“Being Good” vs. “Looking Good”
on Online Consumer Review Systems
: from Reactivity to Reputation Management
of Restaurants in Amsterdam
Bomi Kim
Research Master’s Social Sciences Sociology - Empirical/Analytical Track 11221321 (bomik0129@gmail.com) Supervisor: dhr. dr. O.J.M. (Olav) Velthuis Second Reader: mw. prof. dr. G.M.M. (Giselinde) Kuipers
Date of submission: 7th, July, 2018
Abstract
In recent years, the rapid growth of online consumer review (OCR) systems such as TripAdvisor has greatly reconfigured the operating environment for numerous businesses and organizations. As OCRs become a crucial source of information for consumer decision-making, an increasing number of evaluated organizations are compelled to incorporate OCRs into their management practices. This shifting environment raises question about how the evaluated organizations change their organizational practices in reaction to being evaluated on OCR systems and how they make sense of those practices. Taking restaurants in Amsterdam as the example, this study uses in-depth interviews to identify four restaurant practices regarding OCR systems on two dimensions: ‘production management’ and ‘reputation management’. The paper concludes with the call for the inclusion of the agency of the evaluated entities, which invites us to expand both our theoretical framework and conceptual understanding of OCR systems.
“
W
e're kind of forced to play this whole game with everyone because as a restaurant, you have to be highly ranked, hopefully. So, it's a necessary game you have to play. That's the way I look at it. I don't really necessarily like to ask my clients: “Please leave a good review.” It's not so nice to ask. It's not something I particularly enjoy. We'd rather focus on good cooking and good service.”1. Introduction
In recent years, the rapid growth of online consumer review (OCR) systems such as TripAdvisor (founded in 2000) and Yelp (2004) has greatly reconfigured the operating environment for numerous businesses and organizations. From wines (Vivino, 2010), movies (IMDb, 1990), jobs (Glassdoor, 2007) to professors (RateMyProfessors, 1999), a growing number of commodities, services, and people are being reviewed online. Most of these platforms algorithmically synthesize consumer opinions left in the form of written reviews and ratings into aggregate ratings and rankings, using lines of software codes, various weights, and filters. Sometimes the aggregate information is further abstracted and institutionalized in the form of credentials. One example is TripAdvisor’s Certificate of Excellence (since 2011) granted to approximately 10% of businesses in the hospitality sectori.
According to Smith and Anderson (2016), 82% of American adults consult OCRs before purchasing new items and 40% do so almost always. A number of studies have also established the influence of OCRs on product sales (e.g. Chevalier and Mayzlin, 2006; Ghose and Ipeirotis, 2011). Findings of particular relevance are that a one-star increase in Yelp ratings can result in a 5-9% increase in revenue for restaurants (Luca, 2016) and that a 10% improvement in ratings can increase hotel room sales by 4.4% (Ye, Law, and Gu, 2009). As OCRs become a crucial source of information for consumer decision-making, they also become increasingly substantive for businesses and organizations. For instance, Corley and Gioia (2000) illustrate that US business schools experience the immediate impact of rankings in their student applications, external funding, etc. and this coerces them to conform to the rankings’ measurement criteria, sometimes to the extent of prioritizing “looking good” rather than “being good” (ibid.: 330).
On top of its social relevance, the topic of online consumer reviews, ratings, and rankings merits academic attention for various reasons. To begin with, it offers an excellent vantage point to investigate into broader theoretical subjects that characterize contemporary society—such as standardization, quantification, rationality, calculability, efficiency, transparency, and accountability. From a sociological point of view, it is a good topic to study which can contribute to ongoing debates around valuation studies or market devices.
The aim of this study is to gain an in-depth understanding of this relatively young, yet highly relevant social phenomenon of OCR systems from the perspective of the evaluated entities. The majority of previous studies discuss the topic focusing on users (e.g. consumers) or producers (e.g. developers) of OCR systems and their behaviors, motivations, and perceptions (Ayeh, Au and Law, 2013; Mellet et al., 2014; Park and Nicolau, 2015; Filieri, 2016). Other studies discuss OCR systems as a whole on the system level (Jeacle and Carter, 2011; Bialecki, O’Leary, and Smith, 2017; Fong, Lei, and Law, 2016). On the contrary, how the evaluated entities perceive and react to OCR systems is studied to a lesser degree (with the exception of Scott and Orlikowski, 2012; Orlikowski and Scott, 2014; Beuscart, Mellet, and Trespeuch, 2016; Wang, Wezel, and Forgues, 2016). Similarly, Beuscart, Mellet, and Trespeuch (2016:458) argue that “if the impact of online consumer reviews (OCRs) on the demand side of markets is now well understood and measured, not enough studies have examined the reception of this new evaluation method by those who are assessed.”
Moreover, when people make up the evaluated entities, the dimension of reactivity is revealed due to human reflexivity. In other words, OCRs elicit a change of behavior or way of thinking from the very people, businesses, and organizations they assess. Taking these two points into account, this study asks how restaurants in
Amsterdam change their organizational practices in reaction to being evaluated on Online Consumer Review (OCR) systems and how restaurants make sense of those practices. The restaurant industry was chosen as the example because it is known to be the most susceptible to the impact of OCRs together with the hotel industry (as cited in Mellet et al., 2014:6). Meanwhile, Amsterdam offered an adequate and interesting context for the study with its status as the biggest urban area in the Netherlands and its increasing popularity as a travel destination (Van Zoelen, 2017). The following two sections review relevant literature on OCR systems, reactivity, and reputation management (Section 2) and explain in detail how data was collected and analyzed (Section 3). Prior to presenting the findings, this paper sets the context of the research through a brief explanation of the history and relevance of OCR systems in the Dutch restaurant industry (Section 4) as well as the interviewed restaurants’ awareness and perception of OCR systems (Section 5). Sections 6 and 7 illustrate four practices of restaurants vis-à-vis OCRs on two dimensions of ‘production management’ and ‘reputation management’. The paper concludes with the call to expand our theoretical lens from reactivity to reputation management and widen our conceptual understanding of OCR systems to obtain the full scope of what happens on these systems (Section 8).
2. Literature Review
2.1. OCR systems as devices
Market devices refer to “material and discursive assemblages that intervene in the construction of markets”
(Muniesa, Millo, and Callon, 2007:2). From pricing models, aggregate indicators, trading protocols, to merchandising tools, market devices are widespread and indispensable in the workings of various markets. OCR systems are market devices because they reconfigure markets. For example, TripAdvisor has reshaped the way demand (e.g. individual travelers) and supply (e.g. remote hotels) meet in the market for lodgings.
In understanding market devices, Muniesa, Millo, and Callon (2007) argue that we pay attention to the notion of ‘device’, its ability to articulate actions, and thereby bring objects into the analysis. In order to reject the bifurcation of agency that ‘device’ may suggest—humans on the one side and objects on the other—they argue that market ‘devices’ be understood as economic ‘agencements’. While the notion of ‘agencement’ is similar to the ideas of assemblage and arrangement, the word’s shared root with ‘agency’ stresses that there is no implied divide between humans that assemble and objects that are being assembled (Callon, 2007).
