Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A proposal for improving the meal provisioning process at Canadian Airlines Morency, Vincent 2000

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2000-0121.pdf [ 3.68MB ]
Metadata
JSON: 831-1.0099498.json
JSON-LD: 831-1.0099498-ld.json
RDF/XML (Pretty): 831-1.0099498-rdf.xml
RDF/JSON: 831-1.0099498-rdf.json
Turtle: 831-1.0099498-turtle.txt
N-Triples: 831-1.0099498-rdf-ntriples.txt
Original Record: 831-1.0099498-source.json
Full Text
831-1.0099498-fulltext.txt
Citation
831-1.0099498.ris

Full Text

A PROPOSAL FOR IMPROVING THE MEAL PROVISIONING PROCESS AT CANADIAN AIRLINES  by Vincent Morency B.Eng. Electrical Engineering, McGill University, 1998  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE (BUSINESS ADMINISTRATION) in THE FACULTY OF GRADUATE STUDIES (Department of Commerce and Business Administration)  We accept this thesis as conforming  THE UNIVERSITY OF BRITISH COLUMBIA December 1999 © Vincent Morency, 1999  r  In p r e s e n t i n g t h i s t h e s i s i n p a r t i a l f u l f i l m e n t of the r e q u i r e m e n t s f o r an advanced degree at the U n i v e r s i t y of B r i t i s h Columbia, I agree t h a t the L i b r a r y s h a l l make i t f r e e l y a v a i l a b l e f o r r e f e r e n c e and s t u d y . I f u r t h e r agree t h a t p e r m i s s i o n f o r e x t e n s i v e c o p y i n g of t h i s t h e s i s f o r s c h o l a r l y purposes may be g r a n t e d by the head of my department or by h i s or her r e p r e s e n t a t i v e s . I t i s u n d e r s t o o d t h a t c o p y i n g or p u b l i c a t i o n of t h i s t h e s i s f o r f i n a n c i a l g a i n s h a l l not be a l l o w e d w i t h o u t my w r i t t e n p e r m i s s i o n .  (Signature)  Department of The U n i v e r s i t y of B r i t i s h Columbia Vancouver, Canada Date  -hpn^lttA  H ,  Abstract Catering flights is an important part of an airline's operations. The meal service has a critical impact on customer service quality and represents significant costs. Unfortunately, due to high passenger load variability and minimum production lead-time requirements, it is difficult to get the number of meals to exactly match the passenger count on each flight. The objective of the project is to reduce meal-catering costs due to over-catering while simultaneously improving service level by reducing under-catering. The proposed solution takes the form of a meal bank system. This concept suggests that flights should be systematically under-catered with the possibility of uploading a generic meal to fill in any shortage at the last minute when the final passenger load is known. A thorough investigation of current processes was carried in order to define and recommend process change based on the meal bank approach. The project concludes with a costing analysis of recommended solutions weighing additional investments, cost savings, and service improvement against current performance. Costing analysis results for Vancouver International Airport operations showed that service level could be improved by an approximate reduction of 50 percent in revenue passenger meal shortages. Also, net costs can be reduced up to an estimated $190,000 annually.  11  Table of Contents Abstract Table of Contents List of Tables List of Figures Acknowledgments I - Introduction II - Background A . The Catering Operations  III - Methodology and Approach A. B. C. D.  Preliminary Meal Bank System Analysis Investigation of the Meal Catering System Scope of the Project Model Development & Consolidation of Results  A. B. C. D. E.  Meal Bank Service Options Meal Bank Storage and Lifetime Aircraft Operations Ordering Policy Analysis Scenario Analysis  IV - Analysis  V - Discussion and Recommendations  ii iii iv v vi 1 2 3  11 11 13 14 15  16 16 17 19 21 23  28  A . Meal Ordering B. Meal Production Control and Inventory Management C. Equipment Balancing D. Transportation Logistics E. Aircraft Operations F. Scenario Selection G. Markov Decision Process Model H. Confirmation Study I. Implementation  28 28 29 29 29 30 30 31 32  VI - Conclusion Bibliography Appendix A: LAP Meal Diversity and Quality Appendix B: Preliminary Analysis Appendix C: Costing Analysis Database  33 34 35 36 38  A . Data Source Issues B. Database Structure C. Source Data Tables  Appendix D: Meal Bank Wastage Cost Appendix E: Delivery Van Driver Schedules A . Scheduling Methodology  Appendix F: Costing Analysis Results A . Identification of Outliers, an Example B. Ordering Policy Parameters  38 39 49  50 52 53  57 61 63  List of Tables Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table  1: Delivery vehicle cost 2: Passenger boarding cut-off times 3: Uploading time estimates 4: Total added vehicle costs 5: Added man-hour costs 6: Added cost summary 7: Reduced system analysis results 8: M e a l type variables 9: Current L A P meals 10 : Sample first policy results 11: Sample second policy results 12: Source data tables 13: M e a l bank wastage cost estimation 14: M e a l bank