Customer service quality and benchmarking in public transport contracts

Hensher, David A. Article Customer service quality and benchmarking in public transport contracts International Journal of Quality Innovation Provided in Cooperation with: Springer Nature Suggested Citation: Hensher, David A. (2015) : Customer service quality and benchmarking in public transport contracts, International Journal of Quality Innovation, ISSN 2363-7021, Springer, Heidelberg, Vol. 1, Iss. 4, pp. 1-17, https://doi.org/10.1186/s40887-015-0003-9 This Version is available at: https://hdl.handle.net/10419/176478 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ EMPIRICAL Open Access Customer service quality and benchmarking in public transport contracts David A Hensher Correspondence: [email protected] Institute of Transport and Logistics Studies (ITLS), The University of Sydney Business School, The University of Sydney, Sydney, NSW 2006, Australia Abstract As contracting of public transport services increases in sophistication, there is a growing focus on an increasing number of key performance indicators that emphasis service quality. Although contracts won under competitive tendering or by negotiation are assessed on a number of evaluation criteria, cost efficiency still remains the main basis for selecting a preferred operator. There has been a limited effort to identify the service quality influences that really matter to users of public transport. Ways of incorporating the packaging of service quality offer an improved and behaviourally richer way of representing the role of underlying dimensions of quality in establishing how well a contracted service is delivering services to satisfy customers. In this paper we present a way of doing this using a construct called a Customer Service Quality Index (CSQI), in which a stated preference survey together with actual experience in using public transport is used to obtain preference weights for each significant attribute defining service quality, and which is used then to establish a CSQI for each sampled user, and by aggregation, the performance on service quality of each operator. Such a measure should be considered by regulators when both assessing the merits of each operator’s bid in order to avoid the real risk that cost efficiency dominates at the expense of gains in service performance and in ongoing monitoring of performance. Keywords: Customer service quality; Benchmarking; Bus contracts; Stated preference surveys; Logit models Background As contracting of public transport services grows in sophistication, there is a growing focus on an increasing number of key performance indicators that emphasis service quality. Although contracts won under competitive tendering or by negotiation are often assessed on a number of evaluation criteria, cost efficiency still remains the main basis for selecting a preferred operator. The reasons are in part linked to the need to reduce costs through competition for the market (in tendering), in contrast to other models of contract awarding. In both tendered and negotiated contracts, the ideals of performance improvement are promoted, yet it is often the case that this is poorly understood, and where it is taken as a precise measure of output it is limited to a few measurable criteria that often are defined by supply side measures (e.g., on-time running and accidents per 100,000 operating hours). There has been a limited effort to identify the influences that really matter to users of public transport; however the efforts have been based on simplistic measures of © 2015 Hensher; licensee Springer. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. Hensher International Journal of Quality Innovation 2015, 1:3 http://www.jqualityinnovation.com/content/1/1/3 satisfaction measured of a Likert scale (typically from very unsatisfied to very satisfied), where the contributing influences are treated as independent and hence additive in their impact. In reality, users of public transport purchase a package of attributes that define perceptually what matters to them. Ways of incorporating such packaging of service quality offer an improved and behaviourally richer way of representing the role of underlying dimensions of quality in establishing how well an operator is delivering services to satisfy customers and also their contract obligations. Studies, for example, by Cirillo et al. [1], dell’Olio et al. [2], Eboli and Mazzulla [3] have been critical of methods that treat each underlying service dimension as if it can stand alone in the way it is assessed as an influence on public transport performance (or indeed for any performance of any market delivered service). Bolton and Drew [4] and Boulding