The vast majority of previous studies conceptualize OCR systems as new forms of valuation devices that allow any Internet user to share personal opinions on various products and services and thereby influence the decision-making of other consumers. In the French restaurant industry context, Mellet et al. (2014) conceptualize gastronomic OCR systems like LaFourchette as successors of former valuation devices like the Michelin Guide. In the travel industry, Jeacle and Carter (2011) understand TripAdvisor as a source of reassurance for consumers which provides electronic word-of-mouth (eWOM) and personal recommendations from other consumers. OCR systems are especially relevant to singular goods like films (IMDb), books (Goodreads), and dining experiences (TheFork). Singular goods include, but are not limited to, “cultural goods about which there is considerable uncertainty” (Lamont, 2012:204). In more rigorous terms, singular goods are identified by their multidimensionality, uncertainty, and incommensurability (Karpik, 2010). Unlike homogeneous or differentiated
goods which are valued on a single dimension of price, singular goods require a different modality of valuation that prioritizes quality over price. Karpik names this modality of valuation judgment and various tools that facilitate this process judgment devices. OCR systems become particularly pertinent to singular goods because through the aggregation of eWOM, they provide consumers with powerful judgment devices that are readily available and intuitive in character.
Orlikowski and Scott (2014) provide a deeper theoretical understanding of OCR systems by asking what happens when valuations go online. Through a comparison of an online valuation device for hotels (TripAdvisor) with its traditional counterpart (the Automobile Association (AA) hotel scheme), they challenge the main conclusions of previous valuation literature. In doing so, they emphasize the constitutive role of material conditions such as “the open access Internet, algorithms, databases, and browser interfaces” in valuations (ibid.:872). They argue that certain material conditions make up a tool or apparatus which enables and includes certain aspects of an observed phenomenon while restricting and excluding others (e.g. thermometer-heat-length). In this light, materiality becomes constitutive of and inseparable from observation and thus valuation. By including materiality into their analysis, Orlikowski and Scott (2014:883-5) propose a theoretical framework encompassing both new and old forms of valuation devices which they name formulaic apparatus and algorithmic apparatus, respectively. While “formulae are rules based on standards, principles, or prescriptions for achieving particular ends”, “algorithms are executable procedures that solve a particular problem in a finite number of ways”. Thus, formulaic apparatuses of valuation (e.g. the AA scheme) are standardized, well defined, professional, normative, enduring, and episodic (often annual). In contrast, algorithmic apparatuses (e.g. TripAdvisor) are emerging, open-ended, situational, pluralistic, dynamic, and continuous (almost real-time).
2.2. OCRs and Reactivity
Being a form of natural human reflexivity, reactivity refers to the idea that individuals iteratively alter their behavior in reaction to being evaluated, observed, or measured (Espeland and Sauder, 2007:6). The concept of reactivity originated as a methodological concern for the validity of a study. This is because when an individual under a study modifies her behavior due to her awareness of being observed, it contaminates the measurement (cf. the Hawthorne Effect). However, a more substantial approach such as that of Heimer (1985) and Espeland and Sauder (2007) posits reactivity as a general social phenomenon. In their case study of American law school rankings, Espeland and Sauder (2007) demonstrate how law schools redefine organizational goals and reallocate resources in order to perform better on the rankings.
Since reflexivity is fundamental to human life, similar concepts exist in the sociological literature. For instance, Callon (1998) and MacKenzie and Millo (2003) use performativity to show how economic theory shapes the economy beyond merely describing it. Performativity, however, differs from reactivity as it emphasizes the materiality that lies beyond the human mind. Reactivity, in contrast, focuses only on the human agency by conceptualizing how people react to efforts to study them. By emphasizing human reflexivity and its interactive and reiterative character, reactivity stresses the negotiated character of sense-making. Meanwhile, its connection to methodological validity automatically puts the legitimacy of knowledge as the central issue and creates an interesting dynamic to the concept. That is, when a measure is seen as a neutral, independent depiction of the social world, reactivity contaminates its validity and becomes its threat. Reversely, when a measure is
understood as a means for enacting accountability and auditing, reactivity becomes its desired outcome (Espeland and Sauder, 2007).
The shift of apparatuses from being formulaic to algorithmic also reconfigures organizational reactivity to valuations (Orlikowski and Scott, 2014). Formulaic apparatuses evoke changes in human behavior by imposing normative, isomorphic pressure to conform to explicit valuation criteria set by experts. On the other hand, algorithmic apparatuses alter human behavior in a contingent, volatile manner driven by a distributed and disembodied crowd. In their study, Orlikowski and Scott demonstrate how the AA formulaic apparatus evaluates hotels in a standardized, stable, and singular manner. The AA standards are established upon industry experience, made public, and practiced by trained inspectors once every 12-18 months. This coerces hoteliers into developing strategic plans to conform to the AA standards. By contrast, the TripAdvisor algorithmic apparatus evaluates hotels in a dispersed, dynamic, and multiple manner. It produces unsettling, real-time valuations of hotels by synthesizing reviews from the lay pubic all over the world—with algorithms undisclosed in the name of protecting corporate proprietary rights. Unlike a formulaic apparatus, an algorithmic apparatus gives hoteliers no predefined set of criteria to follow, no time for reflection between valuations, and thus no scope for strategic behavior.
In the French restaurant industry, Beuscart, Mellet, and Trespeuch (2016) study how restaurants make use of OCRs for productive activities. Firstly, OCRs elicit improvements in the food or the ambience. This only happens when required changes are very concrete, rather marginal, inexpensive, easy to implement, and do not contradict the fundamental business model. Secondly, OCRs serve as a remote yet powerful monitoring tool as they allow owners and managers to “delegate surveillance to customers” (ibid.:467). When an issue is flagged on OCRs, restaurants can address it on a collective level or trace down the accountable staff whom they may even dismiss in extreme occasions. Restaurants can also mobilize positive OCRs to congratulate, motivate, and encourage their staff. Although OCRs elicit such behavioral shifts from restaurants, Beuscart, Mellet, and Trespeuch find that OCRs are not perceived to be legitimate among restaurants. In fact, restaurants consider OCRs to be malicious, hypocritical, and coming from unqualified sources. Questioning this ‘reactivity without legitimacy’, Beuscart, Mellet, and Trespeuch go on to argue that the voices of restaurants be heard by OCR systems so as to “gradually legitimize the OCR ranking principles and reinforce the mechanisms of reactivity” (ibid.:472).
2.3. OCRs and Reputation Management
Valuations by former consumers matter to organizations not only directly via reactivity, but also indirectly by asserting significant power over the purchase decisions of prospective consumers. This is because OCRs make up the most prevalent and accessible form of electronic word-of-mouth (eWOM), which greatly influences consumers’ purchase intentions of products and retailers (Chatterjee, 2001; Cantallops and Salvi, 2014). Understanding OCRs as eWOM redirects our attention to their marketing implications and the topic of reputation management.
