviews. Yang et al. (2004) subsequently developed an online retailing quality scale basedon statements extracted from 848 reviews of online banking services. Of particularrelevance to the current study is Simmers and Keith (2015). They utilized consumercomments cards to gain insights of retailing quality. Comments cards are used by manyretailing stores to assess service quality in the United States. Based on a content analysis ofa sample of sixty comment cards collected from diverse retail stores throughout the UnitedStates, Simmers and Keith (2015) compared the dimensions derived from comment cardsand the RSQS dimensions. They found that two RSQS subdimensions – convenience (underphysical aspects) and promises (under reliability) – were not present in comment cards.Additionally, they identified several new themes that were not included in the RSQS. Theirnew additions include friendliness and professionalism, check-out, delivery, loading,availability of service, price, selection, value, condition, usability, styling, preference andthe location of store facilities. Comparing with the RSQS, Simmers and Keith (2015) had a

Sport retailquality

broader operationalization of retail quality. More store policies, including refund/exchange,delivery, were also considered.

Complementary to traditional content analysis approach, one salient advancement inmarketing research methodology is the use of text mining. Text mining is a process ofderiving meaningful information from a large number of unstructured texts using NaturalLanguage Processing (NLP) technology and has been widely used for informationextraction and retrieval, clustering and classification and trend detection (Berezina et al.,2016; Yu et al., 2011a). The text mining approach enables researchers to extract structuredinformation from large datasets with objectivity (Duan et al., 2013), and capture semanticalnuances that might be difficult to be differentiated bymanual coding (Roberts et al., 2014). Ithas been suggested that text mining is epistemologically compatible with grounded theoryand content analysis; it is similar to content analysis in the sense that both aim to extractcommon themes by counting words (Yu et al., 2011b). There are numerous applications oftext mining technology in marketing research. For instance, Aggarwal et al. (2009)suggested that text mining technology can be used by retailors to monitor the positioning oftheir brands against their major competitors; Netzer et al. (2012) proposed an innovativemethod to research market structure using text mining; Humphreys and Wang (2018) andBerger et al. (2019) proposed two methodological guidelines of conducting text mining forconsumer behavior and marketing scholars; Berezina et al. (2016) used text mining tounderstand service attributes associated with hotel industry; Chakrabarti et al. (2018)proposed a text mining approach to capture service quality of private banks in India. Theyfound that sentimental scores of consumer reviews significantly correlated with servicequality dimensions.

There are various open source programming packages and commercial software availableto conduct text mining, such as gensim and Scikit-learn packages in Python, mallet in JAVA,tidytext in R, SPSS Modeler Text Analytics, SAS Text Miner, and Leximancer. Particularly,Leximancer has been extensively used in social science and business research, includingsport management research. For instance, Shilbury (2012) and Anagnostopoulos and Bason(2015) used Leximancer to analyze the topics and contents of sport management–relatedjournals; Tseng et al. (2015) used Leximancer to understand destination image based ontravel blogs.

Roughly speaking, there are three types of text mining: unsupervised learning, wherelearning algorithm only draws inferences from input data without labeled responses(analogous to factor analysis which does not have a dependent variable); semi-supervisedlearning, where learning algorithm draws inferences from a small amount of labeled datawith a large amount of unlabeled data; and supervised learning, where learning algorithmdraws inferences from data with labeled responses (analogous to regression analysis whichhas at least one dependent variable) (Ang et al., 2016). Whereas most existing studies haveonly utilized the unsupervised learning algorithm in Leximancer, Leximancer is able toimplement some supervised learning algorithm through “learning from tags”. Oneadvantage of supervised text mining over unsupervised text mining is that it canincorporate information from the labels assigned to the texts. The “learning from tags”function in Leximancer conceptually is similar to Labeled Latent Dirichlet Allocation(Ramage et al., 2009) and supervised topic modeling (Mcauliffe and Blei, 2008). As far as weknow, Leximancer is one of very few programs that have implemented this function.Utilizing Leximancer as a text mining tool, the current study attempts to identify key retailquality dimensions of sporting goods stores as reflected by yelp reviews. Furthermore,drawing upon the “learning from tags” function, this study also explores the relationshipsamong these identified quality dimensions with consumers’ summative judgment (i.e. thestar ratings).

IJSMS

MethodsYelp review dataThis study utilizes the Yelp reviews to understand the service quality dimensions pertainingto sporting goods retailing under themodernmarketingmilieu. The Yelp dataset, released byyelp.com inAugust 2018, includes five json files, 7.3 Gigabyte in total, containing informationabout businesses, reviews, users, check-ins and tips respectively. Of relevance to this studyare the businesses and reviews files. The businesses file is used to filter out the relevantsporting goods stores based on the business category (i.e. primary business activities). Yelphas over 1200 categories of businesses, which were extracted based on consumers search.Whereas Yelp allows a maximum of three categories to be listed for any business, allbusinesses included in this study contain “sporting goods” in their category list.We identified1481 sporting goods stores, including both full-line general stores and specialty stores. Thesestores operate or have operated in 320 zip-codes in 7 states, namely Arizona, Illinois, NorthCarolina, Nevada, Ohio, Pennsylvania and Wisconsin.

After the identification of the stores, their respective reviews can be identified by linkingthe businesses file to the reviews file through unique business_id variable in the dataset.Subsequently, 27,793 reviews with a total of 3,080,054 words were identified. The dates of thereviews spanned from April 29, 2005 to July 2, 2018. The length of reviews ranged from 1 to978 words, with a mean length of 111 words and standard deviation of 100 words. In additionto the text, the star rating of each review is also recorded. Table 1 further presents the wordcounts of reviews grouped by star ratings.

Yelp did not report the demographic information of the reviewers, but they were a part ofthe users. Comparing with the general US population, yelp users on average are younger,wealthier and better educated. Plausibly, the demographics of yelp users are what mostsporting goods stores target at. In the United States, 33% of yelp users are 18–34 years old;35% are 35–54 years old, and 32% are 55þ years old. Regarding educational level, 64% ofyelp users have some college education, 18% have grad school education and 18% are belowcollege education. Regarding income, 51% of yelp users have an income of greater than 100thousand USD, 24% users have an income between 60 and 99 thousand USD and 25% havean income below 59 thousand USD (Bennett et al., 2003).

Leximancer analysisThe current study used Leximancer 4.0, a commercial software implementing a patentedprobabilistic topic modeling algorithm largely based on Naı€ve Bayes classifier withimprovements for seeded classifier approach (Salton, 1989; Smith, 2003). The 27,793 reviewsare subsequently analyzed by Leximancer 4.0. Leximancer identifies themes in a text basedon the occurrences and co-occurrences of words or text segments. Roughly speaking, ananalysis using Leximancer entails the following stages: text preprocessing, automaticconcept seed identification, concept seeds editing, thesaurus identification, compoundconcept editing and concept map and report generation (Nunez-Mir et al., 2016).

Star ratings N Percentage Mean SD Min Median Max

One 4,016 14.45 150 137 6 108 978Two 1,517 5.46 149 120 9 118 905Three 1,712 6.16 131 106 3 103 943Four 4,448 16.00 118 94 2 94 957Five 16,100 57.93 93 81 1 70 934

Table 1.Descriptive statistics ofreview length by star

ratings

Sport retailquality

Preprocessing. Preprocessing text file is critical to any Natural Language Processingalgorithms. One advantage of Leximancer is that the software can almost automate thecritical data preprocessing step with minimal researchers’ input. Therefore, theunpreprocessed reviews were supplied to Leximancer.