et al. [5] are typical examples of customer satisfaction studies that promote attribute packaging in determining the value of service. In this paper we present a novel way of doing this using a construct developed by Hensher some years ago called a Service Quality Index (SQI) (see [6]), in which a stated preference survey together with actual experience in using public transport is used to obtain preference weights for each significant attribute defining service quality, and which is used then to establish an SQI for each sampled user and by aggregation the performance on service quality of each operator. Such a measure should be considered by regulators when assessing the merits of each operator’s bid in order to avoid the real risk that cost efficiency dominates at the expense of gains in service performance. The focus of this paper is quantifying service quality from a users’perspective, in a way that weights the relevance of each source of service quality, and to use this information to obtain a single measure of service quality, that we call the customer service quality index. We use data from metropolitan and non-metropolitan bus operators in New South Wales (NSW) to demonstrate the way the method can be used in contract negotiation and ex-post monitoring of performance leading up to contract renegotiation or tendering. Concerns about traditional Likert scale metrics of customer satisfaction Service quality in the context of customer experience is typically identified using a Likert scale in which consumers are asked to indicate on a scale (such as from very satisfied to very unsatisfied), how satisfied they are with a specific attribute defining a class of service. Sometimes these satisfaction measures are weighted by a reported importance attached to an attribute, often referred to as the Fishbein-Ajzan importance-satisfaction scale. This approach falls short in two major ways of creating a robust customer service quality index. First, there are interpretations of scale issues with surveys which rely on respondents marking off the different aspects of service quality. We do know that ‘very satisfied’to one user might be rated as ‘very dissatisfied’from another user on exactly the same service. Second, there are independent issues with the way in which respondents are asked to consider the different aspects of service quality. The literature [7] is clear that users do not find all aspects of service quality equally important and, indeed, may be extremely satisfied with an element which is of low importance. Within the broader context of SERVQUAL that has had a dominating influence in consumer and marketing research, Buttle [8]) provides an extensive review and critique Hensher International Journal of Quality Innovation 2015, 1:3 Page 2 of 17 http://www.jqualityinnovation.com/content/1/1/3 of the SERVQUAL method first introduced by Parasuraman et al. [9]. The approach using ordinal scaled (Likert) metrics performs analyses with methods suited to intervallevel data (factor analysis) [10]. It has been criticised on many grounds including that interdependencies among the dimensions of quality are difficult to describe. Importantly it has been criticised for focusing on the process of service delivery rather than outcomes of the service encounter (Richard and Allaway [11]. For this reason, the best approach to the customer service index would be adopt a customer service quality index (CSQI) approach in which an methods such as a stated preference (SP) experiment investigates with users their response to variations around the current level of service attributes, presented in packages (as illustrated in the case study in a later section), which are created using formal statistical design principles (see Hensher et al. [12], 2005). The data would be modelled by combining the SP data from the experiment and the revealed preference (RP) data on currently perceived levels of experience in respect of the attributes of interest from a survey, to obtain estimated parameters for each service quality attribute, which can then be combined to create a customer service quality index. These weights would not be subject to the interpretation of scale issues identified above and a CSQI thus designed would meet the requirements of being robust, capable of benchmarking customer satisfaction over time to inform policy and planning of services, and to inform how the separate elements of service quality contribute to service performance with a view to informing contract management. The CSQI created is critically dependent on the questions asked of users. In turn, this requires clarity in the potential use of the customer service quality index. For example, in the context of