EXPLORING CUSTOMER SATISFACTION IN RESTAURANTS

Copyright Chris Rochester 2010Exploring a new paradigm and methodology for ‘satisfaction’ research in the Restaurant Industry. The origin of this issue relates to my continued focus upon on the vendor-customer relationship. And whilst I’ve read hundreds of articles and books on the topic I feel that many vendors miss the point, which is: the customer is all knowing and in the ascendant position in terms of the purchase decision, including of course, the re-purchase decision. This, from my perspective, is also apparent within the restaurant industry.My other interest relates to restaurants themselves. The nature of the interaction is a business feeding a customer, which in itself has connotations of nurturing and sustaining life. In that sense it performs the function of the family home, in fact many family homes in one setting. It’s this function that sets it apart from the local supermarket, wherein the consumer, for the most part knows what they are buying. With restaurants the onus of responsibility is with the restaurant and trust with the consumer that the restaurant will ‘get it right’.And whilst the restaurant is a small business that ‘sells’ food, it’s the nature of the unknown interaction, the outcome of the food order, that sets up an intrigue that sets it apart from other small business. Further that the awareness created by the core feature (the food), has an ability to heighten the flow-on to other (augmented) aspects of the interaction such as the attitude of the waiter, the ambience, and so on. However, it’s the re-purchase decision which interests me most. Apart from the cost comparisons of keeping customers, versus acquiring new ones, maintaining a high proportion of existing customers is the accepted mark of business success.

My focus in running business has been to evolve loyal customers, ‘enumerated’ within loyalty a program operating as a subset of a business plan. As with most aspects of marketing the approach is formulaic. Generally the structure is one of a data base that dictates repeat customer unprompted communication on a regular basis underpinned by the marketing plan driven imperatives. Nowadays these outbound contacts are either electronic or by letter, and much less by an outbound call centre. In other words the rationale of loyalty programs is that ‘making contact with customers pro-actively’ is the linchpin that holds on to them. In recent times, and with the advent of the internet, there are a plethora of ways to connect with existing customers including, in the case of electronic goods, embedded web-ware to set in place any number of (sometimes unrequested) approaches to the customer. And whilst this work is about satisfaction, ultimately the viewfinder will turn back to loyalty, once I more fully understand satisfaction.

My challenge with historical loyalty approaches is that the researchers (that I've assessed) have structured a quantitative process that presumably is more acceptable to the business world. One example is a research team which rigorously reviewed the scholarly literature for loyalty and its antecedents, and then attributed a comparative ‘numerical’ scale of importance (based on literature frequency) as input data for the ‘quantitative’ approach. In other words this approach attempted to use research literature frequency data as a reflection of a model that could be applied to the consumer marketplace. No attempt was made to correlate this source of data with real marketplace evidence.

The other challenge with loyalty is that it acts as an emotional humanizing construct i.e. people will like you (as a person, brand or company) and therefore will stay with you. Or they will stay with you if you as a company are warm and friendly and stay in touch with them. As with the loyalty literature the perspective is static. In other words it doesn’t deal with the business-customer interface and interaction either holistically or realistically, instead reflecting on its customers as a body of people with whom there is a preferred status which needs to be maintained.

Customer satisfaction (CS), the subject of this review, is defined in the literature within a frame of how a product and/or service performs in lieu of customer expectation. Such judgement can be against one or a number (balanced scorecard approach) of interactions-transactions. Importantly CS is a notion that talks to the interface and interaction of the transaction. In theory CS can be independent of a carry-over relationship (loyalty). In other words a ‘satisfactory’ or ‘unsatisfactory’ event can be adequately measured against a single interaction. This is a key factor in that defining satisfactory events (from single event exemplars) has the potential to establish a model of behaviour against a cast of single events, and doesn’t depend on ‘past glories’.This is in contrast to loyalty (programs) which looks to a series of ‘loyalty accumulating’ customer interactions. Equally, some of the literature also frames satisfaction itself across a number of events, in the sense of satisfaction with the company or ongoing service. At this point in time I am not inclined to go down this (literature) pathway as this definition of satisfaction is more like an antecedent of, or indeed is synonymic with, loyalty. Further, from this initial position, I intend to develop a methodology that singularises satisfaction as the metric/topic under research, and extricated entirely from loyalty. This decision appositionally reflects the weaknesses uncovered of this definitional-melding methodological dilemma of loyalty researchers in my previous literature review. So even if there is a causal relationship between satisfaction and loyalty, by being able to measure satisfaction, loyalty should change accordingly, or not. Further that with a (qualitatively derived) model of satisfaction established, leads the way to perhaps assess a numerical approach to satisfaction, after which could apply to quantifying a relationship with loyalty, or not. Such approaches have been undertaken in the past, however my assessment is that the qualitative aspect of the work has been almost an excuse to get to the quantitative (survey or algebraic model). In doing so leaves the satisfaction research arena where it is today, which is piecemeal and irrelevant on-the-ground to restaurant businesses. Also, from my commercial research (internet and yellow pages) has not established a significant attendant consultant industry, and especially considering the number of restaurants (that exist in first world countries).

