Activity-based costing in restaurants

1. Introduction
Interest in cost and management accounting practices in the restaurant industry is rising (Raab et al., 2009; Annaraud et al., 2008). Pavesic (1985) has initiated research in pricing and cost accounting for restaurants, introducing the concept of profit factor (PF) in menu engineering (ME). Prior studies, such as the one presented in Chan and Au (1998) investigate the implications of not incorporating overhead costs in menu-item profitability analyses in restaurants in Hong Kong. Since then, a number of researchers have examined the application of contemporary cost and management accounting techniques, and particularly activity-based costing (ABC) in a restaurant environment (Raab et al., 2004).

ABC is a cost accounting methodology that aims to allocate overhead costs effectively. ABC traces costs by using resource and activity cost drivers that reveal activities and objects consumption patterns on the basis of a cause and effect relationship. In an ABC, model cost drivers are used to establish a transitional mapping between resources, activities and cost objects. The identification and selection of appropriate and accurate cost drivers is one of the most difficult tasks in ABC models (Cobb et al., 1992). As this survey showed, most companies find it difficult to translate theory into action, mainly due to difficulties relating to data collection, identification of activities and lack of resources. The most common technique used for the identification of activities and the selection of cost drivers is by interviewing the heads of departments.

Other methods include labour reporting systems, work order systems, employee surveys, observations, timekeeping systems and storyboards (Kaplan and Cooper, 1998). Velcu (2002) argued that some data, like cost driver information can be more easily obtained from a company’s information system, such as ERP systems. Although acompany may have implemented a variety of management information systems, obtaining sound information resulting from employee’s interviews, still remains a problem (Kostakis et al., 2011). This study aims at overcoming the problem of estimating cost drivers by introducing a methodology in ABC models in restaurants, a problem that commonly appears in practice. In particular, the study shows how by integrating three management techniques, one can accurately compute values for cost drivers, in cases where real data are not available or are difficult and time consuming to gather. These techniques include ABC, simulation and association rule mining (ARM) and are applied in the restaurant industry. The proposed method equally serves as a tool for improving cost estimation in menu-item profitability analysis.

2. Purpose and background
ABC in restaurants
ABC is one of the most important innovations in the field of cost and management accounting (Bjornenak and Mitchell, 1999; Bjornenak, 1997). ABC systems use a two-stage approach that is similar to the structure of traditional cost systems, such as job order costing and process costing. Traditional costing systems use actual departments or cost centres for accumulating and redistributing costs, while ABC systems use activities for this purpose. Thus, instead of asking how to allocate a service department expense to a production department, the ABC system designer requests a list of performed activities by the service department’s resources. The resource expenses are assigned on the one hand to the number of required activities and on the other hand on the former’s usage to perform these activities (Atkinson et al., 2001; Garrison and Noreen, 2000). The usefulness of ABC in the financial performance of firms has been explored in the literature. Cagwin and Bouwman (2002) use confirmatory factor analysis and structural equation modelling to investigate whether the use of ABC is associated with improved financial performance. Results show that:

[. . .] there is a positive association between ABC and improvement in ROI when ABC is used concurrently with other strategic initiatives, when implemented in complex and diverse firms, when used in environments where costs are relatively important, and when there are limited numbers of intra-company transactions (Cagwin and Bouwman, 2002). To this end Kaplan and Cooper (1998) suggest that service companies are ideal candidates for ABC. One reason for this is that in the services industry a great percentage of total cost is labour cost and processes are highly variant (Kostakis et al., 2008). Studies have shown the possible use and potential benefits of ABC in the hotel industry (Kostakis et al., 2011; Berts and Kock, 1995; Noone and Griffin, 1999). New techniques assisting large-scale ABC implementation have also been developed in recent years (Kaplan and Anderson, 2004).