a bus contract, the interest herein, a CSQI which is to be used for monitoring consumer sentiment and for evaluating policy and planning through the impact of such policies and planning on the service quality attributes, can be less specific than a CSQI which is also to be used for identifying which elements of service quality contribute to performance for the purposes of operator contract development, benchmarking and compliance. In the latter case, the level of service service quality attributes need to be carefully constructed to create, so far as possible, attributes where the responsibility for changes to the level of service can be attributed to a single stakeholder (operator or government). Moreover, if the CSQI is to be used for operator contract development, benchmarking and compliance, a larger sample of bus users in particular will be required as it will be important to capture data effectively for each contract area. In developing a CSQI for use in benchmarking of operator performance, it is critical to recognise that operator’s may have little control over many of the attributes that define a consumer service quality index (or any customer satisfaction metric). Hensher, D.A. [13] has shown this in a study of private bus operators in Sydney (Australia) and concluded that many of the service quality dimensions that matter to users of public transport are often defined in a contract by the regulator (e.g., service coverage by time of day and day of the week and weekend, service frequency, while some attributes are heavily influenced by market forces (e.g., travel times at the peril of traffic congestion and weather). This paper focuses on the role that customer satisfaction expressed as a CSQI can play in the benchmarking of operator performance as part of monitoring bus contract Hensher International Journal of Quality Innovation 2015, 1:3 Page 3 of 17 http://www.jqualityinnovation.com/content/1/1/3 outcomes by the funder. Whilst this paper provides evidence for the bus industry, there is no reason to believe that the conclusions would differ for other public transport modes. In Hensher [13] we investigated the relationship between cost efficiency and CSQI elements under the control of the operator, and concluded that operators who provide services levels that result is a higher CSQI in general are also the more cost efficient, and that investing in higher levels of customer service does not have to necessarily require a great cost outlay. In the current paper we are focussing only on the development of a CSQI, providing details of how this can be obtained and used in a benchmarking context for regulators interested in both annual reviews of incumbent operators as well as for competitive tendering or negotiation (and re-negotiation) on contracts. The inclusion of CSQI in the performance review means that benchmarking becomes are much more valued process than simply one based on cost efficiency which is commonly mistaken as the same as cost reduction or no cost escalation (beyond inflation and other agreed variations). Developing a customer service quality index The concept of customer service quality includes aspects of transport service which are not always well-defined and easily measured. Herein we define service quality in terms of a set of attributes which each user perceives to be the sources of utility (or satisfaction) in bus use. The dimensions of quality, viewed from a bus user’s perspective, are complex. Passengers might, for example, consider the comfort at the bus stop and the time to get a seat, or only the comfort of the seats. Modal choice surveys have identified a large number of influences on the use of buses in contrast to other private and public modes. Service quality can be divided into six broad classes of effects, summarised in Table 1, each containing different quality dimensions (as identified by [14-18], and other studies). Recent contributions by Cirillo et al. [1], dell’Olio et al. [19], Eboli and Mazzulla a [3,20,21] and Marcucci and Gatta [22] have also reinforced the relevance of the attribute set identified in earlier research. Some of these contributions also use a stated preference method, acknowledging the original contribution by Hensher (estimated as multinomial logit and mixed logit models–see [6]), while other studies use a different, more traditional method, in which a satisfaction scale is multiplied by an importance scale (in various forms) to obtain an overall customer satisfaction index. The importance weights are used as proxies for the weights obtained from model estimation herein. Some demand side measures can be translated (or mapped) into a set of supply side equivalences (resources that the operator