An indirect issue relating to customer satisfaction and business is its dependent nature upon supply and demand. At the extreme a monopolistic business, for instance, may have little interest in satisfaction on the basis that they have already maximised their market and profit. The converse being true for highly competitive markets (that are not exclusively price-driven). This may also be true of restaurants i.e.’ the only one for hundreds of kilometres’.

Most important at this outset is the notion that every interaction between a business and its customers can be testable against a standard or a model of practice. In being so staff need to be instructed (trained) towards that model. It is this notion which frames my ‘why’, in that research literature in business needs to have practical applicability that improves the business offering. I should add that this aspect, application of a model into a restaurant operation, is not within the scope of this review, but is most definitely my longer term aim. One last thought relates to how does a restaurant equate to small businesses in general, that turn over $.35m - $0.5m per year? Both most likely don’t have a strategized business plan (perhaps annual targets), or a level of marketing that extends beyond sending flyers to customers (or in the case of restaurants to nearby households), and so on. In other words both share vague or non-existent customer target and post-purchase tactics.

As a summary point the restaurant customer interaction serves a number of functions including pre and post-purchase environment, tailored products, core and augmented benefits, and also over a range of different customer needs at any one time. Also, some of the literature attempts to extract a meaning for satisfaction in order to further predict repeat business i.e. loyalty. I feel it is this linked-construct or connectedness of satisfaction and loyalty that is part of, or is, the problem with developing a singular, clear-thinking and practical satisfaction model that could be applied to non-corporate businesses such as restaurants . Perhaps others. Further, that developing a clear instrument that correlates satisfaction to return patronage may/is complicated by other factors determining return attendance (upward and downward). Examples of these factors are tourists, special occasions, trying new cuisines, invitees of friends or family or business, or convenient on a particular occasion, a friend or family of the staff. This, however, would not apply to satisfaction - if a Lewin disconfirmatory approach was undertaken in which an expectancy scale could be developed (as previous research has undertaken) that set an individualised base level which of course could apply to any subject.

The other problem is that the modern literature attempts to evolve and use ‘mathematical instrument’ to determine satisfaction. And whilst these are applied against restaurants as a predictor they are not used to improve, or show improvement in a restaurant’s operation. From above it is apparent that much of existing literature proposes a strongly causal relationship between satisfaction and loyalty, which as a philosophy, seems to be widely accepted in the business world. At this point of my thinking being able to more fully understand satisfaction as a customer decision/assessment in a moment in time, should be the target for this research to evaluate. Of interest (only) is whether this satisfaction decision can change over time without further interaction with the restaurant.

Most of the literature on CS emanates from the satisfaction theory work in the early 1900s, which in turn drew upon a narrow definition of CS espoused by the cognitive psychologists of the period who oftentimes derived results from laboratory situations framed by situation specific conditions. (Kivela, 1999). Emanating from this era was Kurt Lewin who framed a much broader ‘Expectancy-Disconfirmation Theory (EDT). Positive disconfirmation occurs if the product or service is better than expected, whereas a performance worse than expected results in a negative disconfirmation. A match between the two leads to confirmation, that is, neither satisfaction nor dissatisfaction. (Kivela, 1999)

Lewin’s work is evidenced widely in much of the literature (also often framed as contrast approaches of satisfaction and dissatisfaction). My challenge with this work is that it is more of a model than theory, essentially because it can’t be disproven. That is it is always correct regardless of the outcome of purchase, and defines a customer not purchasing (or re-purchasing) because, in essence, they either want to or don’t want to. What I feel Lewin’s model provides is a thought framework that forms a core for further experiment which logically could be extended to ‘What is the tipping point for a customer to move from positive to negative (and vice versa) disconfirmation?’ This in turn would seem to be a numerical approach. And critical to this singularity of options (i.e. a numerical approach) is that my review shows that the literature on satisfaction, regardless of the research work up, stance or research problem ultimately utilises a quantitative method (oftentimes pre-empted by interview patterning etc) that utilises survey or questionnaire in order to develop a (regressed) quantitative instrument.