Concept Editing. In Leximancer, there are three key terms: theme, concept and thesaurus.Themes are clusters of concepts that are selected based on co-occurrences; concepts in turnare defined by a thesaurus of words. Leximancer firstly generates concepts and themesbefore it identifies the relevant thesaurus of words. After preprocessing, Leximancer canquickly identify the concepts and themes in an exploratory fashion based on frequency andco-occurrences of words. Automatic concept seeds are generated based on a ranked list ofthese words. If desirable, conceptual seeds can also be manually edited by researchers. Forinstance, as reviewers often mention about store names in the reviews, we combine all thenames into one concept and label it as “storeName”; this also applies to city names (cityName),brand names (brandName), staff names (staffName). The reviewersmay alsomention about aspecific product in the reviews, which may not be of central interest to our research. We maymerge all these different product names using one concept “product”. In this study, we usedcamel case words (e.g. storeName) to indicate user-defined seeds.

Compound Concepts. Although Leximancer can automatically extract some multiple-words proper names and treat them as one concept, the software relies on the capitalization ofthe words as a hint of proper names (Leximancer, 2011). Some compound concepts, such as“customer service”, were not able to be automatically identified. After the generation ofthesaurus of words, researchers can manually add compound concepts. For this study, weadded one compound concept “customer service”.

Mapping Concepts. Finally, Leximancer can select concepts and tags to map in the finalreport generation stage. Tags are special names assigned to a folder, a file, a variable,patterns, non-meaningful texts. By default, words associated with tags will be excluded fromthe machine learning process, unless the program is instructed to “learn from tags”. In thisstudy, we instructed the program to learn the star rating tags. Furthermore, to be explained inthe discussion section, we assigned “gun(s)”, “bike(s)”, and “bicycle(s)” as killed concepts. Thefinal output of Leximancer provides an interactive conceptual map that visually representsthe main themes and concepts extracted from the text mining process and containsinformation about how they are associated.

ResultsValidity of extracted informationAn analysis of the 27,793 reviews about 1481 stores that sell sporting goods produced aconcept map consisting of a maximum of 81 concepts. Before we present the substantivethemes and concepts, we attempt to assess the face validity of the extracted information. Textmining is a relatively new method to academic research; as a result, the standards forassessing its validity have not yet fully established. Following the validity types proposed byKrippendorff (2004), Dr. Andrew Smith, the inventor of Leximancer, assessed face validity,stability, reproducibility, correlative validity and functional validity of Leximancer andfound the tool reliable and valid (Smith and Humphreys, 2006).

Humphreys and Wang (2018) suggested that three strategies – comparison, correlationand prediction – may be used to assess various types of validity, which would require thedocument-term matrix or collecting additional external non-textual variables. As aproprietary software, Leximancer does not disclose the exact formula and algorithms usedfor the analyses; neither does it provide the conventional document-term matrix as far as weknow. Therefore, the ability of conducting further analysis is restricted due to the lack ofdocument-term matrix, and we can only conduct some descriptive statistics based on the

IJSMS

available output. In the spirit of Humphreys and Wang (2018) correlation strategy, weexamined the correspondence of the extracted sentimental words and star ratings to assesshowwell the extracted concepts have preserved information in the raw reviews. It is intuitiveto hypothesize that reviews with higher ratings will contain more positive sentimental wordsthan those with lower ratings. The results of our analysis, as shown in Table 2, suggest thathigher ratings are indeed accompanied by more positive sentimental words and fewernegative sentimental words. This pattern is largely consistent across all the sentimentalwords identified Leximancer, which lends credibility to the information extracted by the textmining approach.

Emergent themes and conceptsFigure 1 presents the concept map with theme size at 50% and visible concepts at 100%,which serves as the foundation for further analysis. The Leximancer concept map isinteractive in nature: different clusters of concepts (i.e. themes) can be generated by adjustingthe “% Theme Size” sliding bar. With theme size at 100%, only one theme, “customer” isproduced; with theme size at 0%, each individual concept becomes a theme. Furthermore,although some of the identified concepts were deemed to be generic and less diagnostic (e.g.“feel”, “nice”, “area”), there is nothing to lose by keeping them for potential exploration. Hence,we decided to keep “% Visible Concepts” at 100%.

Five themes, “customer”, “staff”, “prices”, “store” and “return”, with a connectivity rate of100%, 90%, 71%, 59% and 34% respectively, were identified as dominant by Leximancer.Connectivity, indicating the relative importance of the themes, is calculated based on the co-occurrence counts of the concept with every other concept on the map (Leximancer, 2011).

Customer. The strongest theme, “customer”, is the most important theme. Out of the12,962 counts of “customer”, 11,021 are paired with “service”, making it a compound theme,“customer service”. This theme, including concepts like “experience”, “care”, “business”,“review”, “purchased”, “owner”, “time”, “local”, “life”, relates to the holistic judgment ofservice experience and customer care.

Here are some quotations of this theme:

. . . The customer service is ABSOLUTELY TERRIBLE in all of my experiences. Today i went in tobuy some balls and 5 of them were just sitting around not doing anything. . .

Concept One-star Two-star Three-star Four-star Five-star

FAVORABLE 0.65 1.06 1.58 2.03 2.38nice 0.08 0.18 0.26 0.27 0.21love 0.04 0.09 0.12 0.21 0.23amazing 0.02 0.01 0.05 0.05 0.15cool 0.03 0.05 0.09 0.12 0.09fun 0.02 0.02 0.07 0.1 0.1UNFAVORABLE 1.07 0.87 0.57 0.35 0.31bad 0.14 0.15 0.12 0.09 0.06

Note(s): FAVORABLE, UNFAVORABLE are not words used in the reviews. Instead, they are collections ofsentimental words automatically detected by Leximancer.Words in lower cases are the exact words used in thereviews. As reviews of different star ratings on average have different lengths, the frequency was scaled bylength of the review using 150, the average length of one-star review, as the baseline. The scaling coefficientsfor two-star, three-star, four-star and five-star reviewswere 150/149, 150/131, 150/118, and 150/93 (see Table 1).The numbers in the table mean that given the level of star rating, how likely these sentimental words wouldoccur in the review text

Table 2.Occurrence frequenciesof sentimental words

by star ratings

Sport retailquality

. . .He had no intention to serviceme, he did not care if I left unpleased, there was no sense of urgencyin the atmosphere to leave any customer with a lasting impression. A good lasting impression at that.Customer service once meant meeting the needs of the customer by any means, especially whenyou’re more than capable of doing so. . .