The restaurant industry is facing great challenges today, mainly due to the highly competitive environment and the diverse needs of the customers. Restaurants have to achieve a balance between serving these needs, and pricing their menu items to achieve adequate profitability levels (Raab et al., 2009).Traditional costing systemsand simplistic pricing approaches were very popular within the restaurant industry. Overhead cost and operating expenses were usually excluded and prices were a function of food cost, as well as customer demand, historical prices and other factors. This method has the evident disadvantage of leading to misleading information about individual menu items. Research has revealed the significance of labour and energy costs in restaurants (Chan and Au, 1998). This study also revealed that restaurants are labour intensive with labour cost representing a great share of total operating cost.

Although restaurant managers usually are able to monitor their customer’s needs, a study conducted by Raab and Mayer (2003) shows that most of them do not know the true profitability of their various menu items. This study also reveals that although some restaurant managers trace a portion of undistributed operating costs to their menus, the restaurant industry in general does not allocate overhead costs to menu items. Studies in cost and management accounting in the restaurant industry demonstrate the potential benefits of the use of ABC in their operations. Raab et al. (2005) applied an adapted ABC model developed by Cooper (1989) in a buffet-style restaurant in Hong Kong. ABC was able to gain insights about the operating profit margins of individual menu items. ABC has also been applied to a major quick service restaurant chain (Annaraud et al., 2008). A study conducted by Raab and Mayer (2007) combines ME with ABC in restaurants, using direct observation to calculate the percentage of time an employee spends on an activity, while Raab et al. (2009) incorporate PF analysis to compare traditional ME approach (contribution margin-CM based) with ME and ABC (ABC/ME) approach.

This model also used observation as means of allocating labour cost to individual menu items. Recent studies provide empirical evidence of the current general trends regarding the practical consideration, adoption, and use of ABC in the hospitality industry (Pavlatos and Paggios, 2009). The presented methodology aims at overcoming the drawbacks of using observation and similar methods of estimating cost drivers to allocate operating cost to menu items, thus attempting more accurate cost estimation in the restaurant industry. Given that some cost drivers are difficult to calculate, even when observation techniques are employed, this study demonstrates a new method that allows the user to calculate all cost drivers that are included in an ABC model.

Methods for estimating cost drivers ABC is a methodology used to allocate overhead costs effectively. Dunn and Brooks (1990) and Noon and Griffin (1999) discuss the steps involved in the design of ABC systems in hotels: identifying activities, assigning costs-to-cost pools, selecting appropriate cost drivers for assessing the cost of activities to cost objects and assigning the cost of the activities to services and to customers. Thus, ABC is a two-stage approach. In the first phase, resource costs are mapped to activities through resource cost drivers, taking into consideration the resources required by each performed activity. In the second phase, activity costs are mapped to the cost objects that use these activities, by means of activity cost drivers. During the two stages of the aforementioned methodology, relevant data might not be available.

One such example is the time spent by employees in performing various activities. These data include the level of every activity’s usage within the production sector of the final products, and/or services and needs to be estimated (Wieserma, 1995). The restaurant industry numbers hundreds of activities, such as setup, cleaning, cooking, taking orders, etc. Defining the appropriate cost drivers and calculating them accurately is essential, in order to achieve high accuracy of total cost. To deal with the problem of lack of data, three levels of data accuracy can be used to estimate the proportions and levels stated above: educated guess, systematic appraisal and collection of real data. When real data are not available and cannot be obtained or when they are very expensive to obtain, the method of educated guess is used.

This method uses experts’ opinions to gather data, which are then put together to form estimates. Clearly, the method of educated guess is time-effective. However, as Anderson and Kaplan (2007) point out a subtle and serious problem arises from the interview and survey process itself. They suggested that when people estimate how much time they spent on a list of activities, they report percentages that add up to 100 percent and do not record much idle or unused time. Therefore, almost all ABC systems calculate cost driver rates assuming that resources work at full capacity, making cost estimates less accurate. Second, the systematic appraisal method uses quantitative approaches to obtain proportions, such as analytic hierarchical processing. This method offers a higher degree of accuracy. However, personnel with analytical skills and specialised computer software are needed; moreover, the method requires much computational time. Finally, the actual data collection (ADC) method requires systematic data sampling, followed by analysis of data using statistical methods. The main drawback of this method is the high cost associated with its implementation.