has partial or total control of) such as the timetable, fleet age, and/or the buses that are air conditioned; the number of vehicles that are wheelchair accessible, the number of cleaning hours of the vehicles, and the money spent on driver training. The attributes on the supply side are, in contrast to the quality attributes in column two in Table 1, to varying degrees, observable and under the direct control of the bus operator. For example a change in the average fleet size will, ceteris paribus, have a direct influence on the time to get a seat. On the other side we expect the supplied level of service quality to be a function of consumer preferences. If the supplied quality level is a response to customer preferences, and not only to some regulatory restrictions, Hensher International Journal of Quality Innovation 2015, 1:3 Page 4 of 17 http://www.jqualityinnovation.com/content/1/1/3 quality exogeneity cannot be assumed. In this circumstance we need to develop a capability to represent the quality of service as determined by users. The discrete choice approach is an appealing framework (see below). Given these considerations about service quality, we are able to introduce an improved version of the traditional cost model in its reduced form to capture the full dimensionality and service quality. The proposed and preferred service quality measure is constructed by analysing bus user preferences for different levels of bus service quality, and using the resulting weights attached to each underlying dimension of service quaility as perceived by users to derive the level of satisfaction associated with the supplied level of service quality. To this extent we need to identify and quantify the preferences for service levels from bus travellers. We restrict our analysis to actual bus users but recognise that non-users also provide useful information on the levels of service offered by bus operators. Within a performance regime based on the acceptability of service levels to actual users, and with a focus on the service quality that influences operator costs, the emphasis on users is appropriate. To reveal user preferences for service quality, we need to obtain data of sufficient richness to capture the behavioural responses to a wide range of levels of service quality defined on an extended set of attributes such as those given in Table 1. Revealed preference (RP) data is typically restrictive in its variance properties, but is an important input into the assessment. The preferred approach is a stated preference (SP) experiment combined with perceptions of existing levels of service. A sampled passenger would evaluate a number of alternative service levels (known as scenarios) together with the level experienced, and choose the most preferred alternative. Systematically varying the levels of the attributes in repeated scenarios enables us to obtain a profile of each passengers Table 1 Demand side effects and their equivalence on the supply side Dimensions of user perceptions Operator specified measures Getting to the bus stop quality ease, safe, time (distance), knowing where the bus stop is frequency, availability of bus shelter and seats Wait quality wait time at stop, punctuality of bus frequency wait comfort, wait safety availability of bus shelter and seats Trip quality time to board a bus frequency, % of low floor buses time to get a seat number of seats available moving to your seat average speed, network shape travel time travel time trip cost fare Vehicle quality cleanliness hours of vehicle cleaning/vehicle comfort of seats (types), spaciousness percent of buses with cloth seats temperature control (ventilation) percent of buses with air conditioning noise visual surveillance safety average age of the fleet modernity wheelchair access (yes/no) ease of use for those with disabilities Driver quality appearance-helpfulness years of driving experience money spent on driver training Information quality pre-trip information availability of timetable/destination signs Hensher International Journal of Quality Innovation 2015, 1:3 Page 5 of 17 http://www.jqualityinnovation.com/content/1/1/3 preferences for bus services. The data is analysed as a discrete choice model in which we combine the SP and RP data to obtain estimated parameters for each attribute. We estimate the simple multinomial logit model (MNL) in which all random components are independently and identically distributed (IID) (see Hensher et al. [12], 2005) b . Let U nsj denote the utility of alternative jperceived by respondent nin choice situation s. We assume that U nsj may be partitioned into two separate components, an observable component of utility,V nsj and a residual, unobservable component, ε