According to Huang-Horowitz (2015), organizations have organizational identity which can be defined as the central, enduring, and distinct features that an organization uses to describe itself. In addition, organizations also have organizational reputation which refers to perceptions that others have about an organization. In a similar note, Wang, Wezel, and Forgues (2016) argue that organizations have market identity which concerns the shared representations of not only internal members, but also external experts of an organization. The role of external
experts, critics, and other agents is to validate the identity of an organization, set expectations of consumers, and facilitate market exchange. The growth of OCR systems, however, has bestowed a part of the certifying role of experts on ordinary consumers. In this light, OCRs pose a new threat to the market identity of organizations. Especially when faced with negative OCRs, organizations juggle between the expected benefits and risks of engaging in public responses. By responding, organizations can defend their market identity and appear attentive to consumer opinions. On the other hand, they face both the risk of drawing additional attention to negative issues and seeming defensive, self-justifying, manipulative, and thereby alienating potential consumers. This dynamic results in organizations being more likely to engage in public responses when consumer devaluations are severe enough to make the benefits of defending market identity outweigh the risks of negative publicity and alienation of potential customers (ibid.).
Levy, Duan, and Boo (2012) identify typologies of consumer complaints and management responses by carrying out a content analysis of 1,946 one-star reviews. They focus on negative OCRs which damage the reputation of hotels because poor reviews are considered more credible, altruistic, and crucial as well as more closely examined and responded to (as cited in ibid.:49). In a neighboring context, Aula (2010) stresses how social media amplify reputation risks for organizations and what new strategies are needed for reputation management in this new environment. He argues that reputation is a valuable, yet highly fragile asset for businesses and the loss of reputation affects market value, competitiveness, and loyalty of stakeholders among others (as cited in Aula, 2010:44). One of the earlier studies on OCR systems by Malaga (2001) views nascent forms of these systems such as Amazon as ‘reputation management systems’. In doing so, he stresses how reputation becomes key to make quality assessments when there is no inspection or personal interaction.
3. Methodology
The proliferation of OCR systems is a relatively recent phenomenon albeit a rapid one. Moreover, few studies have studied the impact of OCRs from the perspective of the evaluated organizations as Beuscart, Mellet, and Trespeuch (2016) argue. Given the relative novelty of both the topic (OCR systems) and the vantage point (evaluated entities), this study uses a qualitative approach based on semi-structured, in-depth interviews. This approach was chosen to enable a detailed exploration on selected topics while still allowing room for new concepts to emerge during the course of analysis. Prior to actual data collection, five pilot talks with a total duration of four hours were conducted to probe possibilities, adjust expectations, and orient the research. Interview topics developed during the talks changed iteratively throughout the study. For instance, the initial questionnaire was restricted to one OCR system, namely Google Maps. However, the findings showed that restaurants’ experiences are not clear-cut but entangled amongst different OCR systems. Therefore, later questionnaires included more platforms.
The population of this study is defined as “mid-price restaurants in Amsterdam with least 30 reviews on either
TripAdvisor or Google Maps”. Price range was used to capture ‘what is at stake’ for consumers. This element was
interpreted as measuring the consumers’ incentive in investing time and energy on OCR systems prior to selecting a restaurant. First, low-price snack bars and walk-in restaurants were excluded. It was expected that consumers would be significantly less likely to consult OCRs before visiting these places because there is little risk involved. High-end restaurants were also excluded from the population. Given that there is more at stake at these
establishments, it was speculated that people would be inclined to use other valuation schemes with more expertise, authority, and legitimacy like the Michelin guide. These assumptions are in line with previous findings that “cheap establishments remain attractive regardless of their online evaluation” and that fine dining restaurants “have sufficient reputation capital to not economically suffer or benefit from online reviews” (ibid.:462-3).
“Mid-price restaurants” were expected to be not only the most numerous in the city (73%)ii, but also the most compatible with OCR systems. Firstly, their price is high enough to incentivize consumers to consult OCRs prior to making a visit. Secondly, their large number (availability) in the city invites consumers to deploy OCR systems to make comparisons before selecting a restaurant. Thirdly, assuming that younger age groups with decent digital literacy would use OCR systems more actively, it is reasonable to limit the population to inexpensive restaurants. This study defined mid-price restaurants as inexpensive restaurants with ‘an average price range of 15-25 euros for a main dish’ that are also ‘classified as mid-range on TripAdvisor’. Restricting the geographical scope to “Amsterdam” was expected to control for the variability in economy, tourism, climate, and industry opportunity among restaurants (adapted from Orlikowski and Scott, 2014). Lastly, the cut-off line of “at least 30 reviews on either TripAdvisor or Google Maps” shows that only those restaurants with a certain minimum level of online visibility on either one of the most prominent OCR systems were considered adequate for the study. With the defined population in mind, the researcher selected restaurants on Google Maps. Given the high compatibility of tourism and OCRs, this study put more weight on the inner districts of the city. Hence, among the seven official districts of Amsterdam—Centrum, Oost, Zuid, West, Noord, Nieuw-West, and Zuid-Oost—samples are concentrated on the first four districts where most tourist attractions are found (figure 1). Nevertheless, restaurants in the immediate vicinity of tourist attractions (e.g. Leidseplein, Rembrantplein) were avoided following Beuscart, Mellet, and Trespeuch’s (2016:463) finding that these restaurants only perceive marginal economic impact of OCR systems.
When selecting samples, restaurants responsive to OCRs on Google Maps were preferred over non-responsive ones. This was done in order to select restaurants with sufficient awareness of OCR systems as well as certain degrees of reactivity to OCR systems. Due to this selection bias, findings are by no means representative of the whole restaurant population in Amsterdam. Especially, the extremely high proportion of responsive restaurants (71%) does not represent the whole population. Rather, this study aims at understanding how restaurants make sense of OCR systems and what drives them to react to these platforms, under the presumption that OCR systems will continue to grow and increase the number of restaurants attentive to such tools.
The final sample consisted of 17 restaurants whose owners or managers volunteered to participate in an interview upon request (table 1). Samples were continuously recruited alongside interviews until when the researcher reached the ‘saturation point’ and new information regarding the research question had ceased to appear. Interviews were conducted between the 20th, March and the 26th, May of 2018 and lasted between 25 and 55 minutes (average: 38 minutes) adding up to a total of 11 hours. Interviews were recorded using a mobile phone after each respondent had read and signed the Interview Consent Form on an iPad. The audio files were transcribed using oTranscribe, iteratively coded and analyzed with Atlas.ti by the researcher.
Figure 1. A map of seven districts in Amsterdamiii.
Table 1. A summary of restaurants in the sample.
* Iens has a maximum rating of 10 instead of 5.
a) This place is a student-owned restaurant with ca. 30 owners/managers. b) The exact title of this respondent’s position was ‘management trainee’.