This study aims at overcoming the drawbacks of the three methods of estimating cost drivers, by introducing a new model based on the integration of discrete-event simulation and ARM in an ABC context in restaurants. The method can be proved particularly useful, as it models process variability through the simulation of cost drivers. Moreover, it uses data mining to extract associations between cost drivers that are easy to estimate and cost drivers that are either difficult or time consuming to estimate. The association rules will lead to the estimation of the values of the cost drivers that were difficult to calculate before. The proposed methodology complements the three methods of estimating cost drivers, namely educated guess, systematic appraisal and ADC, rather that substituting them.

Simulation modelling in ABC
In business, manufacturing and services, such as the restaurant industry, processes involve a great number of interconnections. This means that cost drivers usually interact with each other; this interaction affects the total cost and should be accounted for when building an ABC model (Kostakis et al., 2008). Traditional ABC models produce only point estimates of product cost; cost sensitivity analysis is then performed to test various cost scenarios. This method has the disadvantage of not incorporating the interaction of cost drivers in the production process. Thus, in practice, the application of traditional ABC techniques may be problematic, particularly in the services industry. To overcome this problem, Beck and Nowak (2000) introduced a model forABC based on discrete-event simulation in manufacturing.

They used simulation modelling to supply the ABC model with values for the cost drivers, thus producing confidence interval estimations for the cost at varying system conditions. Their model uses a range of values for activity drivers to model process variation. For example, for estimating the set up cost of a machine, they used a duration activity driver, such as “set up time of the machine”, instead of “number of setups”. This shows that some setups take less time than others, depending on the final product; selecting the “number of setups” as an activity driver would distort the final cost.

The model proposed by Beck and Nowak (2000) is built on the concept of entities flowing through the system. In a discrete-event simulation, physical items flow through the sequence of the manufacturing processes, while in ABC models cost flows through the model driven by cost drivers. The simulation model uses statistical distributions of cost drivers developed to supply the ABC system, thus modelling the stochastic nature of the system’s activities. These distributions are calculated based on performance data. In practice, however, it is rather difficult to gather performance data and develop statistical distributions accordingly for all cost drivers that will be used in the simulation model. In order to overcome this bottleneck, the presented model builds on the simulation model proposed by Beck and Nowak (2000) and introduces the concept of ARM. It is based on the extracted association rules between cost drivers that are easy to estimate and cost drivers that are difficult to estimate.

3. Simulation/arm model in ABC
The proposed methodology integrates three techniques: discrete-event simulation, ARM and ABC in a restaurant context. ARM is one of the core data mining techniques and has been initially developed to be used in market basket analysis. The purpose of ARM is to discover the important relations in data, using knowledge discovery methods, while the algorithms for ARM extract dependencies between various factors in databases (Agrawal et al., 1993). In market basket analysis, the purchase of a list of items is regarded as a single transaction. The objective is to discover trends and relations among a series of transactions, which in turn could be analyzed to reveal patterns that will be used as information (Witten and Frank, 2001). An association rule might be of the form: 70 per cent of the customers who purchased bread, also purchased milk and 60 per cent of the transactions contain bread and milk.