Zeithami et al (Zeithaml et al., 1993) describes it as ‘consensus exists that expectations serve as standards with which subsequent experiences are compared, resulting in evaluations of satisfaction or quality’. In other words the research worldview is a framework of customer expectation (Lewin), scaled in a particular way that attempts to make sense of a decision to purchase. However, the following by Zeithami delineates the current (diverse) bodies of thought on customer satisfaction and dissatisfaction (CSD) customer expectation paradigms, and in doing so, reflects that there is no ‘comprehensive theoretical framework that….integrates these....different....standards’:1. Expectations as Predictions Standard: predictions made by customers about what is likely to happen during an impending transaction or the objective calculation of probability of performance. The dominant standard in CSD.2. Expectations as Ideal Standard: the level at which the customer wants the product to perform. A wants, need and desires orientation towards an idealised model.3. Experience-based norms: what the product should provide based on what the customer thins is possible.4. Comparative expectations: consumer expectations versus other brands.

In other words the observation by Zeithami (and well understood throughout the literature) is that there is no single overarching or even cogent CSD framework. Reflecting on the above ‘satisfaction research history’ there seems to be an ongoing funnel-effect borne from the narrow research and theory emanations of the 1930s that is extremely creative, but within a singular worldview. The further assertion is that there would seem to be a fundamental disconnect between this worldview and the complexity inherent in satisfaction. This may be further complicated by the diversity of the restaurant industry.

Further to this emerging problem I researched to see if any single instrument approaches to the restaurant industry had been achieved, and came across SERVQUAL (Bojanic, 1994) and DINESERV (Stevens et al., 1995) both defined within a CSD framework .

Even though service quality is a factor (antecedent) of CS my search found no correlation even in the article, which for the most part assumed CS to be the natural consequence of good customer service. A similar approach was undertaken by Stevens et al developing their instrument. Of most import is that the research didn’t undertake an indicative-of-the-industry broad sample of restaurants. In the case of SERVQUAL seeks to quantify the difference (gap) between customer expectation of service and the actual service received. A discussion within the Kivela article (Kivela, 1999) follows ‘restaurateurs should not believe that merely meeting or exceeding what is anticipated, e.g. expectations (orPrDp - a CSD construct), will satisfy customers, but should understand that the actual performance (EM in the model, Figures 1 and 2) will have a far greater predictive effect on satisfaction.’The above approach is further supported by Johns and Tyas (Johns et al., 1996) ‘service quality precedes customer satisfaction, and that performance alone is a more useful measure than the performance-expectations gap, because customers may perceive expectations in the same terms as performance and also because expectations are complex structures.’

My reflection on the above is that firstly the same logic could and should be applied to satisfaction itself. That is, satisfaction could be accessed using an appropriate methodology, and more importantly applied (via theory and/or model) within a restaurant to elicit the best that the restaurant can achieve - obviously within the worldview constraints of the theory or model. In other words the Lewin-originated CSD model becomes the CS model. Further to SERVQUAL and DINESERV is that subsequent attempts by the workers themselves, as well as other, have failed to replicate the original work, made more difficult by the non-disclosure of key experimental factor-structures.

One final article by Maddox (Maddox, 1981) reflects a different paradigm to satisfaction, and is based on the work of Herzberg et al (Herzberg, 1959). Known as Herzberg’s Two Factor Theory this approach views satisfaction and dissatisfaction as entirely disconnected and unrelated. As opposed to traditional workers ascribing a ‘continuum of satisfaction and dissatisfaction’ (described further above) this construct embraces that dissatisfaction and satisfaction may co-exist (even to the fullest extent) at any point in time within a person, and on the basis that the factors relating to satisfaction are different to those relating dissatisfaction.