nsj ,suchthat Unsj ¼Vnsj þεnsj:ð1Þ The observable component of utility is typically assumed to be a linear relationship of observed attribute levels, x, of each alternative jand their corresponding weights (parameters), β,with a positive scale factor, σ n such that Unsj ¼σnX K k¼1 βnkxnsjk þεnsj;ð2Þ where β nk represents the marginal utility or parameter weight associated with attribute kfor respondent n. The unobserved component, ε nsj ,is often assumed to be independently and identically (IID) extreme value type 1 (EV1) distribution. We will develop the implications of the distributional assumption in detail below. The individual scale factor in Equation (2) is normalised to one in most applications. (We refer to such models as constant variance models.) An alternative representation that preserves the preference order in Equation (2), as long as σ n does not vary across alternatives, is U nsj ¼X K k¼1 βnkxnsjk þεnsj=σn  :ð3Þ It can be seen that the variance of ε nsj , is inversely related to the magnitude of σnXK k¼1βnkxnsjk via σ n .Ifε nsj has an EV1 distribution with this scale parameter, then Var(ε nsj /σ n )=π 2 /6. In order to make any progress at modelling choices, it is necessary to make a number of assumptions about the unobserved components of utility. The most common assumption is that for each alternative, j,ε nsj ,willbe randomly distributed with some density, f(ε nsj ), over decision makers, n, and choice situations, s. Further assumptions about the specific density specification adopted for the unobserved effects, ε nsj (e.g., the unobserved effects are drawn from a multivariate normal distribution) lead to alternate econometric models. Assuming there exists some joint density such that ε ns =〈ε ns1 ,…,ε nsJ 〉represents a vector of the Junobserved effects for the full choice set, it becomes possible to make probabilistic statements about the choices made by the decision makers. Specifically, the probability that respondent nin choice situation swill select alternative jis given as the probability that outcome jwill have the maximum utility; Pnsj ¼Prob Unsj >Unsi;∀i≠j  ¼Prob Vnsj þεnsj >Vnsi þεnsi;∀i≠j  ð4Þ which can also be written as Pnsj ¼Prob εnsj−εnsi >Vnsi−Vnsj;∀i≠j  :ð5Þ Hensher International Journal of Quality Innovation 2015, 1:3 Page 6 of 17 http://www.jqualityinnovation.com/content/1/1/3 Equation (5) reflects the probability that the differences in the random terms, ε nsi −ε nsj will be less than the differences in the observed components of utility, V nsi −V nsj . The probabilities for a multinomial logit model given equation (5) can be computed in closed form. It has been shown in many sources, such as [12,23], Train, [24]), that for a multinomial logit (MNL) model: Prob Alt j is chosenðÞ¼ exp Vnsj  XJ j¼1exp Vnsj  ;j¼1;…;J:ð6Þ Assuming that the utility functions themselves are straightforward, the probabilities in (6) can be computed simply by plugging relevant quantities into the formula, with no approximations required. This is one of the appealing features of the multinomial logit form of a choice model which we use in this study. A CSQI for each bus operator (or contract region) can be derived from the application of the parameter estimates obtained form the estimation of the MNL model to the current RP levels which each operator-specific passenger sample currently experiences. This index is not a probability (of choice) weighted indicator that is typically derived from a choice model; rather we seek to establish an indicator based solely on the levels of service currently on offer. The SP-RP model’s role is to provide a rich set of parameter estimates to weight each attribute of service quality. To assist in the selection of attributes for the CSQI, we undertook an extensive review of the literature as well as a survey of bus operators who have a wealth of experience on what customers look for in a good service (see [25]). We found that thirteen attributes describe the major dimensions of service quality from a user’s perspective. The range of levels of each attribute in Table 2 provided us with a mechanism for establishing the weights that signal the contribution of each attribute to the overall SQI. Through a formal statistical design, the attribute levels are combined into bus packages before being translated into a survey form. The full factorial design (i.e., all possible bus packages) consists of 3 13 combinations of the 13 attributes each of three levels. To produce a practicable and understandable design for the respondents, we restricted the number of combinations to 81 (i.e., 81 choice sets) using a fractional design. Fractional designs permit the reduction in the number of combinations (i.e., the number of