An association rule is an implication of the formX $ Y, whereXandYare sets of values. The rule X $ Y holds in the transaction set with confidence c, if c% of transactions that contain X also contain Y. The rule X $ Y has support s in the transaction set, if s% of transactions contain X and Y. The purpose of mining association rules is to generate all association rules with support and confidence above a threshold value. The introduction of ARM in ABC in restaurants solves the problem of defining all cost drivers, a condition that is a prerequisite in the methodology by Beck and Nowak (2000). The methodology operates on the scenario that there are some cost drivers in the ABC model that are difficult or extremely time-consuming to estimate. If an association is found between those cost drivers with another cost driver, whose distribution is known, then the former can lead us to the estimation of the latter. The proposed methodology was adapted fromthe model presented by Kostakis et al. (2008) (Figure 1). Figure 1 shows the intersection of simulation modelling, data mining and ABC. ARM is used to supply discrete-event simulation with values to be used in ABC modelling. An example of how ARM can assist simulation modelling in ABC in restaurants will be presented.

The methodology was applied to one restaurant specializing in Apple pees restaurant in Kuwait. The items that were selected to be included in the ABC model were the most popular ones, in terms of sales: Greek salad, tzatziki, baked chicken with potatoes, fried meatballs with rice, home-made French fries and a pudding. Data were collected in site for one day (one shift) on a Sunday.

Three activities were considered in the ABC model: “preparation and cooking time”, “dish assembly time (dat)”, which form the activity centre back-of-the-house (BOH) before the food is served and the activity “washing time”, that includes the time the plates are cleaned and put into the washing machine to be washed and forms another BOH activity. In the model presented in Raab et al. (2009), the activity “washing” is grouped under the main activity “preparation”. This means thatwhen calculating the resources used (in time units) for these activities, accuracy is low, since the main activities involve numerous other activities, whose values in practice vary greatly. In our model, the time spent on these activities were recorded in seconds. The BOH activities and their corresponding activity drivers for the six menu items under consideration are shown in Figure 2. Figure 2 shows only the part of the ABC model under consideration. A whole ABC restaurant model would include the cost pools (such as “personnel” and “direct operating supplies”) and their corresponding resource drivers. Figure 2 shows that the activities materialize in accordance with the six different menu items, while Table I summarizes the activities and their activity drivers.

In the following analysis, we present only data for the dish “baked chicken with vegetables and potatoes”. Similar analysis has been done for all six menu items.

This particular dish included: cooked chicken, baked potatoes, boiled peas and fresh vegetables, like tomato and lettuce. The chicken, the potatoes and the peas were prepared and cooked before the restaurant opened andwere stored and waited to be served. In total, 20 dishes were prepared for this menu item, while sales amounted to 15 dishes. The total time to prepare and cook the aforementioned items was 25 minutes for the 20 dishes. The total preparation and cooking time is calculated following the equation: pct (preparation and cooking time) = k/n + Lx

where k is the time the 20 plates are prepared and cooked before the restaurant opens, n ¼ 20 and x is the time the dish iswarmed in themicrowave and the fresh vegetables are cut before they are assembled; L is a constant.

When an order was placed for this particular menu item, the dish had to be assembled. This activity was measured using the activity driver “dat”, and it was measured in seconds. Therefore, the total time for preparation, cooking and assembling before the dish was served is:

tt = pct + dat
Table II shows the time recorded (in seconds) for each activity driver for the menu item “baked chicken with vegetables and potatoes” for one shift. It should be noted that

the first part of equation (1) (k/n) corresponds to dishes produced, while the second part of equation (1), as well equation (2) corresponds to dishes sold. Cost includes only labour cost, while direct raw material cost, as well as overhead cost is excluded from the analysis. Daily salary is calculated on a e36/day basis. Therefore, the final column of Table II is calculated using the following equation:

Cost
item
¼ tt *
36
8
; where 8 corresponds to an 8 2 hour shift ً3 The activity driver “washing time” comprised of two terms; one term that could be measured and included the cleaning and washing time of the various cooking utensils (a/n, where a is the total time the cooking utensils were cleaned and washed and n is the number of dishes produced); and another term that could not be easily measured and included the time the plates were cleaned and put in the washing machine. In order to be able to estimate the aforementioned time, a hypothetical relation needed to be developed between one cost driver that could easily be measured (such as pct) and the “washing time” (wt) that could not be measured. In the final stage, this hypothetical relation will have to be verified. Table III shows this relation that materializes in accordance with the six different menu items. It is expected that a sample of real measurements of “preparation and cooking time” and “washing time” will contain this relation, depending on the sample size. For the menu item “baked chicken with vegetables and potatoes”, a3/n3 is 200 seconds/20 dishes produced, that is 10 seconds (time was measured in site).