For this research Maddox further includes, from the theory of Swan & Coombs (Swan and Combs, 1976) ‘expressive performance’ which produces satisfaction, and ‘instrumental performance’ which leads to dissatisfaction. The examples provided relate to clothing where by the performance of the item is instrumental-based and the user’s response to it is expressive-based. I should add that the intent of the Maddox article is to reproduce the two-factor portion of the Swan & Combs work to then explore other application. Interestingly Maddox changes part of the original design, which includes replacing deep interview with questionnaire amongst other elements, and not surprisingly cannot duplicate the original results. Thereafter Maddox essentially dismisses the Two-Factor Theory as difficult, time-consuming (of data collection), with ‘interpretation (that is) highly subjective’.

My interest in the Herzberg Two Factor Theory is that it provides a different construct to the traditional dogma of a satisfaction-dissatisfaction framework, and in doing so re-affirms that it’s possible to do so. I should add that there may indeed be no such thing as dissatisfaction (or satisfaction), that disappointment could easily act appositionally to satisfaction.

The above is a brief synopsis of my research of the literature. There are numerous other issues in relation to satisfaction, including the debate on antecedents, industry-wide and national satisfaction tools, and numerous other (frameworks of) definitions. Importantly, and after I re-affirmed in my mind that the research seems to have ‘bogged down’ in CSD approaches, I formed a summary view that there is no current dominant or consensual approach that could evolve a pragmatic and perhaps across-the-board instrument for restaurants.

In turn I reflected a stance ‘to find alternate viewpoints’, examples of which came from Kivela et al (Kivela, 1999), Johns and Tyas (Johns et al., 1996), and Herzberg (Herzberg, 1959) via the work of Maddox (Maddox, 1981). Further that the methodological approaches were/are equally stagnated within the CSD frame, for the most developing a questionnaire, sometimes from exploratory interview, with the unitary intention (via regressive techniques) to evolve a mathematical instrument. Disappointingly, my research cannot find evidence that instruments have been applied across other restaurant satisfaction related contexts, at least not without significant further work. More relevant, and referencing the examples of well known SERVQUAL and DINESERV, there seems to be scant evidence of predictive instrumentality for the fast-food restaurant industry, and categorically none for the traditional restaurant industry.

The outcome from the above literature review and analysis is that I intend to research CS as a unitary item, and even further, detached from dissatisfaction and expectation. Whilst this may seem hypothetical in nature, it is more an attempt to move away from previous constructs, in order to move towards ‘fresh air’.

My overarching philosophy is that ‘science that should start from the facts’ and ‘theory that should be able to be disproved’ (Chalmers, 2006). In doing so will, I feel, set me off on a research path much different to Lewin and his successive supporters. Further from the literature, the questions that are apparent (as of now) are:a. Why don’t research instruments operate across different restaurant environments?b. What are the inherent factors that could enable a restaurant instrument?c. How can a single factor (continuum) be researched to evolve an instrument?d. Can a restaurant environment be ‘created’ to systematically eliminate irrelevant factors versus relevant factors and develop a testable theory?e. Can an instrument be applied to improve the known factors relevant to improving customer satisfaction?f. Can improving customer satisfaction be structured to improve restaurant profitability?

For now I choose option c. ‘How can a single factor (continuum) be researched to evolve an instrument?’ as it has much appeal as an exploratory work, that more importantly opens the way, from a methodological viewpoint, to then work on the unitary ‘satisfaction’.

Intellectual Framework:1. Should I explore CS and ‘how to optimise CS’ it in a ‘laboratory situation’ wherein I can (cross-sectionally) control (by not varying) other factors, including ambience, delivery of the method (interview), note taking or recording, video options, approaches to subject reflections on concepts, process of explore new ideas. a. If so I would apply deep interview as a method to better understand. From this would come themes and patterns across interviewees. b. I probably need to equally review and include themes that imply dissatisfaction. In part this shows no bias, and it may also open up a new way to look at the relation (Leximancer approach) between satisfaction and dissatisfaction. Whether they are connected or not may not matter to the research as it is looking at the optimal degree of satisfaction that can be operationalised ‘in situ’. c. In fact, if it did show a correlation (or dependency) may even be better because:i. Shows no biasii. Shows that, even though a dependency relationship (with dissatisfaction), satisfaction still operates in some implicit way on its own to achieve a level of customer favour. This would be new knowledge (cf literature). I could then design a further methodology against this question i.e. intentionally include dissatisfaction to show that it exists with, and is correlated to, satisfaction - but that satisfaction is the ‘operant conditioner’ that dominantly determines the ‘favourable impression’. This is a combination of CSD and Herzberg, and is simply a ‘thought-piece’ at this point.iii. If it shows that both (satisfaction and dissatisfaction) exist, and no correlation (Herzberg, 1959) - possibilities are that one goes up whilst other down, or that one is less than the other.d. The reason for the work is not fundamentally to evolve a new theory, but to elicit new knowledge that may have the potential to drive new theory. e. The previous ‘satisfaction’ work that I’ve reviewed is researcher reflexive. It may be argued that the approach of evolving a ‘narrow’ theoretical definition of satisfaction without an instrument for either eliciting, or interpreting as, new knowledge, is not new knowledge. It’s just new theory.f. This research is targeting new knowledge. One mechanism I know for creating new knowledge applies to eliciting implicit (internally held) knowledge that in social interaction. (Lagerström and Andersson, 2003). Further to this is that knowledge creation significantly post-dates traditional ‘satisfaction’ research.g. Further to f. above, the social interaction that is described in the research nominates a cohort who have an obvious commonality and framework of thought/communication, which is the company they work for, and their IT role. It may be that this criterion carries over to my research with restaurant loyalty groups, and the like, as a potential cohort.