bus packages) without losing important statistical information (see [23]). A pre-test of the survey showed that respondents were able to evaluate consistently three choice sets (i.e., different scenarios of bus packages), resulting in 27 different survey forms. To allow for a rich variation in the combinations of attribute levels to be evaluated as service packages in the SP experiment, each bus operator received 8 sets of 27 different survey forms (i.e., 216 forms) and instructions on how to organise the survey. An example of an SP question is shown in Additional file 1: Table S3a, with the questions on a recent trip and background data shown in Additional file 1: Table S3b. Scheduled c bus users of 25 private bus operators in NSW participated. Survey forms were distributed and collected during the first half of 1999. A total of 3,849 useable observations (out of 4,334 returns) were incorporated in the estimation of the discrete choice model. A multinomial logit (MNL) specification was selected. This is appropriate for a model form in which the utility expressions associated with the current trip Hensher International Journal of Quality Innovation 2015, 1:3 Page 7 of 17 http://www.jqualityinnovation.com/content/1/1/3 and two attribute packages are unlabelled (or unranked) alternatives. Consequently all design attributes were generic across the three alternatives. In addition, in the current trip alternative we considered alternative-specific characteristics of the passenger (income, gender, age and car availability) and of the operator together with a number of other potential influences on relative utility such as treatment effect, trip purpose and access mode. The user preference model results The user attribute choice model is summarised in Table 3 9 with acronyms defined in the Appendix. The model includes the attributes of the SP experiment, operatorspecific dummy variables and three user characteristics. The nine sets of two dummy variables per service attribute are defined relative to a third level which is set to zero (given the three levels in the design). The overall goodness of fit (adjusted pseudo-R 2 ) of the model is 0.324. The great majority of the design attributes are statistically significant. Service reliability (i.e., the extent to which buses arrive on time), fares, access time and travel time are all highly significant with the expected negative sign. Relative to ‘reasonably unsafe’, we find a positive (almost) significant parameter estimate for ‘reasonably safe’(0.1510) and for ‘very safe’(0.1889). The higher estimate for ‘very safe’in contrast to ‘reasonably safe’is plausible. The infrastructure at the bus stop appears not Table 2 Set of attributes and attributes levels in the SP experiment Attribute Interpretation of levels Attribute Interpretation of levels Reliability on time Info at the bus stop on time 5 minutes late 5 minutes late 10 minutes late 10 minutes late Frequency every 15 minutes Travel time 25% quicker than the current travel time every 30 minutes same as now every 60 minutes 25% longer than the current travel time Walking distance to the bus stop -same as -now Bus stop facilities bus shelter with seats 5 minutes more seats only 10 minutes more no shelter or seats at all Waiting safety very safe Fare 25% more than the current one-way fare reasonably safe same as now reasonably unsafe 25% less than the current one-way fare Access to the bus wide entry with no steps Driver attitude very friendly wide entry with 2 steps friendly enough narrow entry with 4 steps very unfriendly Air conditioning available with no surcharge Safety on board the ride is very smooth with no sudden braking available with a surcharge of 20% on existing one-way fare the ride is generally smooth with rare sudden braking not available Cleanliness of seats very clean the ride is jerky; sudden braking occurs often clean enough not clean enough Hensher International Journal of Quality Innovation 2015, 1:3 Page 8 of 17 http://www.jqualityinnovation.com/content/1/1/3 defined attributes to the mix of attributes offered in a bus service, we use a discrete choice multinomial logit model to identify the role that each attribute in a package (‘an alternative’) plays in defining the level of utility (or satisfaction) applicable to each bus travelling member of the population. Importantly the stated preference method has as its sole purpose an enriched strategy to ensure a richer understanding of preferences for attributes describing service in a situation where the variability is actual service levels may not be rich enough to enable a fully revelation of customer preferences for each attribute. Then implementing an estimated model on actual experiences (revealed