Of respondent Of restaurant As of 31st, May, 2018 (only when N>=30)
Seats Staff rating N rating N rating N rating N
P1 Owner West 16 90 8 40% 4 95 - - 3.8 124 4.4 66 P2 Owner West 6 40 6 30-40% 4.5 43 - - 4.8 126 4.9 109 P3 Owner/ Manager Centruma) 11 30 30 30% 4 138 - 35 4.3 200 4.4 151 P4 Owner Centrum 1.5 75 17 50% 4.5 173 - - 4.4 138 4.8 120 P5 Owner Centrum 7 45 22 20% 4.5 292 8.4 78 4.4 185 4.6 96 P6 Owner Oost 0.5 140 20-5 5% - - - - 4.4 92 5 94 P7 Owner Centrum 4 45 31 40-50% 4.5 126 9 288 4.6 144 4.9 118 P8 Owner Zuid 1 130 30 30% 4.5 58 - - 4.1 178 4.6 157 P9 Manager Oost 2 480 43 10% 4 51 - - 4.1 200 4.6 67 P10 Manager Centrum 1 50 6 50% 4.5 158 - - 4.2 126 - -P11 Manager Centrum 12 30-40 15 60% 4 435 8.6 169 4.2 151 4.3 162 P12 Owner Centrum 1 45 10 50% 5 103 9 90 4.5 69 4.9 85 P13 Manager Zuid 5 250-60 40-50 20-30% 4 58 8.2 51 4.1 222 4.6 98 P14 Owner Zuid 2 50 9 60% 4.5 376 - - 4.7 203 4.8 71 P15 Managerb) West 1 50-60 20 30-40% 4.5 95 8.7 84 4.2 112 4.7 156 P16 Owner West 3 100 11 10-15% 4.5 38 8.7 64 4.5 102 4.8 70 P17 Owner Centrum 16 80 5 30-40% 4.5 190 9.3 353 4.6 60 4.8 86 Size (#)
4. OCR systems in the Dutch Restaurant Landscape
This section aims at setting the context of this research through a brief introduction of the history and relevance of the four major OCR systems that appeared during interviews. Founded in 2000 in the US, TripAdvisor started as a vertical search engine for travel. While the initial business model was B2B, it made its turn to a consumer-facing site in 2001iv. Currently, TripAdvisor claims to be the world's largest travel site with an average of 455 million monthly unique visitors (in seasonal peak during 2017). It offers over 630 million user-generated reviews and listings of 7.5 million accommodations, airlines, attractions, and restaurants among which 4.6 million are restaurants. Its non-hotel segment (comprised of restaurants, attractions, and vacation rentals) has grown in importance in recent years, accounting for 15%, 20%, and 23% of its total consolidated revenue for the years ended December 31, 2015, 2016, and 2017 respectivelyv. TripAdvisor has become more relevant for the Dutch restaurant industry after taking over Iens in 2015.
Iens was founded in 1998 in the Netherlands as a democratic local restaurant guidebook where all restaurants could receive assessments in contrast to “snobbish” guides such as the Michelin guide and Lekker. 10 year later, Iens had expanded to the most visited website in the Dutch restaurant sectorvi and it started allowing visitors to book restaurant tables via its websiteiii (De Ronde and Sahadat, 2008). By the time it joined TripAdvisor in January, 2015, Iens had an annual turnover of six million euros and a network of more than 20 thousand restaurants (De Waard, 2015). There were over 4,000 reviews published on Iens weekly and around 200,000 yearly (De Ronde and Sahadat, 2008). Iens further consolidated its position in the industry by taking over
Couverts, another Dutch restaurant reservation site, in 2016. With about 1,500 additional restaurant partners
using its reservation system, 5,000 restaurants or about 20% of all restaurants in the Netherlands can now be booked on Iens/TripAdvisor (Posthumus, 2017).
Google Maps was launched in 2005 in the US and was introduced in the Netherlands a year later (Van Ammelrooy, 2006). In 2007, it permitted its users to leave ratings and reviews of local businesses. Prior to that, it offered a collection of reviews from other web sources (Goldman, 2007). The introduction of this feature marked an important turning point that allowed user-generated content to take on a bigger role in Google Maps. From 2010, verified business owners could publicly respond to customer ratings and reviews (Maguire, 2010). This new feature aimed to help businesses grow by getting feedback and building relationships with customers. By 2014, business owners could see and respond to reviews on their smartphones aided by real-time alerts (Poddar and He, 2014).
In 2004, Facebook started as a Harvard University social-networking website. Within a year, it had grown into a popular service across the country and by 2007, it had 30 million registered users globally (Phillips, 2007). Currently, it has an average of 1.45 billion daily active users from all over the world (as of March 2018)vii. Facebook introduced the star-rating system for businesses with physical addresses in late 2013 (Childress, 2014). Facebook differs from other OCR systems in the sense that it allows businesses to opt out from reviews and ratings altogether, both good and bad, and that it discloses how the ratings are calculatedviii.
There are a number of key differences that distinguish TripAdvisor and Iens from Google Maps and Facebook. Firstly, while OCRs lie at the core of the service for TripAdvisor and Iens, they form a complementary part for Google Maps (mapping service) and Facebook (social networking service). Secondly, while the former two platforms rank (order) restaurants, the latter two do not. Thirdly, the former two require that reviews meet a
minimum character count of 100 and 120 each and that they conform to pre-defined formats including assessment standards such as ‘waiting time’ and ‘price quality’. On the contrary, one can simply rate a place with no further explanation on Google Maps and Facebook.
5. Awareness and Perception of OCR systems
Prior to illustrating specific practices of restaurants regarding OCRs, it is necessary to investigate whether restaurants are aware of OCR systems and how they perceive OCRs and OCR systems in general. To begin with, it was discovered that all interviewed restaurants are acutely aware of being reviewed, rated, and ranked online. Not only are they able to name all major platforms on which they are reviewed, but they are also keeping up to date with reviews. Restaurants check OCRs in an irregular and unstructured manner. They do not set time aside to monitor OCRs, for example, on a fixed day of week. All but two restaurants check them on a daily basis and the two restaurants do it weekly. The vast majority react to notifications from OCR systems on their mobile devices every time a new OCR is published and quickly check them on the spot in between other tasks. Time spent on OCR systems is fragmented, dispersed, and squeezed in between larger activities as on other social media. As a result, restaurants feel like they are checking OCRs “all the time”.
The biggest reason for checking OCRs immediately is “because reviews are often written the day after or even the same night that your customers came. So, then it's easier to communicate with staff or communicate with your chefs on what was wrong” as noted by one restaurant owner. Since OCR systems do not ask reviewers for the exact date of visit, restaurants can never know for sure when, sometimes if, reviewers visited the restaurant unless they made a reservation via these systems. Nevertheless, restaurants have an educated guess on the production cycle of OCRs through years of experience as explained by one manager:
It's usually safe to say that they post the review the next day. So, when you get a really bad review and you check with the people that worked the day before, then usually people can say somehow: “Yeah, I think I know who that was. Something happened.”
Hence, it is the underlying assumption that OCRs are produced almost in real-time which invites restaurants to check them in an instant, alert-based manner. This is in line with Orlikowski and Scott’s (2014:884) theorization of OCR systems as algorithmic apparatuses where the valuation cycle is “continuous (almost real-time)” rather than “episodic (often annual)”.