2. Summarising from the work-up of 1. above the driver in this research is not to look for theory, but to create new knowledge (at least initially) that relates to either CS per se, or CS relating to the restaurant environment. The exploratory question ‘How can a single factor (continuum) be researched to evolve an instrument? may best be explored by first determining discussion topics salient to the research question as outlined by Lagerstroemia and Andersson, 2003, and other workers, within a defined framework of likeminded experimental subjects. This also reflects an approach aligned to the overall philosophy of ‘starting with facts’.

3. From 2. I anticipate emergent themes that relate to the question. At this point it’s worth noting that the exploratory research may have been a Case Study methodology in order to more fully understand and perhaps develop, through some degree of iteration, a precise definition of ‘satisfaction’ within the scope outlined above. A content analysis approach would no doubt lift the understanding of the topic in a short time -frame. However, in doing so may change the nature of the work from exploratory to more formalised - as content analysis requires a degree of pre-emptory structuring. Another approach could be a phenomenology, perhaps providing a further (truer) perspective.

4. Perhaps the point of differentiation is that the exploratory work could show that new knowledge is occurring, even if at an embryonic level, and that the advancement of the work had promise. In any event the opportunity exists to continue with the group, or define a more appropriate one, or several for better rigour, if sufficient information presents. A level of emergent overarching cognition would need to have occurred at this point in order to decide if the work should stay on track, or identifies the need for a new approach, necessitating of course a new research question.

5. Whilst I cannot right now answer the questions emergent from 3. and 4. I can suggest possible methodological approaches (post-exploratory) that I sense would need to addressed. a. Firstly methods such as questionnaire or survey are options at any time from this point onwards. The issue that I see though is staying true to capturing new knowledge that meets the philosophy and scope of the work, both of which I feel are ‘watchers of the research’ that stop it slipping into the flawed platform of previous workers i.e. surveys framed against a plausible, yet non-scientific paradigm. Secondly that if the emergent results indicate that there is no singularity of satisfaction, or likely to be shown, no amount of questionnaires will fix the problem (as they cannot provide foundation data in themselves because they follow the inherent researcher line of thinking in the questions) b. One decision that needs to be considered in moving to a formal research project (i.e. my thesis) is whether to re-locate the experiment to an actual restaurant environment. Most certainly pre-and post-dining surveys have dominated in the literature, but once again these will need to be applied against a rigorous concept as mentioned above. Further, whether or not further rich data is required (after the new knowledge work) is uncertain. It may be that the first data suggests a new or varied cohort is needed for triangulation or a differences approach is required to re-assess the data. At a practical level rich data from interview may be difficult to elicit from restaurant patrons. One option if needed may be (for the restaurant) to offer a free meal or similar for patrons to come in one hour earlier (or at their home, which is more difficult to achieve). c. Of most importance is to optimise the process of building a theory, which is the explicit aim of the work (at this point in time). From my reading Case Study approach would, I feel, provide the most rigorous methodology to firmly establish the emergent data (and not resort to premature survey etc). Further, the literature is rich in outlining approaches to evolving theory from case study (Eisenhardt, 1989, Kennedy, 1979, Woodside, 2003). d. Grounded Theory (GT) presents as the strongest candidate for the methodological approach to achieving the aims of my research, underpinned by prescribed interview (and other) procedures for moving through analysis then to theory, and is often used when existing theory is inadequate.