Staff. The second strongest theme, “staff” (merged with “employee(s)”) has a word count of14,442, which corresponds to the Sales Staff dimension by Simmers and Keith (2015). Themost important concepts under this theme are “helpful”, “friendly” and “knowledgeable”; andthe theme connects to “needed” (mergedwith “need(s)”), “recommend”, and “staffName”. Thisfollowing quote highlights the quality of sales staff expected by consumers and positivebehavioral consequences associated with high quality of sales staff:

. . .Wewere at [storeName] the day before and left empty handed because the girl who worked in thesnowboarding dept. had no idea what she was talking about. [Another storeName] was the exactopposite. A very helpful and informative guy by the name of [staffName] answered all of ourquestions. He gave us his personal advice over what size board would be right for me . . . Afterspeaking with him I felt confident in the purchase I was making, and so happy that I ended by a new

The black line in the middle, which is not a Leximancer output, is manually added by theresearcher. Additionally, the researcher also added the names of themes to the original Leximancer output

Figure 1.Leximancer conceptualmap with VisualConcepts at 100% andTheme Size at 50%.The Leximanceroutput is interactive asthe users can decide thenumber of conceptsand the number ofthemes to be displayedby adjusting thesliding bars (% VisibleConcepts and %Theme Size,respectively) below theconceptual map. Thethemes are displayed ina conventional heatmap scheme, with themost important themedisplaying in red,followed by orange,green and so on; thesize of the circles has nobearing as to itsimportance in the text(Leximancer, 2011).The words in redstarting with “stars_”are the ratings of thebusinesses and weretreated as “tags” byLeximancer. Theconcepts that clusternear a tag indicate thatthey are more specificto the tag

IJSMS

pair of pants that I did not really even need. I am just sooooooo impressed by . . . the great customerservice that was provided.

Price. The third strongest theme, “price”, with a word count of 9170, is really onesubdimension of “product”which has a word count of 7106. This theme also aligns well withthe product dimension identified by Simmers and Keith (2015). Apparently, “price” is themost important concept under this theme. Other highly connected concepts are “selection”,“quality”, “inventory” / “stores” (as a verb, merging with “storing”). Here are a few reviewexcerpts that highlight the product dimension:

. . . Stopped by today just to check out the store, and Iwalked outwith a new backpacking tent. Pricesare very reasonable. . .Not a huge selection, but they definitely have quality products.

“. . .This place has all the snow and skateboard gear onemight need. Great priced helmets for skatingand skiing, great penny board selection.” “The [brandName], [brandName], [brandName] and[brandName] shoe selection is insane. Seriously, blows [competitorStore] out of the water. . . kids andadult sizes! And [brandName] are hard to find. . . my husband loves them . . .

Store. The fourth theme, “store”, with a word count of 12,851, includes the store facilitiesdimension by Simmers and Keith (2015) and also overlaps with the service dimension bySimmers and Keith (2015) and policy dimension by Dabholkar et al. (1996). The criticalconcepts associated with this theme include “sales”, “location”, “items”, “size”, “line” (i.e.checkout line), “order”, “brandName”. More interestingly, this theme is closely related to“online” and “competitorStores”. This theme is also associated with “holidayName” as moststores have sales promotion during holiday seasons. The following are a few reviews that arerepresentative of this dimension:

. . .The rest of the store I do not mind. It’s well organized and occasionally you can find good sales.

. . .I really like this [storeName] location . . .Great location within theOakville downtown shops. Onlydownside is hard to find street parking, and you can only park for 1 hour.

. . .The only thing this [storeName] store has going for it is the convenience of its location.

Return. The last theme, “return”, with a word count of 2931, is closely related to “order”,“trying”, “money”, “manager”, “bad”, “problem”, “perceivedValue”, “online”, “online order”,and time related concepts (i.e. “minutes”, “week”, “days”). Given the fact that this theme ishighly associated with one-star reviews, an examination of the reviews indicates that thistheme is primarily about problem handling, order returning and re-patronage intention.

Further examination of the concept map also reveals that this theme is split by tworelatively distinct clusters of concepts. The upper left sphere of this theme, where adjoins the“customer” theme, clusters concepts of “manager”, “bad”, “problem”; and the lower rightsphere of this theme, where adjoins the “store” theme, clusters concepts of “order”, “home”,“online”, “trying”, “online order”. The former, generally relating to the policy and problem-solving dimensions identified by Dabholkar et al. (1996), has highlighted the role of storemanagers. The reviews reveal that a lack of service orientation from the management leveloften leads to poor service and negative behavioral consequences. The latter, which has notbeen documented in the existing literature, reveals how the presence of online channelsinteracts with the perceived offline retail quality. Given the novelty of this finding, we decideto separate the return theme into “manager” and “online”.

Online. In this section we further explore the “online” concept, which has a word count of1906. Comparing with online stores, better customer service is the reason that consumersmight remain loyal to offline channels, which is reflected in many of the reviews. It is alsoevident that service quality, as consumers’ perception, has been impacted by the onlinechannels in multiple ways. First, many consumers have higher expectation about service

Sport retailquality

quality in the offline channels, due to the threats of the online channels. The consumers feelthey are empowered to switch channels and thus entitled to receive good services at theoffline channel. Second, consumers expect that the prices at the online channels to be lower.They constantly compare the prices at the offline stores with online channels. Most sportinggoods are shopping goods as they are typically valuable, consumers often have significantshopping involvement, such as quality inspection and price comparisons. Consumers will bepleasantly surprised when they find out that the offline channel has competitive prices vis-�a-vis the online channels.

However, with the easiness of price comparison enabled by smartphone and price checkerapps, the online channels have pushed the offline channels to bundle their products withcreative value-adding services. Branded variants (e.g. exclusive products; Bergen et al., 1996;Shugan, 1983), in-house brands (i.e. private labels) and service bundles are effective tools toavoid fierce price competition. However, smaller store may not have the bargaining powerwith the manufacturers to create their own product variants or private products, only theservice bundle may be feasible. This again highlights the importance of customer service.

. . .Will be back to this store and purchase other items instead of online. What a difference being ableto customize an item and have accessories installed at the same stop. You can always order shoesonline, but it does not beat trying them in on the store and getting advice from an actual runner.

Third, although most transactions of sporting goods occur in the offline channels, thee-commerce is on the rise. Leading sporting goods chains are prioritizing their investment ononline channels (Dick’s Sporting Goods, 2018). Many sporting goods stores have their ownonline stores; they may have “buy online pick up in store” service; they may have “in storepreorder” service if they do not have items in stock; they may have an “online stock check”service. To compete with dominant online channels, going online is another strategic movefor most sporting goods stores. However, the integration of the online channels and offlinechannel becomes a new challenge in running a sporting goods store, as evident in thefollowing reviews.

Returning an online order to an offline channel:

. . .Ordered a pair of shoes online and they did not fit brought them to a [brandName] corporate storewhich they are listed on [brandName] website as and the brilliant manager proceeds to tell me he hasnever encounter a return like this and do not know how to do it or even go about finding a resolutionto my problem. . .

Online inventory check:

. . . I called store based on online inventory indicating that the store had 1 pair of the gloves that I waslooking for. . .After driving the 30miles to this location they did not have anything on hold for me. . .

In store ordering:

. . .So because their inventory was so messed up, we proceeded to check out . . . we were buying 2pairs of shoes from the store, then 3 pairs from online since sizes were not available in the store. . . .This was a 30 minute þ process and was A TOTAL WASTE OF TIME!

Through this analysis, it becomes apparent that online channels have exerted significantinfluences on consumers’ evaluation of retail quality. More importantly, an effectiveintegration of online–offline channels has become a new dimension of retail quality (Chiuet al., 2011; Ofek et al., 2011).