In order to be able to model process variability and produce confidence interval estimates for the cost, a simulation model was built. Figure 3 shows the activity-cycle diagram (ACD) for this particular simulation model. The model incorporates the BOH activities, as described previously, before and after the food is served. The serving, as well as the whole front-of-the-house (FOH) operations are not included in the model. One kitchen assistant is working for the preparation and cooking, as well as the dish assembly activities and one kitchen assistant is doing the washing and cleaning. The statistical distributions that were developed were based on the performance data that were gathered and are presented in Table II.

The statistical distributions are assumed to be normal for the activities “preparation and cooking” and “dish assembly” (for “preparation and cooking” N(246,33.5) and for “dish assembly” N(184.93, 30.01). The activity “washing time” materializes in accordance with the equations in Table III. The simulation model was built using Extend software. The model was run for seven consecutive shifts (one week). The model allowed for a warm-up period of two hours and results were recorded after the termination of this period. For each simulation run, data regarding the activity driver “washing time” were recorded, based on the equations in Table III and for each value of the other cost drivers that the model produced. The next step was to categorize these continuous values in ranges; this categorizationwas based on the algorithmfor discretization reported inBerka and Bruha (1998). Following, these ranges were used as the training set of the association rule algorithm Apriori (Agrawal and Srikant, 1994). Figure 4 shows some of the extracted association rules for the menu item “baked chicken with vegetables and potatoes”.

It was observed that these association rules were depicting the aforementioned dependencies presented in Table III. Each association rule presents the ranges that were categorized, according to the discretization mentioned before. For example, if we consider the association rule: “if ‘preparation and cooking time’ is between (225, 243), then ‘washing time’ is between (13.5, 15.2)” for menu item “baked chicken with vegetables and potatoes”, this association confirms the relation in Table III. Based on the extracted association rules, it is possible to define the empirical distribution of “washing time”. A new table, similar to Table II can be drawn, which contains the final results of the simulation model, after ARM is performed. Sales were recorded for all seven consecutive days. All measurements regarding the distributions of the activity drivers were recorder and clustered. The mean value of every cluster was used in the final ABC model.

Finally, the ABC model, which was implemented at Microsoft Excel, utilized these values of the activity drivers, to build confidence interval estimates for the cost per menu item. Table IV shows the total weekly cost (only direct labour cost is included in the model) for all menu items for all items sold (total weekly cost is the sum of all confidence interval estimates of the cost per menu item).
Table IV presents the final weekly direct labour cost for all menu items. The presented methodology is an innovative technique allowing for accurate cost estimates that are based on: performance data for those cost drivers that are easy to estimate; . statistical distributions that could not be developed easily, but whose values can be estimated using verified hypothetical relations; and confidence interval estimates rather than point estimates of the cost with the aid of discrete-event simulation.

4. Discussion
In this paper, we present an integrated model for effectively calculating values for cost drivers in an ABC model in restaurants. The model utilizes three techniques: discrete-event simulation, ABC and ARM, which is one of the core data mining techniques. The purpose of this model is to allow for accurate estimations of cost drivers, whose estimation is difficult to make with the common methods of defining cost drivers. The benefits of using ABC models have been extensively reported. ABC systems are able to trace overhead costs to individual products (Garrison and Noreen, 1997). Traditional costing seems inappropriate, in cases where processes are highly variant and product/service lines are diverse. These are typical conditions of service companies, where overhead costs today represent approximately two-thirds of the total cost (Raab et al., 2009).