In addition to being aware of OCRs, restaurants perceive OCRs to be very relevant for their businesses. They believe that OCR systems offer an important navigation tool for customers, that they are “a way of people finding out about your business”. Restaurants learn that customers greatly resort to OCRs when deciding whether or not to visit a restaurant mainly through three channels. One, restaurants directly ask customers how they found the place; two, restaurants see people being guided by mobile phones; three, restaurants know this from their own experiences as customers. As in the case of a remote hotel studied by Scott and Orlikowski (2012), restaurants acquire visibility and online publicity through OCRs which lead to more business. In this sense, restaurants perceive OCR systems also as a useful medium for public relations and marketing that complements traditional word-of-mouth. One manager notes,
(…) people find your restaurant through those platforms and otherwise, it would be only (...) word-of-mouth. Then, it would be just that. And now people also find you through those. If you look up [restaurant], the first thing you see is the TripAdvisor one. So, I think it's good for your public relations; [it’s a] good marketing tool as well.
Restaurants believe that tourists use OCR systems as a navigation tool more than do locals. One restaurant owner, half of whose customers are tourists, argues that “[a]s a stranger in the city, you find your way by relying on the opinion[s] of other people.” Another restaurant owner with a similar proportion of tourist customers believes that “[i]n general, if you come to a restaurant which you like a couple of times, you never look at the ratings anymore.” Hence, the role of OCRs is bringing in new traffic. OCRs assert their maximum power prior to making a visit and prior to establishing one’s own assessment. In this light, OCR systems become strong judgment devices that resolve uncertainty and facilitate consumers’ judgments on restaurants (Karpik, 2010; Bialecki, O’Leary, and Smith 2017).
‘Local optimization’ is one strategy that restaurants believe customers use when navigating reality through OCR systems. When the pool of selection becomes sufficiently narrow—e.g. by neighborhood or cuisine— restaurants believe customers tend to select a place based on a flat comparison of ratings amongst restaurants. Customers choose this strategy because of its intuitive and affordable character as echoed by two restaurant owners,
If there are a few choices, they will choose a better-rated restaurant. So, for me, it's very important how people rate us, you know. (…) They search, for example, a vegan restaurant. We have vegan options, a lot. So, somehow, we show up as a vegan-option restaurant. But there are another four around us and people would actually choose a better-rated restaurant.
If you look for a restaurant and you see it has like a 3.2 out of 5 and the other one has a 4.2 and you think: ‘Hey, wait a minute, I think the 4.2 should be much better.’ (…) I think people think that way because that's easy.
Another noteworthy point is that restaurants pay much more attention to bad reviews although in reality, OCRs are positively skewed with the average rating being a 4 out of 5 (Mellet et al., 2014:27). Most concrete examples restaurants provide during interviews are negative. Even among restaurants that do not respond online, negative reviews are sometimes “tempting” to respond to. Restaurant owners and managers take harsh criticism personally as they have put their hearts and souls into their businesses. For instance, one restaurant owner responded to a review for the first and last time because it made him “really pissed.” Moreover, restaurants believe negative reviews influence potential customers more than do positive ones. In short, restaurants display ‘negativity bias’ towards OCRs, proving that negative stimuli have stronger psychological effects than positive ones (as cited in Fong, Lei, and Law, 2016). The same tendency is found amongst French restaurants by Beuscart, Mellet, and Trespeuch (2016:468) who explain it in the context of economic psychology “that the cost of a loss is often much stronger than the benefit of a gain.”
6. Practices on Production Management
Restaurants’ acute awareness and perceived relevance of OCR systems result in restaurants engaging in four different practices towards OCRs. These practices are found on two dimensions that can each be named ‘production management’ and ‘reputation management’. The first two practices are related to staff management and quality improvement and resonate with the two types of reactivity that Beuscart, Mellet, and Trespeuch (2016) find in French restaurants.
6.1. Using OCRs for loose staff management
When restaurants receive OCRs that are either particularly good or bad, they find it important to share and discuss them with staff. Positive reviews are generally shared by the whole staff to boost their morale. Mostly, restaurants take screenshots and share them instantly via group chats with everyone so that the staff “know people like what they do” and “feel good about themselves”. Meanwhile, negative reviews that “sharpen the team” are usually discussed with employees who are expected to know the story. However, reviews do not reveal much about specific staff members as explained by one manager: “(…) nobody wrote, you know, like blond girl, tall girl from Saturday.” Therefore, the tricky task of “matching a critical account with the appropriate employee” needs to be based on the educated guess that OCRs are produced almost immediately (ibid.:467). One restaurant owner sums this up nicely:
The good reviews I would share with everybody. The bad reviews would be more something I would discuss with people that worked that night, “Hey, listen. What happened? Do you know?”
Talking to people “that worked that night”, nevertheless, is not aimed at tracing down the line of accountability. Rather, it aims at setting the record straight and deciding how to deal with the issue as a team. No respondent says that she has ever used OCRs to reward or punish staff members. As a matter of fact, the vast majority either laugh or display discomfort at the question. They perceive the very idea of rewarding or punishing staff to be absurd, arguing that the business is strictly team-oriented. Still others believe that OCRs do not provide legitimate enough grounds for employee evaluation since OCRs are “really subjective” and “there's so much more than just what the people [customers] are seeing”. A manager of a restaurant with around 50 staff members and the owner of a restaurant with 5 employees deliver these points as follows:
Um... punishing, yeah, pff. I think punishing isn't the right word. Obviously, I'll speak to them and give them feedback. But I wouldn't punish someone through something that's been said in an online review.
No. We... the business that we are in is a very team-oriented business. Nobody can do it on its own. Some mistakes can happen. We work with people, so, it's normal to make mistakes. (…) and not always mistakes are being reflected in reviews, either. So, it's also not fair that if one review gets a negative thing in it that we punish a member.
This rebuts Beuscart, Mellet, and Trespeuch’s (2016:467) argument that OCRs can be “an extremely powerful staff management tool” that allows even for a delegation of surveillance and a dismissal of employee. This also does not resonate with Scott and Orlikowski’s (2012:38, brackets in original) argument that TripAdvisor reconfigures accountability in the travel sector to the extent that it renders “the hotelier the (mostly) passive recipient of distributed judgement.”
Restaurants in Amsterdam do not use individual OCRs as an employee performance measure, but they do utilize the tangible character of aggregate ratings and rankings for effective staff motivation. The underlying belief is that however partial and subjective individual OCRs may be, when a sufficient amount is put together, they do reflect a restaurant’s performance. Next to three respondents that vaguely aim at improving their overall performance on OCR systems, five have (had) specific goals. One restaurant in Zuid with several hotels nearby, aimed to be ranked #1 of all restaurants in Amsterdam on TripAdvisor within the second year of business (and achieved it). The owner of a recently opened restaurant in Centrum states,
We are now, I think, listed in TripAdvisor in the 194 which is really good—so, the top 200. Our goal is to be in the top 100. You can also use it as a motivation tool: “Let's get into the top 100 with real reviews, with real responses, with real good experiences and then we do the job.”