Retail quality of sporting goods storesAs indicated in Figure 1, some identified circles overlap with each other, whereas others donot. The overlapped areas suggest that the themes are semantically related to each other; thenonoverlapping on the hand suggests distinctiveness of the themes. There are some

IJSMS

interesting observations by examining the overlapped themes and concepts: “love” is at theintersection of “staff” and “product”; “online” and “perceivedValue” are at the intersection of“store” and “return”; “problem” and “review” are at the intersection of “customer” and“return”; “product”, “inventory” and “deal” are at the intersection of “prices” and “store”.Furthermore, we can hypothetically draw a line to separate the five themes into two(Figure 1): the upper left being customer service and the lower right being store aspects,which more or less is consistent with the existing conceptualization of retailing quality(Westbrook, 1981). The current analysis, however, provides not only empirical evidence forthe relative separation of these two themes but also insights about where these two themesinteract. It seems that there are two main interaction areas (i.e. overlapping areas): the firstbeing “staff” and “product”, and the second “manager” and “online”. Additionally, thisseparation can be confirmed by adjusting the theme size of the mind map to 97%, when onlytwo themes would emerge.

Although the themes and concepts identified by Leximancer have provided valuableinsights about quality of sporting goods retailing, they are indicative in nature. Based on theabove analysis, we re-name the themes to better reflect the meanings of the concepts and re-cluster them in a hierarchical structure. Table 3 presents the final conceptual model of RetailQuality of Sporting Goods Stores (RQSGS). On the very top level, customer service and storeaspects serve as the two primary dimensions of RQSGS. Under customer service, three mainthemes, including holistic experience, manager and staff have emerged. Holistic experiencefocuses mainly on customer care and shopping experience. Manager highlights the role ofmanagers in elevating retail quality; problem handling, terms of sales, and a promptrefunding policy have been particularly pronounced in this dimension. Staff are the frontlineservice individuals who typically provide direct customer service to consumers. Helpfulness,knowledge and friendliness are the three major positive qualities that are valued bycustomers. Under store aspects, three main themes, including product, B&M store andonline–offline integration, have emerged. Regarding product, consumers care about price,selection, quality and brand. Regarding B&M store, consumers care about inventory,

Second-orderdimensions Dimensions Subdimensions Leximancer concepts

Customer service Holistic experience Customer care CareShoppingexperience

Experience

Staff Helpful helpfulKnowledgeable knowledgeableFriendly friendly

Manager Problem handling manager, owner, problemTerms of sales money, returnPromptness week, days, minutes

Store aspects Product Price price(s), pricingSelection selectionQuality qualityBrand brand, brandName

B&M Store Inventory inventory, storesLocation locationIn-store sales sales, holidayNamePerceived value perceivedValue,

competitorStoreOnline-offlineintegration

online, order, return

Table 3.The hierarchical model

of retail quality ofsporting goods stores

Sport retailquality

location, sales promotion and perceived value. Finally, online–offline integration (Nguyenet al., 2018) becomes a salient dimension in RQSGS.

RQSGS vis-�a-vis star ratingsBased on the relative positions of the star rating tags in Figure 1 and the numericrepresentation of conditional probability of word occurrence in Table 4, it is evident thatdifferent star ratings have focused on different aspects of retail quality. A general conclusionis that one-star and five-star reviews focus more on customer service, whereas two- to four-star reviews focus more on store aspects. Figure 2 presents a heat map of the occurrences ofconcepts by normalizing the data points in Table 4. Green color on the figure indicates morelikely to occur and red color least likely to occur. By grouping concepts into their respectivedimensions, Figure 3 presents the probability of dimensions by star ratings. It is evident thatfrom one-star to five-star ratings, some dimensions are monotonically increasing (i.e. Staff),some monotonically decreasing (i.e. Manager), some peaks at the middle (i.e. Product andB&M Store) and others peak at the extremes (i.e. Holistic Experience).

Customer Service. Holistic experience, as indicated by “care” and “experience”, ismentioned more often in one-star and two-star reviews, followed by five-star reviews, three-star and four-star reviews do not discuss this theme as often. Regarding positivecharacteristics of staff, as indicated by “helpful”, “knowledgeable” and “friendly”, we canroughly observe a positive correlation between star ratings and frequency of word

Concept Counts One-star Two-star Three-star Four-star Five-star

Care 1970 10 9 6 4 7experience 4856 21 19 15 13 18helpful 5514 6 11 17 25 23knowledgeable 4674 4 9 9 15 22friendly 7157 13 19 21 30 29manager 1805 20 12 7 4 3owner 2300 14 8 4 4 8problem 1834 11 11 8 6 5money 2227 20 14 9 7 5return 2937 28 20 9 8 6week 2464 17 15 7 7 7days 1791 13 10 7 5 5minutes 2724 20 18 10 9 7prices 9170 29 41 49 42 29selection 4936 6 21 35 30 15quality 2092 5 7 9 10 7Brand 1842 6 9 11 10 5brandName 936 3 6 6 6 2Inventory 684 2 3 4 4 2Location 3293 13 17 21 19 8holidayName 432 1 2 3 3 1Sales 4116 22 24 25 21 9competitorStores 479 2 2 4 3 1perceivedValue 436 2 2 2 2 1Online 1906 11 12 10 7 5

Note(s): The numbers on the columns 3–7mean the frequencies of word occurrences. For instance, in each 100one-star reviews, the concept “care” would occur 10 times; whereas in each 100 four-star reviews, the sameconcept would only occur four times. Whereas some concepts are distributed relatively evenly among allgroups, other concepts are more or less prominent with certain group. For instance, “knowledgeable” and“sales” with five-star reviews; “manager” with one-star reviews

Table 4.Frequencies of conceptoccurrences in every100 reviews

IJSMS

occurrences under this theme. Five-star and four-star reviews are strongly related to thepositive characteristics of staff. Whereas four-star reviews are more likely to use “helpful”and “friendly” than five-star reviews; five-star reviews are more likely to use“knowledgeable”. “Knowledgeable” is one the two most diagnostic terms (the other being“sales”) that separates five-star reviews from the rest.With regard to themanager dimension,we observe a roughly monotonic decreasing occurrence of such words as “manager”,“owner”, “problem”, “money”, “return”, “week”, “days”, and “minutes”, from one-star to five-star reviews. One-star and two-star reviews tend to focus much more on this dimension thanthe rest.

Store Aspects. Regarding price, selection, quality and brand of the product dimension, weobserve that three-star and four-star reviews tend to discuss most often, followed by two-starreviews; one-star and five-star reviews tend to discuss least. The same phenomenon holds forthe location, promotion aspects of the B&M store dimension. Furthermore, five-star reviewstend to discuss “sales” the least. Three-star and four-star reviews are strongly related to theB&M store dimension. Particularly, these reviews are closely related to “competitorStore”and “online” concepts. In those reviews, a comparison with competing stores, such asAmazon andWalmart, or other online channels is often mentioned, indicating an assessmentof quality is anchoredwith competing stores and influenced by online channels. Two-star andone-star reviews are strongly related the manager dimension. Essentially, the lack of servicefrom the management level leads to order returning behavior and negative repatronageintentions. Value-laden words, such as “worth” and “value”, are often mentioned in thesereviews. “Online” again is a closely connected concept, indicating impacts of online channels.