In the restaurant industry, which is characterized by high degree of competitiveness, small profit margins and relatively high failure rate in the USA (Bell, 2002), the use of sophisticated cost systems is considered essential. For example, a study conducted by Raab and Mayer (2003) to a sample of 100 managers in US restaurants found that restaurant managers are increasingly aware of the need to trace some of their overhead costs, such as salaries and wages, to individual menu items. ABCmodels have been developed and applied in the restaurant industry before (Raab and Mayer, 2007). These studies reveal the importance of applying ABC in restaurants and conclude that the distribution of expenses between departments changed drastically when ABC methods were applied, as opposed to more traditional pricing techniques (such as ME). However, the limited application of ABC in restaurants is attributed to the difficulties of tracing costs to activities and activities in customers and products in restaurants (Kunst and Lemmink, 1995).

The study conducted by Raab and Mayer (2003) revealed that, although 50 per cent of the 100 restaurant managers attempted to measure processes and their costs, only one restaurant company was able to gain knowledge of their labour costs by calculating activity-based labour costs. The majority of studies that apply ABC in restaurants use observation techniques and interviews in order to identify and calculate values for cost drivers (Raab et al., 2009). However, as Anderson and Kaplan (2007) point out, a subtle and serious problem arises from the interview and survey process itself. They suggest that when people estimate how much time they spend on an activity, they do not record much idle or unused time. Therefore, almost all ABC systems calculate cost driver rates assuming that resources work at full capacity; this might lead to less accurate cost estimates.

Finally, some activities, such as the time an employee has spent performing a particular task, are difficult to calculate in all cases, as these can change frequently (Munoz and Oksan, 2006). This is especially true in the services industry. This study has extended the current knowledge in the area of cost accounting and cost driver estimation in the restaurant industry in three ways. First, it showed how to deal with diversity and heterogeneity in processes in the restaurant industry with the aid of discrete-event simulation, hence revealing the interaction between the cost drivers. The simulation model that was built to assist the ABC system was based on the methodology presented by Beck and Nowak (2000). The model produces values for the cost drivers, representing variations in the processes (i.e. time spent on an activity).

These values are in turn used by the ABC system to produce confidence interval estimations for the cost. Nevertheless, the use of simulation requires the model to be kept to low-complexity levels, as this ensures the minimum variation in the results. Second, it enabled the activity centres to be divided into more detailed activities, without raising concerns about compounding of errors; the simulation model produces a range of values; average values for the cost are therefore considered. Beck and Nowak (2000) report similar results when simulation average cost and ABC point estimates without simulation are used. The division of main activities into more detailed activities enables managers to gain insight about the utilization of resources by the activities. In cases where activities are grouped under main activities, this division is not possible, hence useful and precise information is lost.

Third, it introduced the method of ARM, which is one of the core data mining techniques, to assist simulation modelling in the calculation of the values of cost drivers. In the model produced by Beck and Nowak (2000), it is a prerequisite to have empirical distributions of all simulated cost drivers. The presented model overcomes this constraint and utilizes ARM to facilitate this process. The model is based on the proposition that associations between cost drivers, which are easy to estimate and cost drivers that are difficult to calculate, can supply the model with values for the latter. The presented model may equally be empowered by any other cost driver calculation method, such as systematic appraisal and collection of real data. However, these may be limited to a number of cost drivers.

One of the most important features of the methodology is the use of low-complexity algorithms. For the same purpose, various methodologies could be used. However, techniques such as regression analysis are too complex and are based on some working hypothesis developed by its users. On the contrary, ARM is based on a user-friendly platform and is free of assumptions. It also extracts all dependencies in one run (Kostakis et al., 2008).