preference data) enables us to identify customer satisfaction overall from the package of experienced attribute levels, named the customer service quality index. This index when aggregated across a sample of users of an operator’s services enables us to obtain an operator specific CSQI, which can be benchmarked against other operators, distinguishing between those attributes that have levels under the control of the operator and those that are controlled by the regulator or the market at large. All attributes, no matter who has ‘responsibility for their level’, clearly matter to varying degrees to the end user and hence must all be taken on board in changes designed to create additional value in the bus use experience, and consequent justification through value for money of the taxpayers outlay to support existing and improved bus services under government contracts. In this paper we have quantified CSQI based on customer feedback from the performance of incumbent operators. This can be used to monitor the performance of incumbents under contract with the proviso that the contributing influences to the overall CSQI are distinguished in respect of who has control of and hence is responsible for each underlying source of relative satisfaction. However, this approach is also valuable in a competitive tendering or negotiated contract setting in that it can be used Table 5 Comparison of overall service performance between two years for a Singapore operator Attributes Coefficient Attributes Coefficient Constant 1.14* (0.22) Seated part of the way −0.11 (0.11) Fares −0.04* (0.0066) Very smooth journey 0.51* (0.11) Travel Times −0.20* (0.027) Slightly jerky journey 0.42* (0.11) Waiting Time −0.41* (0.076) Information (Reliable) 0.16 (0.13) Very Clean 0.54* (0.11) Information (Unreliable) −0.10 (0.12) Moderately Clean −0.19 (0.12) Facilities −0.11 (0.082) Temperature just nice 0.31* (0.11) Range of Waiting Time −0.035 (030) Seated all the way 0.65* (0.10) * denotes statistical significants at 5% level YEAR 1 YEAR 2 Components Data Components Data Fares 95 cents Fares 120 cents Travel Time 22 mins Travel Time 22 mins Waiting Time 8 mins Waiting Time 6 mins Seat Availability Standing all the way Seat Availability Seated all the way Smoothness of Ride Very smooth journey Smoothness of Ride Very smooth journey Cleanliness Moderately Clean Cleanliness Very Clean Temperature Just Nice Temperature Just Nice Hensher International Journal of Quality Innovation 2015, 1:3 Page 15 of 17 http://www.jqualityinnovation.com/content/1/1/3 to set targets or standards that are aligned with what is already achieved (or better) under current contracts that are being replaced. Endnotes a Eboli and Mazzulla [3] is a very useful review of the literature on methods used to study public transport service quality. b The MNL and more advanced methods are discussed in detail in Hensher et al. [12]. c School children were excluded from the sample, as they are captive users and might have a biased perception towards the attributes. Appendix Notation URELI Late minutes UTARIF Bus fare UACCESST Access time UTRATIM Travel time UVSAFE Very safe URSAFE Reasonably safe USEATS Seats only at bus stop USEATSHEL Seats plus shelter at stop UAVALFREE Free Air conditioning UAVALPAY Air conditioning at 20% extra fare UGSBRAKE Smooth ride UVSNBRAKE Ride very smooth UCENOUGH Clean enough UVCLEAN Very clean UWIDE2STP Wide entry and 2 steps UWIDENSTP Wide entry no steps UFRIENDN Friendly drivers UVFRIEND Drivers very friendly UTIMWMAP timetable and map UTIMNOMAP Timetable, no map UFREQ60 Frequency 60 minutes UFREQ30 Frequency 30 minutes Additional file Additional file 1: Table S3a. A Typical Stated Preference Exercise. Table S3b. Revealed Preference data collected. Competing interests The author declares that he has no competing interests. About the Author David Hensher is Professor of Management, and Founding Director of the Institute of Transport and Logistics Studies (ITLS) at The University of Sydney. A Fellow of the Australian Academy of Social Sciences, Recipient of the 2009 International Association of Travel Behaviour Research (IATBR) Lifetime Achievement Award in recognition for his long-standing and exceptional contribution to IATBR as well as to the wider travel behaviour community; Recipient of the 2006 Engineers Australia Transport Medal for lifelong contribution to transportation. David is also the recipient of the Smart 2013 Premier Award for Excellence in Supply Chain Management. Honorary Fellow of Singapore Land Transport Authority, and a Past President of the International Association of Travel Behaviour Research. He has published extensively (over 550 papers) in the leading international transport