RATING ONE TWO THREE FOUR FIVE

care 1 0.83 0.33 0 0.5

experience 1 0.75 0.25 0 0.63

helpful 0 0.26 0.58 1 0.89

knowledgeable 0 0.28 0.28 0.61 1

friendly 0 0.35 0.47 1 0.94

manager 1 0.53 0.24 0.06 0

owner 1 0.4 0 0 0.4

problem 1 1 0.5 0.17 0

money 1 0.6 0.27 0.13 0

return 1 0.64 0.14 0.09 0

week 1 0.8 0 0 0

days 1 0.63 0.25 0 0

minutes 1 0.85 0.23 0.15 0

prices 0 0.6 1 0.65 0

selection 0 0.52 1 0.83 0.31

quality 0 0.4 0.8 1 0.4

brand 0.17 0.67 1 0.83 0

brandName 0.25 1 1 1 0

inventory 0 0.26 1 0.71 0

location 0.38 0.69 1 0.85 0

holidayName 0 0.5 1 1 0

sales 0.81 0.94 1 0.75 0

competitorStores 0.41 0.49 1 0.6 0

perceivedValue 1 0.42 0.32 0.79 0

online 0.86 1 0.71 0.29 0

Figure 2.A heat map of the

occurrences ofconcepts by

normalizing the datapoints in Table 3 into 0-

1 range. For eachconcept (i.e. row ofdata), the following

normalization formulawas used:

zi ¼ xi −minðxÞmaxðxÞ−minðxÞ,

where x ¼ ðx1; . . . ; x5Þand zi is the normalizeddata. Green color on thefigure indicates more

likely to occur and redcolor least likely

to occur

Sport retailquality

DiscussionSporting goods retailing is a significant supporting sector of the sports industry(Eschenfelder and Li, 2007). The main objective of this research is to investigate the retailquality of sporting goods stores using a relatively novel approach – text mining. Drawingupon themes and concepts extracted by Leximancer on voluminous consumer self-generatedonline reviews, this study identifies key quality dimensions within sporting goods retailing.We subsequently re-theme and re-cluster the Leximancer map and propose a model of RetailQuality of Sporting Goods Stores (RQSGS). Consistent with the existing retail service qualityliterature (Dabholkar et al., 1996; Simmers and Keith, 2015), this model takes a broadperspective of retail quality and identifies customer service and store aspects as two higher-order dimensions within RQSGS. Customer service includes holistic experience, manager andstaff; store aspects include product, B&M store and online–offline integration. These twodimensions are relatively distinct, but they also interact through store, manager, staff andproduct. Furthermore, this study explores the relative importance of each concept to differentlevels of ratings. To our knowledge, this is also the first study examining service quality inthe context of sporting goods retailing.

Theoretical and practical implicationsThe current study to certain extent cross-validated the existing retailing literature that isdeveloped on alternativemethods (Acquila-Natale and Iglesias-Pradas, 2020; Dabholkar et al.,1996; Simmers and Keith, 2015): most themes identified in RQSGS are consistent with theexisting literature. Whereas the emerging multi-channel retailing quality literature hasfocused on the integration of consumer experiences in both physical stores and online stores

O N E -S T AR T W O -S T AR T H R E E -S T AR F O UR -S T AR F I VE -S T AR

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Holistic Experience Staff Manager

Product B&M Store Online-offline Integration

Figure 3.Likelihood ofdimension occurrencesby star ratings

IJSMS

(Acquila-Natale and Iglesias-Pradas, 2020; Murfield et al., 2017; Zhang et al., 2019), it does notfit the sporting goods retailing completely as sporting goods stores rely on revenues fromphysical stores. The current study in the context of sporting goods retailing revealed severalpoints that had not been discussed in the existing literature.We focus our discussion on thesefindings.

First, knowledgeable staff is ultimate important to create satisfied consumers in sportretailing. Staff has been the focus of service quality literature: four out of the five SERVQUALdimensions, namely reliability, responsiveness, assurance and empathy, are staff related. Inretailing literature, the importance of sales staff has also been highlighted (Simmers andKeith, 2015). However, the existing literature often considers only the sales staff. In thecurrent context, we do not limit staff to sales staff. Instead, staff are the frontline employeeswho help with the sales and/or provide other value-added services (e.g. a biomechanicsspecialist in a specialty running shoe store, a club repairer in a golf pro shop). Whereas threepositive characteristics of the staff – knowledge, friendliness, and helpfulness – wereidentified important in sporting goods retailing business, “knowledgeable”, however, is onethe most diagnostic concepts that separate five-star reviews from four-star reviews.

Second, managers play a critical role in elevating consumers’ satisfaction. Manager,loosely referring to both owners and professional managers, has three subdimensions:problem handling, terms of sales and promptness.Whereas the role of frontline staff has beenacknowledged in the literature, the role of manager is less clear. Our results suggest thatconsumers tend to blame managers/owners for poor service and experience. It might be dueto the fact that many sporting goods stores are small local businesses. Traditionally, sportinggoods retailers aremerchandisemovers and are typically low in service orientation (Garg andChan, 1997). Faced with severe competition and easiness of price comparisons due to nationalbranding, sporting goods stores are urged to differentiate themselves from their local andonline competitors. Among exclusive distribution, branded variants, private labels andservice bundling, augmenting merchandises with value-adding services become one of veryfew viable approaches sporting goods stores might take (Homburg et al., 2002). Therefore, thedemarcation of merchandise and service has become increasingly blurred. Sporting goodsstores are no longer in the merchandise mover business; rather, they are shifting towardsservice business. As a result, managers and owners’ service orientation must be emphasizedin the current marketing landscape (Garg and Chan, 1997; Homburg et al., 2002). Storepolicies, such as return/exchange, warranty, credit cards, installation, are the materializationof manager’s service orientation philosophy.

Third, online–offline integration is a new dimension of retail quality. Consistent with theexisting literature, the online channels have considerable impacts on consumers’ shoppingbehavior. Consumers rely on the online channels to find out product information, stockinginformation, store information, and pricing information; or to place orders online and thenpick them up in store. As a result, traditional B&M stores have started to expand their onlinepresence. However, the lack of effective integration between the online and offline channelshas created much consumers’ dismay. An effective integration of the channels is likely animportant area of investment for sporting goods stores.

Finally, hedonic and social values are important to create superior satisfaction.Ostensibly, from our data, extreme reviews (i.e. one-star and five-star) tend to focus oncustomer service andmediocre reviews (i.e. two-star, three-star and four-star reviews) tend tofocus on store aspects. Fundamentally, we argue that different value dimensions have playeda role in consumers’ satisfaction. Rintam€aki (2006) has decomposed values of shopping intoutilitarian, hedonic and social values. Utilitarian values focus on monetary savings,convenience of shopping; hedonic values focus on emotional, entertainment aspects ofshopping; and social values focus on the symbolism and relationships. It is evident from thereview data that utilitarian value is the bedrock of shopping experience, but is in itself often

Sport retailquality

unable to produce ultimate satisfaction. Verhoef et al. (2009) suggest social environment,service interface, retail atmosphere, assortment, price, customer experiences in alternativechannels and retail brand are the major drivers of customer experiences. In the context ofsport retailing, there are still rooms to grow consumers’ experience by utilizing the driversmentioned in Verhoef et al. (2009).