Nevertheless, the inherent nature of algorithmic apparatuses allows little room for any strategic behavior. Including the restaurant in Zuid that climbed up over 100 ranks to be ranked #1 on TripAdvisor, several restaurants realized their goals on OCR systems. However, when asked how, all they say is “hard work by all of us”. Restaurants cannot do much that will translate directly into improvements in ratings and rankings. What they can do is stick to the principle and hope that their hard work will be reflected on OCRs. Therefore, setting goals with ratings and rankings is an effective, yet still loose form of staff management. One restaurant owner notes,
So, what we deliver here [in the restaurant] is what we aim for and the reviews come along, in that way. So, it's not that we have said "we need to be having a 9 on Iens, then we're doing good." No. We need to do good and the reviews follow along with that.
6.2. Making quality improvements from OCRs
Obviously, restaurants are interested in adapting customer feedback to improve their businesses. After an OCR that complained that “the staff always go smoking and do not wash their hands”, a restaurant owner told the staff to “walk all the way around the back” when going for a smoke. Another restaurant in Centrum which had a surcharge on different egg styles, cancelled it after a couple of OCRs commenting on it:
We had two complaints about it. I, as a boss, had even forgotten about the whole surcharge thing and then obviously your customers are allergic to these situations, which I didn't like also. So, we cancelled it. So, no more surcharge for having your eggs in another way (chuckles).
However, in reality, restaurants say they do not adapt that many changes from OCRs. Many respondents struggle or fail to remember concrete changes they have made to their places, menus, amenities, or management practices from OCRs. They lay out various reasons for not mobilizing OCRs for quality improvement so actively. To begin with, “at the end, reviews are reviews and they’re very subjective”. Instead of objectively reflecting a restaurant’s performance, OCRs are believed to portray subjective expectations and prior experiences of individual reviewers. Many are also quite generic (e.g. “Bad service!”) and not specific enough to act upon. Moreover, some criticisms are simply invalid as they stem from a failure to understand the restaurant’s concept (e.g. “The dishes are too small” at a Spanish tapas restaurant).
On an aggregate level, it is simply not feasible to react to all the comments since they often contradict each other. For example, while some people enjoy the “lively” atmosphere, others complain that it is “too loud”, making it impossible to act on both. The sheer volume of complaints coming in, in proportion to the number of customers, also makes it difficult to take everything into consideration. To borrow Orlikowski and Scott’s (2014:889) analogy, OCRs are akin to stones slung through a thousand slingshots every day by unknown adversaries, from all possible directions. This compels restaurants to be selective and somewhat strict in adapting changes from OCRs.
Not surprisingly, restaurants act on points that are recurring, reasonable, detailed, small, and low-cost. For instance, restaurants gladly remember to leave plates until a table is completely finished or stop using metal sponges in the kitchen. However, a restaurant that identifies itself to be lively and vibrant will not turn the music down despite repeated complaints on the noise. Restaurants emphasize the need to be especially strict with feedback that collides with what they believe lies at the core of the business and therefore is unnegotiable. Note the resolute voices of owners of two restaurants with clear concepts:
If there's any sort of like same comment through a couple of reviews, you use it, of course. But if it's different to the way you want your business to be, (…), then you don't use it. Then, you [reviewer] should go to a different restaurant.
It's just small changes on plating or something. But on taste, we don't do a lot. We think we're doing okay and we have to believe in that. We have to believe that we're sort of right on taste.
This confirms Beuscart, Mellet, and Trespeuch’s (2016:467) finding where “the adjustments made by restaurants are both very concrete and rather marginal. The changes made are inexpensive, easy to implement and do not involve a transformation of the organization.” It also confirms Orlikowski and Scott’s (2014:886) finding where hoteliers “selectively revise their everyday practices in response [to user generated contents]”. Here, the selective attitude stems from the innate character of an algorithmic apparatus which generates dynamic and multiple accounts of a hotel by the lay public. This is in stark contrast to a formulaic apparatus which produces a stable and singular account of a hotel grounded in industry expertise. Faced with continuous assessments that pull the organization in different directions at the same time, hoteliers as well as restaurant owners and managers have no choice but to be eclectic.
The selective reactivity observed with algorithmic apparatuses of valuation can be named as ‘deliberated reactivity’ in contrast to ‘mechanical reactivity’ of formulaic apparatuses. With mechanical reactivity, each and every assessment gets translated passively and almost automatically into concrete organizational changes. On the other hand, with deliberated reactivity, assessments will first be contemplated, doubted, and juxtaposed with an organization’s identity. The cost of changes they require will also be analyzed. Only regarding OCRs that pass a series of rigorous tests, some form of reactivity may ensue.
One crucial point yet to be made is that restaurants argue that there is no fundamental difference between the way they would receive customer feedback offline and online. In other words, the visibility and publicity of OCRs do not compel restaurants to prioritize them in any way over offline feedback. In fact, OCRs easily get conflated in their heads with face-to-face conversations with customers. Nevertheless, OCR systems still contribute to the overall quality improvement of restaurants by their mere existence. The amplified risk of customer
dissatisfaction heightens the tension in everyday production practices. Below two quotes from a restaurant manager and a restaurant owner explain these points:
But if you can do something about it, you should, to avoid the same mistake or the same problem later on. (…) But that doesn't matter whether it's a review online or whether it's an angry customer that's here, you know?
And these online platforms help for business owners to work for something and to everyday keep on striving for the best quality to deliver to the customers. Because if one day you don't do your best, it might affect your standards.
7. Practices on Reputation Management
Besides the production of goods and services, OCRs also spark changes in restaurant practices regarding their ‘reputation management’. This is because OCRs do not end as one-off assessments but constitute an organization’s reputation which influences purchase decisions of potential consumers. This dimension consists of two practices: soliciting OCRs and responding to OCRs.
7.1. Soliciting OCRs
Based on their understanding of the importance of OCRs, seven out of 17 respondents ask customers to review their restaurants online. Four restaurants occasionally ask customers they have a good click with; one hands out TripAdvisor cards with bills to all customers; and the remaining two do both. The reason for encouraging customers to write OCRs is simply the resulting online publicity that is expected to bring more business. Obviously, restaurants want positive publicity. Hence, they approach customers selectively as one manager explains,
Well, if I ask someone like "Oh, leave a review for us" or whatever, I'm not going to say that to the table that's angry or that didn't have a good night or... you know. I would say that to the people that I know that had a good time.
Despite the expected economic benefits, however, restaurants are generally very hesitant to ask customers for OCRs mainly for two reasons. First, restaurants are afraid to appear “pushy”, “invasive”, or “desperate” to their customers. Second, restaurants feel uncomfortable as reviews resulting from solicitations are not fully “honest”. One manager of a restaurant at a touristy corner of Centrum, who finds TripAdvisor reviews highly important, explains her discomfort by comparing it to “asking for tip[s]”—in a country with no tipping culture. Another restaurant manager explains that he sometimes thinks about soliciting reviews from satisfied customers, but never actually does so because it seems “kind of cheap”. Yet another owner of a restaurant situated in a slightly remote area in West, thus for whom being visible online is crucial to attract customers, speculates that his reluctance stems from values of Dutch culture—both that of respecting individual boundaries and being honest:
I don't know, maybe it's a little bit of a Calvinistic thing in me but... I don't do it. (…) But like I say, if I'd be a clever, hard businessman then I'd probably put on the bills: “Did you enjoy? [Please leave a] [r]eview. If you didn't like us, please tell us.” You know, something like that. But I think it's a little bit too much. (...) I know it's
very important, but I don't like to push people. It must come from people themselves. It must be honest. If I look at [competitor], he has 500 reviews but I know that he asks people.