journals and key journals in economics as well as 12 Hensher International Journal of Quality Innovation 2015, 1:3 Page 16 of 17 http://www.jqualityinnovation.com/content/1/1/3 books. David has advised numerous government and industry agencies in many countries (notably Australia, New Zealand, UK, USA and The Netherlands), with a recent appointment to Infrastructure Australia’s reference panel on public transport, and is called upon regularly by the media for commentary. Acknowledgment The comments from Sang Lee are appreciated. Received: 13 September 2014 Accepted: 15 December 2014 Published: 21 May 2015 References 1. Cirillo C, Eboli L, Mazzulla G (2011) On the asymmetric user perception of transit service quality. Int J Sustainable Transportation 5(4):216–232 2. dell’Olio L, Ibeas A, Cecìn P (2010) Modelling user perception of bus transit quality. Transport Policy 17:388–397 3. Eboli L, Mazzulla G (2010) How to capture the passengers’point of view on a transit service through rating and choice options. Transport Rev 30(4):435–450 4. Bolton RN, Drew JH (1991) A multistage model of customers’assessment of service quality and value. J Consum Res 17:375–84 5. Boulding W, Kalra A, Staelin R, Zeithaml VA (1993) A dynamic process model of service quality: from expectations to behavioral intentions. J Mark Res 30:7–27 6. Hensher DA, Prioni P (2002) A Service Quality Index for Area-wide Contract Performance Assessment. J Transport Econ Pol 36:93–113 7. Hensher DA, Stopher PR, Bullock P (2003) Benchmarking and service quality at a market segment level. Transport Res Part A 37:499–517 8. Buttle F (1996) SERVQUAL: review, critique, research agenda. Eur J Mark 30(1):8–31 9. Parasuraman A, Zeithaml V, Berry LL (1985) A conceptual model of service quality and its implications for future research. J Mark 49:41–50 10. Andersson TD (1992) Another model of service quality: a model of causes and effects of service quality tested on a case within the restaurant industry. In: Kunst P, Lemmink J (eds) Quality Management in Service. van Gorcum, The Netherlands, pp 41–58 11. Richard MD, Allaway AW (1993) Service quality attributes and choice behavior. J Serv Mark 7(1):59–68 12. Hensher DA, Rose JM, Greene WH (2005) Applied Choice Analysis: A Primer, Cambridge University Press. Second edition published in 2015, Cambridge 13. Hensher DA (2014) The relationship between bus contract costs, user perceived service quality and performance assessment, (presented at Thredbo 12, Durban, South Africa, September 2011). Int J Sustainable Transportation Spec Issue 8(1):5–27 14. Hensher DA (1992) Performance evaluation and passenger transit: what are the relevant measures? The Proceedings of the Second International Conference on Privatisation and Deregulation of Passenger Transport, Finland, pp 61–72 15. Fielding GJ, Babitsky TJ, Brenner ME (1985) Performance evaluation for bus transit, Transportation Research, 19 (1), 73–82. 16. Kittleson and Associates (1996) Development of Transit Capacity and Quality of Service Principles, Practice and Procedures, Transit Cooperative Research Program Project A-15. Transportation Research Board, Washington D.C 17. Swanson J, Ampt L, Jones P (1997) Measuring bus passenger preferences, Traffic Engineering and Control, June, 330–36. 18. Cunningham LF, Young C, Lee M (1997) Developing customer-based measures of overall transportation service quality in Colorado: Quantitative and qualitative approaches. J Public Transportation 1(4):1–22 19. dell’Olio L, Ibeas A, Cecìn P (2010a) The quality of service desired by public transport users, Transport Policy, doi:10.1016/ j.tranpol.2010.08.005. 20. Eboli L, Mazzulla G (2008) An SP experiment for measuring service quality in public transport. Transportation Plann Technol 31(5):509–523 21. Eboli L, Mazzulla G (2008) Willingness-to-pay of public transport users for improvement in service quality. Eur Transport 38:107–118 22. Marcucci E, Gatta V (2007) Quality and public transport service contracts. Eur Transport 36:92–106 23. Louviere JJ, Hensher DA, Swait J (2000) Stated Choice Methods: Analysis and Applications in Marketing, Transportation and Environmental Valuation. Cambridge University Press, Cambridge 24. Train K (2009) Discrete Choice Methods with Simulation. Cambridge University Press, Cambridge 25. Prioni P, Hensher DA (2000) Measuring service quality in scheduled bus services. J Public Transport 3(2):51–74 doi:10.1186/s40887-015-0003-9 Cite this article as: Hensher: Customer service quality and benchmarking in public transport contracts. International Journal of Quality Innovation 2015 1:3. Hensher International Journal of Quality Innovation 2015, 1:3 Page 17 of 17 http://www.jqualityinnovation.com/content/1/1/3