It should also be noted that sporting goods retailing only has subtle differences withgeneral retailing. These differences are not only nuanced but also may be shared by otherspecific retailing contexts. For instance, sporting goods are mostly shopping goods thereforefree riding is more likely; grocery retailing typically does not share this characteristic butelectronics retailing might; in the grocery retailing case, friendly staff might be moreimportant than knowledgeable staff but in the latter case the opposite might be true. Piano isoften considered as shopping goods, but piano retailing might have much less free riding dueto limited distribution, therefore online–offline integration might be less an issue. Manysporting goods stores are small in scale and cater to local markets, therefore managers andowners may play a more significant role in delivering services. Sporting goods stores areoften augmenting their products with ancillary services such as gait analysis, repair services,renting services or training classes, therefore the definition of staffmay bemuch broader thanother retailing contexts. Sporting goods are mostly national/international brands that aredistributed across many channels therefore price competition is severe in comparison withother retailing such as high fashion. Nonetheless, it is almost always possible to find anothercontext that will share the same characteristics with sports retailing.

Limitations and future researchOne challenge and limitation of this research is the identification of sporting goods stores.Although theNAICS has provided a straightforward definition of sporting goods stores, frommarket definition point of view, limiting stores to NAICS 451110maymisidentify true playersin this industry, thus inappropriate (Shugan, 2011). For instance, bothWalmart and Amazondo not belong to NAICS 451110. As a compromise, this study relies on stores’ self-identification. We include all the stores who claim they sell sporting goods. On the Yelpplatform, many gun shops and shooting ranges are listed as sporting goods stores; granted,shops sell sporting guns are NAICS 451110 per the NAICS. Also, in the dataset, we have alarge proportion of bike shops which both sell and repair bikes. We even included a smallnumber of vision centers that may sell eyeglasses for sports use. As Leximancer extractconcepts based on frequency, the large amount of gun shops and bike shops we have in thedataset could have biased the findings. Hence, we excluded some bike and gun shops byassigning “gun(s)” and “bike(s)” “bicycle(s)” as killed concepts. Essentially, we underweightconcepts associated with bike and gun shops. Due to the lack of a precise definition ofsporting goods stores, the applicability of RQSGS to certain sporting goods stores warrantsfurther research.

A second challenge of this research is the representativeness of Yelp reviewers. We do notbelieve Yelp reviewers are a representative sample of all consumers. However, as the unit ofanalysis is individual review instead of individual reviewer, at the issue really is therepresentativeness of the reviews; and the exact nature of this concern cannot be empiricallyassessed. We do have confidence in our findings: as in many qualitative studies, a saturationof themes can often be achieved with a small number of participants. Given the large amountof reviews we have, the topics that are covered by the current study are likely more extensivethan the traditional methods. The chance of omission will be smaller. This approachessentially has the same spirit as the “maximum variance sampling” (aka heterogeneoussampling) technique: the goal is not to be representative of the average views on an issue butto look at it from all angles. The key empirical task is to determine the weights of these

IJSMS

concepts. Although a closer examination of rare concepts will be of interest in our futureresearch, it is beyond the scope of the current project.

The generalizability of RQSGS beyond the sporting goods retailing industry is asecondary consideration of this project. We do not intend to claim the framework works forother retailing businesses, as different sectors likely have their own salient characteristics.However, we do believe that the framework provides a supplementary foundation for otherretailing contexts; and the RQSGS may be generalizable to other retailing businesses thatshare similar characteristics with sporting goods retailing (such asmusic instruments stores).

This research was exploratory with the goal of producing a better understanding ofservice quality in sporting goods retailing. We neither examine service quality relating tospecific types of stores nor quantify the relative importance of each dimension. Lastly,validating the findings from text mining remains a promising task. Cross-validating findingsbased on Humphreys and Wang (2018) strategy can be fruitful. Future studies may addressthese topics.

References

Acquila-Natale, E. and Iglesias-Pradas, S. (2020), “How to measure quality in multi-channel retailingand not die trying”, Journal of Business Research, Vol. 109, pp. 38-48, doi: 10.1016/j.jbusres.2019.10.041.

Aggarwal, P., Vaidyanathan, R. and Venkatesh, A. (2009), “Using lexical semantic analysis to deriveonline brand positions: an application to retail marketing research”, Journal of Retailing, Vol. 85No. 2, pp. 145-158, doi: 10.1016/j.jretai.2009.03.001.

Anagnostopoulos, C. and Bason, T. (2015), “Mapping the first 10 years with Leximancer: themes andconcepts in the sports management international journal choregia”, Sports ManagementInternational Journal, Vol. 11 No. 1, pp. 24-41.

Ang, J.C., Mirzal, A., Haron, H. and Hamed, H.N.A. (2016), “Supervised, unsupervised, and semi-supervised feature selection: a review on gene selection”, IEEE/ACM Transactions onComputational Biology and Bioinformatics, Vol. 13 No. 5, pp. 971-989, doi: 10.1109/TCBB.2015.2478454.

Bennett, G., Henson, R.K. and Zhang, J. (2003), “Generation Y’s perceptions of the action sportsindustry segment”, Journal of Sport Management, Vol. 17 No. 2, pp. 95-115.

Berezina, K., Bilgihan, A., Cobanoglu, C. and Okumus, F. (2016), “Understanding satisfied anddissatisfied hotel customers: text mining of online hotel reviews”, Journal of HospitalityMarketing and Management, Vol. 25 No. 1, pp. 1-24, doi: 10.1080/19368623.2015.983631.

Bergen, M., Dutta, S. and Shugan, S.M. (1996), “Branded variants: a retail perspective”, Journal ofMarketing Research, Vol. 33 No. 1, pp. 9-19.

Berger, J., Humphreys, A., Ludwig, S., Moe, W.W., Netzer, O. and Schweidel, D.A.J.J.o. M. (2019), “Unitingthe tribes: using text for marketing insight”, Journal of Marketing, Vol. 84 No. 1, pp. 1-25.

Cabela’s (2016), “Cabela’s incorporated form 10-k”, available at: http://www.annualreports.com/Click/7867.

U.S. Census (2018), “Advance monthly sales for retail and food service”, available at: https://www.census.gov/economic-indicators/#qfr_retail.

U.S. Census (2019), “North American industry classification system definition: 451110 - sporting goodsstores”, available at: https://www.naics.com/naics-code-description/?code5451110.

Chakrabarti, S., Trehan, D. and Makhija, M. (2018), “Assessment of service quality using text mining –evidence from private sector banks in India”, International Journal of Bank Marketing, Vol. 36No. 4, pp. 594-615, doi: 10.1108/IJBM-04-2017-0070.

Chelladurai, P. and Chang, K. (2000), “Targets and standards of quality in sport services”, SportManagement Review, Vol. 3 No. 1, pp. 1-22.

Sport retailquality

Chiu, H.-C., Hsieh, Y.-C., Roan, J., Tseng, K.-J. and Hsieh, J.-K. (2011), “The challenge for multichannelservices: cross-channel free-riding behavior”, Electronic Commerce Research and Applications,Vol. 10 No. 2, pp. 268-277.

Cronin, J.J. Jr and Taylor, S.A. (1992), “Measuring service quality: a reexamination and extension”,Journal of Marketing, Vol. 56 No. 3, pp. 55-68.

Cronin, J.J. Jr and Taylor, S.A. (1994), “SERVPERF versus SERVQUAL: reconciling performance-based and perceptions-minus-expectations measurement of service quality”, Journal ofMarketing, Vol. 58 No. 1, pp. 125-131.

Dabholkar, P., Thorpe, D. and Rentz, J. (1996), “A measure of service quality for retail stores: scaledevelopment and validation”, Journal of the Academy of Marketing Science, Vol. 24 No. 1, pp. 3-16.