When the expected gain exceeds moral reluctance, restaurants start soliciting OCRs to customers. One restaurant owner that has recently given in to the growing importance of TripAdvisor and started asking customers, illustrates the clash of the perceived necessity and moral reluctance and describes the triumph of the former as “a necessary evil”:
It [soliciting OCRs] is not something I would typically love to do, you know, because... yeah. As far as I'm concerned, it’s a necessary evil. I mean, the whole rating game, online game. (…) We're kind of forced to play this whole game with everyone because as a restaurant, you have to be highly ranked, hopefully. So, it's a necessary game you have to play, that's the way I look at it. I don't really necessarily like to ask my clients: “Please leave a good review.” It's not so nice to ask. It's not something I particularly enjoy.
Restaurants that do encourage customers in defiance of moral discomfort, nevertheless, still emphasize the importance of treading lightly. To ease off the possible burden on customers, restaurants choose to approach them in a light rather than a serious tone.
And I don't want to kind of bother people too much with it. I mean, well, if someone wants to review, they will do it. And if someone is really positive, then maybe in a way of like making a little joke "You're welcome to put in online if you'd like!", you know? But... really saying "Oh, can you help us by..." then I think that would go one step too far.
7.2. Responding to OCRs
A more common practice related to reputation management is engaging in public responses on OCR systems. If soliciting OCRs from customers is managing an organization’s online reputation with others’ words, this would be doing so with one’s own words. 12 out of 17 restaurants respond online albeit with varying degrees. Five restaurants respond to all OCRs on selected platform(s). That is, while one restaurant owner responds to all reviews on all major platforms, another does so only on TripAdvisor which is their main platform. Another five respond to OCRs occasionally and randomly (“If I have the time, I do it.”) and the remaining two respond only to negative OCRsix.
Revisiting the responding behavior
The foremost point to remember about responding to OCRs online is the visibility it entails. It is the instant reach and scope of readership (spatiality; that anyone in the world with Internet access can see everything) and the preservative character (temporality; that everything seems to be engraved online forever) of OCR systems that set the first rule of the game. When restaurants respond to OCRs, they are never just talking to reviewers alone. Instead, they are making a speech to countless potential customers out in the world whose decisions hinge on those OCRs. Below are quotes that effectively capture the public nature of responding to OCRs, offered by two restaurant owners that do and do not respond respectively:
You respond to him, but uh… a little bit indirectly, you are basically responding to everybody in the whole world that is going to read that article [response].
If you're a singer and you're going to sing in front of a million people, you'll be a little bit… [nervous] you know? It's different when you sing to yourself at home. (...) If it was me responding, I would be thinking how to write it in a nice way because I know people can read that. It's open. People can read it. So, it would affect me if I was writing back.
Besides the visibility, the delay in time between an event and a resulting OCR also changes the nature of responding. When a customer complains about something at the restaurant, restaurants can take all possible measures to solve the issue and make the customer happy. However, when the issue is taken away and presented only later as an OCR, restaurants do not get that chance. The visibility and delay in time together redefine the simple act of responding to be more strategic. It becomes less about replying to the reviewer and more about speaking to bystanders (potential customers). It becomes less about addressing the issue and more about managing the possible repercussions of the review. One restaurant owner who diligently responds to all OCRs notes,
If he would be sitting here, you could talk to him like “I'm sorry about it. Here's a cup of coffee” or “Have a beer for me” or “You don't have to pay for the meal” or whatever. But you can't because he's gone. So, that's the big difference. Then you have to make sure that what you write is for other people that are going to read that, to influence the other people that they will come to you and not think: ‘Oh, the food is bad there.’ So, there is a very big difference, very, very big difference basically in what you're trying to do.
Three underlying motives of responding online
The visibility and delay in time of OCRs result in a strategic responding behavior imbued with various motives. During the interviews, three motives were identified. Firstly, restaurants respond online to signal professionalism to the crowd—that they value customer opinions. Regarding positive OCRs, they do this by reciprocating and appreciating the time and effort of reviewers. With negative OCRs—especially those with insufficient information on platforms with no minimum character counts (cf. Google Maps, Facebook)—they probe into the issue and demonstrate sincere interest and dedication in improving themselves. Figure 2 offers a real example of a one-star rating with no explanation (left) and the restaurant owner’s response (right) which is driven by the first motive.
[OCR on Google Maps] ★☆☆☆☆
[Response]
Dear [reviewer], thank you very much for rating us. In order to understand your experience and rating in more detail would you mind to contact me via [email address] or via [phone number]? It would mean a lot to us to make it right.
Figure 2. OCR and response retrieved from Google Maps on 28, June, 2018.
Demonstrating professionalism by responding online is believed to be of great importance among restaurants. Some even believe that this brings them a competitive edge over nonresponsive restaurants. As one restaurant owner who finds it important to respond to all OCRs on all major platforms puts it:
I believe that when a visitor comes by and you see kind of a same rating [similar ratings] but there's one place responding to the customers, I think it could trigger somebody to definitely enter that place because
there is a dialogue between the place and its customers. (…) I would appreciate it. If I see that a business is replying back to the feedback, I would say: “Oh, they care about the customer.”
TripAdvisor officially vouches for this motive as wellx: “Responding to reviews clearly demonstrates—to both former and prospective guests—that you are interested in feedback, and that you take customer service seriously.”
Secondly, restaurants publicly respond to negative OCRs to negotiate their online reputation. This is commonly done by explaining what the situation was like, what has or will be done about the issue. The ultimate aim here is to mitigate the impact of these devaluations—that of alienating prospective customers. A restaurant on a busy street in Amsterdam Centrum offers a nice example. Once, they accidentally included a piece of metal sponge in a customer’s salad. Despite sincere apologies, the incident resulted in a permanent display of a possibly misleading review on TripAdvisor (figure 3-left). In her response, the manager explains that “a big piece of sharp metal that could have cut the reviewer’s throat” was actually “a part of a metal sponge” and promises to stop using metal sponges altogether to prevent similar incidents (figure 3-right).
[OCR on TripAdvisor (trimmed)] ★☆☆☆☆
Stay away! I found metal shape in my salad!
Hey I have been at [restaurant] and ordered a salad. Suddenly, I felt something sharp in my mouth!!!! It was a sharp piece of metal!! (…) I definitely don’t think you should it there! A big piece of sharp metal could have cut my throat!!! Do not recommend!!!
[Response (trimmed)]
First of all I would like to thank you for visiting our restaurant. (…) What you found in your salad was most likely a part of a metal sponge which we use to clean our pans- however we are now going to change to different types of sponges which will never leave pieces behind.
On behalf of [restaurant] and our team, I sincerely apologize for your experience, and hope to welcome you another time.
Figure 3. OCR and response retrieved from TripAdvisor on 08, June, 2018.