Cooper and Kaplan (1988) introduced the concept of the optimal cost system; this system aims at minimising the sum of the cost measurement, i.e. those costs associated with themeasurements required by the cost system.AnoptimalABCmodel should aimat balancing the cost of errorsmade frominaccurate estimates with the cost of measurement. The presented methodology accomplishes this goal, as not only does it minimize the cost of measurement, but also reduces the cost of errors, by using modelling techniques, which can be tested for their sensitivity and verified with the physical system. Al-Omiri and Drury (2007) suggest that ABC cost systems can be either less or more sophisticated. The level of sophistication depends on the number of cost drivers and cost pools that these systems use.

The use of more sophisticated cost systems increases the accuracy of cost information. Therefore, if the management of a restaurant is interested in using a more detailed and refined cost system, it might have to use some cost drivers that are difficult, expensive or time-consuming to estimate. Although this study has contributed to cost and management accounting in the restaurant industry, some limitations should be noted. This study has focused only on three activities and hence, the application has been limited to the BOH of the restaurant. The FOH activities have not been taken into account, somehow affecting the results of the simulation model. This, however, does not prevent the user to expand the model by incorporating more activities, providing that there exists a true relationship between the activities to be modelled. This will enable numerical dependencies to be extracted. Finally, the model does not incorporate customer demand for the different menu items. This parameter should be included in future work to allow for better recommendations regarding the pricing of various menu items.

5. Conclusion
This study has presented a new technique in the area of cost and management accounting in the restaurant industry. It utilizes three techniques to model ABC in restaurants. Discrete-event simulation is used to generate values, which are in turn used to produce confidence interval estimates for the cost in an ABC model. The use of simulation reveals the dynamic behaviour of cost-parameters in a production process. ARM assists the ABC model, by providing values for those cost drivers that are difficult to calculate.

The method results in considerable time saving, since it reduces the interview and survey practices for cost driver estimation. It also reduces the probability of making inaccurate estimations of cost drivers. This results into a more accurate and efficient cost accounting information in the restaurant industry.

References
(Raab et al., 2009; Annaraud et al., 2008).
(Raab et al., 2004).
(Cobb et al., 1992).
(Kostakis et al., 2011).
(Bjornenak and Mitchell, 1999; Bjornenak, 1997).
(Atkinson et al., 2001; Garrison and Noreen, 2000).
(Cagwin and Bouwman, 2002).
(Pavlatos and Paggios, 2009).
(Witten and Frank, 2001).
(Kunst and Lemmink, 1995).
Agrawal, R. and Srikant, R. (1994), “Fast algorithms for mining association rules”, Proc. of the 20th VLDB Conference, Santiago, Chile, pp. 487-99.
Agrawal, R., Imielinski, T. and Swami, A. (1993), “Mining association rules between sets of items in large database”, Proceedings of the ACM SIGMOD International Conference on the Management of Data, Washington, DC, USA, pp. 207-16.

Al-Omiri, M. and Drury, C. (2007), “A survey of factors influencing the choice of product costing systems in UK organizations”, Management Accounting Research, Vol. 18 No. 4, pp. 399-424.
Anderson, S.R. and Kaplan, R.S. (2007), Time-driven Activity-based Costing: A Simpler and More Powerful Path to Higher Profits, Harvard Business School Press, Boston, MA. Annaraud, K., Raab, C. and Schrock, J. (2008), “The application of activity-based costing in a quick service restaurant”, Journal of Foodservice Business Research, Vol. 11, pp. 23-44. Atkinson, A., Banker, R., Kaplan, R. and Young, M. (2001), Management Accounting, Prentice-Hall, Upper Saddle River, NJ.

Beck, V.U. and Nowak, J.W. (2000), “The merger of discrete event simulation with activity based costing for cost estimation in manufacturing environments”, Proceedings of the 2000 Winter Simulation Conference, Orlando, FL, USA, pp. 2048-54. Bell, D. (2002), Food and Beverage Cost Control-course Packet, Department of Reprographic Services, University of Nevada, Las Vegas, NV.