Davis-Sramek, B., Droge, C., Mentzer, J.T. and Myers, M.B. (2009), “Creating commitment and loyaltybehavior among retailers: what are the roles of service quality and satisfaction?”, Journal of theAcademy of Marketing Science, Vol. 37 No. 4, pp. 440-454.

Dick’s Sporting Goods (2018), “2017 annual report”, available at: www.annualreports.com/Company/dicks-sporting-goods-inc.

Duan, W., Cao, Q., Yu, Y. and Levy, S. (2013), “Mining online user-generated content: using sentimentanalysis technique to study hotel service quality”, Paper presented at the 2013 46th HawaiiInternational Conference on System Sciences.

Eschenfelder, M.J. and Li, M. (2007), Economics of Sport, Morgantown, Fitness InformationTechnology, WV.

Foroughi, B., Iranmanesh, M., Gholipour, H.F. and Hyun, S.S. (2019), “Examining relationships amongprocess quality, outcome quality, delight, satisfaction and behavioural intentions in fitnesscentres in Malaysia”, International Journal of Sports Marketing and Sponsorship, Vol. 20 No. 3,pp. 374-389, doi: 10.1108/IJSMS-08-2018-0078.

Gagliano, K.B. and Hathcote, J. (1994), “Customer expectations and perceptions of service quality inretail apparel specialty stores”, Journal of Services Marketing, Vol. 8 No. 1, pp. 60-69.

Gale. (2018), Encyclopedia of American Industries: Sporting Goods Stores, Gale Cengage, FarmingtonHills, MI, Vol. 2.

Garg, R.K. and Chan, K.K. (1997), “Service orientation and small business marketing”, Journal ofProfessional Services Marketing, Vol. 15 No. 2, pp. 131-143, doi: 10.1300/J090v15n02_10.

Gr€onroos, C.J.E.J.o. m. (1984), “A service quality model and its marketing implications”, Vol. 18 No. 4,pp. 36-44.

Homburg, C., Hoyer, W.D. and Fassnacht, M. (2002), “Service orientation of a retailer’s businessstrategy: dimensions, antecedents, and performance outcomes”, Journal of Marketing, Vol. 66No. 4, pp. 86-101.

Humphreys, A. and Wang, R.J.-H. (2018), “Automated text analysis for consumer research”, Journal ofConsumer Research, Vol. 44 No. 6, pp. 1274-1306, doi: 10.1093/jcr/ucx104.

Hyland, R. (2018), Ibisworld Industry Report 45111 Sporting Goods Stores in the Us.

Jain, P. and Aggarwal, V.S. (2018), “Developing a service quality scale in context of organized groceryretail of India”, Management Decision, Vol. 56 No. 9, pp. 1969-1990, doi: 10.1108/MD-08-2017-0790.

Janda, S. (2002), “Consumer perceptions of internet retail service quality”, International Journal ofService Industry Management, Vol. 13 No. 5, pp. 412-431, doi: 10.1108/09564230210447913.

Jiang, B. and Zhang, X.J.M.S. (2011), “How does a retailer’s service plan affect a manufacturer’swarranty?”, Vol. 57 No. 4, pp. 727-740.

Kamerer, D. (2014), “Understanding the yelp review filter: an exploratory study”, Retrieved 2019/6/29available at: https://firstmonday.org/ojs/index.php/fm/article/view/5436/4111.

Kim, D. and Kim, S. (1995), “Quesc: an instrument for assessing the service quality of sport centers inkorea”, Journal of Sport Management, Vol. 9 No. 2, pp. 208-220.

IJSMS

Ko, Y. and Pastore, D. (2005), “A hierarchical model of service quality for the recreational sportindustry”, Sport Marketing Quarterly, Vol. 14, pp. 84-97.

Kotler, P. and Keller, K.L. (2006), Marketing Management, 12 ed., Person Prentice Hall, NJ.

Krippendorff, K. (2004), Content Analysis: An Introduction to its Methodology, Sage Publications,Newbury Park, CA.

Lam, E.T.C., Zhang, J.J. and Jensen, B.E. (2005), “Service quality assessment scale (sqas): an instrumentfor evaluating service quality of health-fitness clubs”, Measurement in Physical Education andExercise Science, Vol. 9 No. 2, pp. 79-111.

Leximancer (2011), “Leximancer manual”, available at: https://doc.leximancer.com/doc/LeximancerManual.pdf.

Lin, H.-H. (2012), “The effect of multi-channel service quality on mobile customer loyalty in an online-and-mobile retail context”, Service Industries Journal, Vol. 32 No. 11, pp. 1865-1882, doi: 10.1080/02642069.2011.559541.

Mao, L.L. and Zhang, J.J.J.J.o. A.M.T. (2012), “Exploring the factors affecting the perceived programquality of international sports education programs”, The Case of a Chinese Sports University,Vol. 3 No. 2, pp. 14-35.

Martinelli, E. and Balboni, B. (2012), “Retail service quality as a key activator of grocery store loyalty”,Service Industries Journal, Vol. 32 No. 14, pp. 2233-2247.

Mcauliffe, J.D. and Blei, D.M. (2008), “Supervised topic models”, Paper presented at the NeuralInformation Processing Systems Conference.

McDonald, M., Sutton, W. and Milne, G. (1995), “Teamqual measuring service quality in professionalteam sports”, Sport Marketing Quarterly, Vol. 4 No. 2, pp. 9-15.

Mehta, S.C., Lalwani, A.K. and Han, S.L. (2000), “Service quality in retailing: relative efficiency ofalternative measurement scales for different product-service environments”, InternationalJournal of Retail and Distribution Management, Vol. 28 No. 2, pp. 62-72.

Mukherjee, A., Venkataraman, V., Liu, B. and Glance, N.S. (2013), “What yelp fake review filter might bedoing?”, Paper Presented at the Seventh International AAAI Conference on Weblogs andSocial Media.

Murfield, M., Boone, C.A., Rutner, P. and Thomas, R. (2017), “Investigating logistics service quality inomni-channel retailing”, International Journal of Physical Distribution and LogisticsManagement, Vol. 47 No. 4, pp. 263-296, doi: 10.1108/IJPDLM-06-2016-0161.

Netzer, O., Feldman, R., Goldenberg, J. and Fresko, M. (2012), “Mine your own business: market-structure surveillance through text mining”, Marketing Science, Vol. 31 No. 3, pp. 521-543, doi:10.1287/mksc.1120.0713.

Nguyen, D.H., Leeuw, S. and Dullaert, W.E.H. (2018), “Consumer behaviour and order fulfilment inonline retailing: a systematic review”, International Journal of Management Reviews, Vol. 20No. 2, pp. 255-276, doi: 10.1111/ijmr.12129.

NunezMir, G.C., Iannone, B.V., Pijanowski, B.C., Kong, N. and Fei, S. (2016), “Automated contentanalysis: addressing the big literature challenge in ecology and evolution”, Methods in EcologyEvolution, Vol. 7 No. 11, pp. 1262-1272.

Ofek, E., Katona, Z. and Sarvary, M. (2011), ““Bricks and clicks”: the impact of product returns on thestrategies of multichannel retailers”, Marketing Science, Vol. 30 No. 1, pp. 42-60.

Oliver, R.L. (1980), “A cognitive model of the antecedents and consequences of satisfaction decisions”,Journal of Marketing Research, Vol. 17 No. 4, pp. 460-469.