With the detailed account and such a promise, the manager hopes to minimize the worries of potential customers and reassure them that it is “safe” to visit the restaurant:
The most important thing in my opinion is just to make sure they know that you're doing something about it. (…) Even if she doesn't come back, at least other people that see the review and see that we did something, they will know: 'Oh, we're coming here. They must be safe', you know?
Sometimes restaurants move beyond ‘explaining’ to ‘contesting’ negative OCRs. This can be done when there is “a factual misstatement” as TripAdvisor puts itxv. For instance, one restaurant owner once received a one-star review where the reviewer had ordered a “double”. Having no “double” whatsoever on the menu, the owner was suspicious of whether it was true or meant to be directed at his restaurant. By writing “[w]hat double did you order as we have nothing on the menu that is double” (retrieved from Google Maps on 05, May, 2018), he opens doors for all readers to doubt the credibility of the review:
So, I responded to that and I said “Be more specific and let me know. I'm sorry to hear about it. And what did you order? We don't have a double.” (…) You know, you have try to make sure that the people that read it think 'Oh, maybe he wasn't even there. What was it?' See, you have to influence it. That's a very strong influence and you have to use it.
The second motive of responding confirms Wang, Wezel, and Forgues’ (2016) argument that organizations respond to consumer devaluations to mitigate consumer devaluations and defend their market identity. That is, when an OCR damages an organization’s reputation, the organization can refute such evaluation to protect its market identity. TripAdvisor also encourages this motivexv: “It’s generally a good idea to respond to reviews that are negative, as well as those where you can correct a factual misstatement or write about an action you’ve taken to correct problems addressed in the review.”
Thirdly, restaurants can go on to appropriate the response function on OCR systems as a promotion channel. For example, responding to a five-star review, one restaurant owner goes beyond signaling professionalism to advertising his products: “The mixed platters to share are one of the favourites” (retrieved from Google Maps on 08, June, 2018). He explains,
A review like this, he is clearly really happy. I was not here. I don't know who it is. But still, I like to share because I think the shared platter… it's nice to let people know that we do that. So, I use the response also maybe to add a little note so people know: ‘Oh, they have shared mix platters. Okay, great. It's not only breakfast or only lunch.’ I think we can communicate this as well.
Another restaurant owner aims for the ‘imprinting’ effect with the responses. By closing his responses with positive points, he aims to imprint readers with a positive image of the restaurant. For instance, he captures “tasty Cod” in a two-star review otherwise completely negative (figure 4-left) and remembers to stress and promote their “fresh piece of Cod” at the end of his response (figure 4-right). TripAdvisor encourages this strategy as wellxv: “Refer to the reviewers’ positive comments about your business to both personalize your response and reiterate the compliment to your potential visitors.”
[OCR on Google Maps] ★★☆☆☆ Cod was very tasty! However chips were not great, mushy peas were cold and came in a very small 5cm diameter 1 cm deep dish. I've had many better fish 'n chips. Not really recommended.
[Response (trimmed)]
(…) We are very sorry the Mushy Peas came out cold, that should not happen. Thank you for telling us straight away, so we were given the chance to make it up straight away and get you a portion of hot Mushy Peas.
(…) Thanks for leaving a review, always appreciated. Glad you like a fresh piece of Cod!
Kind Regards, [restaurant]
Figure 4. OCR and response retrieved from Google Maps on 08, June, 2018.
Below is a part of the interview in which the restaurant owner explains how and why he uses this strategy when responding to OCRs:
And then, sometimes there's some positives as well, so emphasize in that as well. It's really important. In the negative reviews, there's sometimes… from time to time, there's a positive thing as well. "The food was really great, but the service was really slow" or something. So, always [mention] "Luckily, you like the food." (…)
Researcher: Is there a reason for doing that?
Yeah, emphasize on the last part of your comment on the good things. That's the thing that's going to stay in the mind.
Researcher: Of the reviewer or the readers?
8. Discussion and Conclusion
The vast majority of previous studies see OCR systems as new valuation devices in contrast to their more traditional counterparts (Jeacle and Carter, 2011; Mellet et al., 2014; Orlikowski and Scott, 2014; Beuscart, Mellet, and Trespeuch, 2016; Bialecki, O’Leary, and Smith, 2017; Wang, Wezel, and Forgues, 2016). This approach introduces a divide between evaluat-ing users and evaluat-ed objects and a resulting focus on the former. Still others understand OCR systems as social media platforms where people with common interests share information in the form of user-generated content (UGC) (Scott and Orlikowski, 2012; Ayeh, Au and Law, 2013; Filleri, 2016). The latter is no less user/consumer-centered than the former.
Both approaches fail to deliver the full scope of the changes that OCR systems evoke when people, businesses, and organizations make up the evaluated entities. This is mainly because the agency of these evaluated entities is excluded from the picture. By asking how restaurants in Amsterdam change their organizational practices in reaction to being evaluated on OCR systems, this study has successfully brought their agency into analysis. This resulted in the call for the expansion of both our theoretical framework and our conceptual understanding of OCR systems as a whole.
The first two practices on the dimension of ‘production management’ are captured using the theoretical framework of reactivity. Firstly, restaurants use OCRs for loose staff management. Especially, some make use of ratings and rankings to set tangible goals for effective staff motivation. However, unlike previous studies by Scott and Orlikowski (2012) and Beuscart, Mellet, and Trespeuch (2016), this study found that OCRs do not provide powerful enough a means for stricter staff management. Generally, restaurants are resistant to using OCRs as a performance measure mainly due to their subjective and non-representative character.
Secondly, restaurants use OCRs to make quality improvements. When implementing organizational shifts from OCRs, restaurants show greater acceptance towards feedback that is recurring, convincing, specific, low-cost, and rather marginal for the business. This selective attitude stems from the multiple and variable character of assessments made by the lay public using algorithmic apparatuses. In contrast to ‘mechanical reactivity’ of formulaic apparatuses, this eclectic attitude can be named ‘deliberated reactivity’. This pattern confirms the findings of Scott and Orlikowski (2012) and Beuscart, Mellet, and Trespeuch’s (2016). On top of this, OCR systems elicit a general improvement in production activities by amplifying the risks of customer dissatisfaction.
More strikingly, OCRs reconfigure restaurant practices concerning ‘reputation management’. Reactivity, however, ceases to be an effective theoretical lens here since restaurants go beyond merely being ‘reactive’ to particular assessments. Rather, they go on take a ‘proactive’ stance towards these systems. Firstly, some restaurants solicit OCRs from customers in order to improve online publicity which would lead to more business. Albeit the expected benefits, restaurants are generally reluctant to ask for OCRs either for fear of overburdening customers or manipulating reviews. This compels restaurants to weigh up the expected gain against moral reluctance. Those who give in to the “necessary evil”, as expressed by one respondent, still remember to take a light approach with their requests.
Secondly, some restaurants respond to OCRs publicly. The foremost rule of engaging in public responses is keeping in mind that things are open for everybody to see. On top of this, OCRs entail a delay in time, which leaves little room, if any, for the complaint to be solved. The visibility and delay in time collectively redefine the simple act of responding into a strategic act with multiple, and possibly ulterior, motives. The three identified motives