Papadimitriou, D. and Karteroliotis, K. (2000), “The service quality expectations in private sport andfitness centers: a reexamination of the factor structure”, Sport Marketing Quarterly, Vol. 9 No. 3,pp. 157-164.

Parasuraman, A., Zeithaml, V.A. and Berry, L.L. (1985), “A conceptual model of service quality and itsimplications for future research”, Journal of Marketing, Vol. 49 No. 4, pp. 41-50.

Sport retailquality

Parasuraman, A., Zeithaml, V.A. and Berry, L.L. (1988), “SERVQUAL: a multiple-item scale formeasuring consumer perceptions of service quality”, Journal of Retailing, Vol. 64 No. 1, pp. 12-40.

Ramage, D., Hall, D., Nallapati, R. and Manning, C.D. (2009), “Labeled lda: a supervised topic model forcredit attribution in multi-labeled corpora”, Paper presented at the 2009 Conference onEmpirical Methods in Natural Language Processing.

Rintam€aki, T. (2006), “Decomposing the value of department store shopping into utilitarian, hedonicand social dimensions”, International Journal of Retail and Distribution Management, Vol. 34No. 1, pp. 6-24.

Roberts, M.E., Stewart, B.M., Tingley, D., Lucas, C., Leder-Luis, J., Gadarian, S.K. and Rand, D.G.(2014), “Structural topic models for open-ended survey responses”, American Journal of PoliticalScience, Vol. 58 No. 4, pp. 1064-1082, doi: 10.1111/ajps.12103.

Salton, G. (1989), Automatic Text Processing: The Transformation, Analysis, and Retrieval ofInformation by Computer, Addison-Wesley, Reading, MA, Vol. 169.

Sancihak, L. and Seuna-Chana, L. (2013), “The influence of sport sponsorship communication on sportfans’ rating of retail service quality”, International Journal of Sport Communication, Vol. 6 No. 3,pp. 312-324.

Shilbury, D. (2012), “Competition: the heart and soul of sport management”, Journal of SportManagement, Vol. 26 No. 1, pp. 1-10.

Shugan, S.M. (1983), Branded Variants, University of Chicago, Center for Research in Marketing,Chicago, IL.

Shugan, S.M. (2011), Lecture on Market Definition, Personal Collection of Steven Shugan. University ofFlorida, Gainesville, FL.

Simmers, C.S. and Keith, N.K. (2015), “Measuring retail store service quality: the disparity between theretail service quality scale (RSQS) and comment cards”, Academy of Marketing Studies Journal,Vol. 19 No. 2, pp. 117-126.

Siu, N.Y. and Cheung, T.-H.J. (2001), “A measure of retail service quality”, Marketing IntelligencePlanning, Vol. 19 No. 2, pp. 88-96.

Siu, N.Y. and Chow, D.K. (2004), “Service quality in grocery retailing: the study of a Japanese supermarketin Hong Kong”, Journal of International Consumer Marketing, Vol. 16 No. 1, pp. 71-87.

Smith, A.E. and Humphreys, M.S. (2006), “Evaluation of unsupervised semantic mapping of naturallanguage with Leximancer concept mapping”, Behavior Research Methods, Vol. 38 No. 2,pp. 262-279.

Smith, A.E. (2003), “Automatic extraction of semantic networks from text using Leximancer”, Paperpresented at the Annual Conference of the North American Chapter of the Association forComputational Linguistics, Edmonton, Canada.

Statista (2018), “Sporting goods industry-statistics and facts”, Retrieved 2019/6/29 available at: https://www.statista.com/topics/961/sporting-goods/.

Teas, R. (1993), “Expectations, performance evaluation, and consumers’ perceptions of quality”,Journal of Marketing, Vol. 57 No. 4, pp. 18-34.

Thanawala, S. (2018), “California high court: yelp can’t be ordered to remove posts”, available at:https://apnews.com/531970dd2b8d482d82ec192dd28b0240/California-high-court:-Yelp-can’t-be-ordered-to-remove-posts.

Trocchia, P.J. and Janda, S. (2003), “How do consumers evaluate internet retail service quality?”,Journal of Services Marketing, Vol. 17 No. 3, pp. 243-253.

Tseng, C., Wu, B., Morrison, A.M., Zhang, J. and Chen, Y.-c. (2015), “Travel blogs on China as a destination imageformation agent: a qualitative analysis using Leximancer”, Tourism Management, Vol. 46, pp. 347-358.

Verhoef, P.C., Lemon, K.N., Parasuraman, A., Roggeveen, A., Tsiros, M. and Schlesinger, L.A. (2009),“Customer experience creation: determinants, dynamics and management strategies”, Journal ofRetailing, Vol. 85 No. 1, pp. 31-41, doi: 10.1016/j.jretai.2008.11.001.

IJSMS

Westbrook, R. (1981), “Sources of consumer satisfaction with retail outlets”, Journal of Retailing,Vol. 57 No. 3, pp. 68-85.

Yang, Z. and Fang, X. (2004), “Online service quality dimensions and their relationships withsatisfaction”, International Journal of Service Industry Management, Vol. 15 No. 3, pp. 302-326,doi: 10.1108/09564230410540953.

Yang, Z., Peterson Robin, T. and Cai, S. (2003), “Services quality dimensions of internet retailing: anexploratory analysis”, Journal of Services Marketing, Vol. 17 No. 7, pp. 685-700, doi: 10.1108/08876040310501241.

Yang, Z., Jun, M. and Peterson Robin, T. (2004), “Measuring customer perceived online servicequality”, International Journal of Operations and Production Management, Vol. 24 No. 11,pp. 1149-1174, doi: 10.1108/01443570410563278.

Yoshida, M. (2017), “Consumer experience quality: a review and extension of the sport managementliterature”, Sport Management Review, Vol. 20 No. 5, pp. 427-442.

Yu, C.H., Jannasch-Pennell, A. and DiGangi, S.J.Q.R. (2011a), “Compatibility between text mining andqualitative research in the perspectives of grounded theory, content analysis, and reliability”,Qualitative Report, Vol. 16 No. 3, pp. 730-744.

Yu, C.H., Jannasch-Pennell, A. and DiGangi, S.J.Q.R. (2011b), “Compatibility between text mining andqualitative research in the perspectives of grounded theory, content analysis, and reliability”,Vol. 16 No. 3, pp. 730-744.

Zeithaml, V.A., Parasuraman, A. and Berry, L.L. (1990), Delivering Quality Service: Balancing CustomerPerceptions and Expectations, Free Press, New York.

Zeithaml, V.A., Berry, L.L. and Parasuraman, A. (1996), “The behavioral consequences of servicequality”, Journal of Marketing, Vol. 60 No. 2, pp. 31-46.

Zhang, J., Onal, S., Das, R., Helminsky, A. and Das, S. (2019), “Fulfilment time performance of onlineretailers – an empirical analysis”, International Journal of Retail and Distribution Management,Vol. 47 No. 5, pp. 493-510, doi: 10.1108/IJRDM-10-2017-0237.

Corresponding authorLuke Lunhua Mao can be contacted at: lmao@unm.edu

For instructions on how to order reprints of this article, please visit our website:www.emeraldgrouppublishing.com/licensing/reprints.htmOr contact us for further details: permissions@emeraldinsight.com

Sport retailquality