wastage cost summary 15: V a n driver schedules 16: Flight assignments 17: Detailed costing results 18: Detailed service level results 19: Outlier events 20: Identification o f an outlier 21: Booked load history for an outlier 22: Ordering policy parameters  18 19 20 25 25 26 27 35 35 36 37 49 50 51 52 53 58 60 61 62 62 63  iv  List of Figures Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure  1: M e a l quantity vs. passenger load 2: Catering operations overview 3: M e a l production process 4: F l o w o f information and parts in the flight kitchen 5: M a i n order delivery procedure 6: M e a l delivery containers 7: L A P meal uploading procedure 8: Analysis date ranges 9: Second ordering policy performance 10: Reduction factor sensitivity 11: Intermediary upload parameter sensitivity 12: Sample final passenger load variability 13: Costing database user interface  2 3 4 5 6 9 10 12 13 22 23 36 40  v  Acknowledgments M y thesis was completed with the help and support o f numerous persons and organizations. I would like to take the opportunity to acknowledge their efforts. I would like to thank Associate Dean Martin L . Puterman for his help i n writing the thesis document and carrying out the project, and Professor Derek Atkins for taking the time to be part o f my thesis examining committee. I would also like to thank the Centre for Operations Excellence for providing facilities and funding to master student projects like mine. Special thanks go to Stephen Jones and Andre Powell for their guidance and assistance. I would also like to thank all students and staff involved i n the project. I extend m y gratitude to Canadian Airlines International for providing an interesting project as well as financial support. I would especially like to thank the V i c e President o f Inflight Services Marshall Wilmot, Director o f Catering Services, Michael Joss, Manager Aircraft Provisioning, M a n l y Sitter. Thanks to Mandy Zhou from the Inflight Services finance department for her patience and essential help. I would also like to thank David Lee and Steve Passmore from L S G Lufthansa Skychefs. A l l staff mentioned and many others were instrumental in providing resources, describing the systems, and facilitating access to data. I would like to thank my parents for their long-distance encouragement and support through the rougher times. Financial assistance received from the Natural Science and Engineering Research Council ( N S E R C ) through the Industrial Postgraduate Scholarship was greatly appreciated.  vi  I - Introduction This document presents the final report of my Master of Science in Management Science thesis project. This project was done as part of the Centre for Operations Excellence (COE) Master Programme between September 1998 and December 1999. As for all COE projects, it was conducted with one of the Centre's partners: Canadian Airlines International (CAI). Canadian Airlines is the second largest national carrier. In 1998, Canadian carried more than 12.5 million passengers to over 245 destinations worldwide in 23 countries on five continents. In Canada, Canadian flies to 109 destinations. 1  The work done in the present project concerned the catering area of Canadian Airlines' operations. The Meal Provisioning Project consisted in the development of a proposal for improvement in the meal ordering, production, and delivery system as a whole. The study and proposal covered CAI operations at Vancouver International Airport. One of the key features of the airline industry is the variability in passenger demand. Variable demand creates uncertainty as to the number of passengers that will board any given flight. This uncertainty makes the task of consistently providing the appropriate number of meals to each flight very challenging. As a result, large expenses are currently incurred by overages in meal provisioning. In addition, service quality can suffer significantly when a flight is under-catered. The Meal Provisioning Project has two key objectives. The first objective is to reduce meal-catering costs. Simultaneously, the project strives to improve service level, by reducing under-catering instances. These twin objectives have been addressed through an in-depth review of the catering process and the application of management science techniques. A similar document was presented to Canadian Airlines management outlining the potential costs and benefits associated with implementing a revised catering policy. The approach has been to build a management science tool that allowed us to study the impact of various catering policies on service level and cost. This thesis document gives a more detailed background and explanation of the methods used to arrive at the recommendations presented to CAI management.  ' Canadian Airlines International  web site, http://www.cdnair.ca  il  II - Background 2  The Meal Provisioning Project follows on the work accomplished by Jason Goto . In his project, Goto completed an investigation of the current catering system performance. The final report identified a potential opportunity to reduce under-catering of flights as well as meal wastage costs through a comprehensive analysis of the catering policy. The scatter plot in Figure 1 below summarizes some of the key findings of his study. It shows the number of over-catered flight instances (dots above the 45-degree line) and the number of under-catered instances (dots below the line). The objective is to bring all dots as close as possible to the 45-degree line by more closely catering at the final passenger load. Meal  Quantity 400 350' 300 250 200 150 100' 50 0  |I I I I | l l I I |I I I I | I I I I | I I I I | I I I I | I I u  0  50  | I I I I |  100 150 2 0 0 2 5 0 3 0 0 3 5 0 4 0 0 Passenger  Load  F i g u r e 1 : M e a l quantity vs. passenger load  Goto also developed a Markov Decision Process (MDP) model for the meal ordering system at Canadian Airlines. This model provided an optimal decision rule for each individual flight at multiple decision epochs prior to departure. For example, the optimal policy for the decision epoch at 3 hours prior to departure would tell the system to increase the order quantity to 54 meals given that the current passenger load is 58 and that the meal order at the previous decision epoch was 48 meals. The approach for my project is to propose process changes for the catering operations and evaluate their feasibility with regards to costs and logistic concerns. The ordering policy is only one of the areas of the project. Goto's model was considered as a potential solution for the meal ordering aspect of the proposed changes. 2  Goto, Jason H . (1999). A Markov Decision Process Model For Airline Meal Provisioning. 2  A. The Catering Operations  The current operations involved in provisioning flights with a meal service are described here. Understanding and mapping these operations has been an essential part of the project. Flight Kitchen  Meal Production  Commissary  Equipment Assembly  Delivery  Figure 2: Catering operations overview  The catering operations are divided between two main departments: the flight kitchen, and the commissary department. The flight kitchen is responsible for meal production, assembly, and storage. A meal refers to the main service for any given flight. The meal is usually composed of two parts: a main dish or casserole, and a Tray Set-Up (TSU) that usually includes a salad, a bun, a dessert, and utensils, served on a tray. Other meals can be served in a basket or lunch box, but are still stored on a tray. The commissary department is responsible for assembling all items that make up the galley. The galley is the set of containers and equipment that is loaded and unloaded on the aircraft before and after each flight. It includes all food items like the meals coming from the flight kitchen, but also dry stores (e.g. peanuts), alcoholic and non-alcoholic beverages, cooking equipment and utensils, magazines and newspapers, and headsets. Every day, the commissary staff assembles the galley equipment for each flight inside the caterer's warehouse and delivers it to the aircraft via high-lift truck. Meal Production The flight kitchen operates in a batch process flow structure. The meal service for each flight is treated as an individual entity that goes through the process flow illustrated in Figure 3. The production quantity is scheduled in the preliminary production sheet completed by the information services group and transmitted to the Chefs department (kitchen). The Chef will then order ingredients from the store room. The store room staff delivers batches of ingredients mapping the requirements of each individual flight. The ingredients are assembled and cooked in the kitchen resulting in the main dish or casserole part of the meal. Simultaneously, the material handling & delivery staff assembles necessary components for the TSU's. The ingredients, which can include lettuce, cheese, plates, and buns, are delivered on a rack to the assembly personnel. The assembly staff is organized in individual workstations according to flight batch sizes. They assemble the TSU and casserole on individual trays and store them in galley containers. These containers can either be trolleys, large metal boxes on wheels, or carriers, smaller metal boxes with a handle (see Figure 6). The containers are grouped by flight number and by the location of their destination inside the aircraft (forward, center, 3  or aft). The grouped containers are stored i n the refrigerated holding room. The order checker w i l l verify that meal quantities inside the containers match the updated and final order sheet before commissary employees deliver them to the aircraft.  Time prior to departure Preliminary sheet production  36 to 24 hours  24 hours  Order ingredients t  TSU assembly  \  18 to 14 hours  Carrier/trolley assembly  14 to 3 hours  Check order  3 to 1 hour Delivery  V Aircraft  Figure 3: Meal production process The information services group controls the flow o f information and parts throughout the flight kitchen as illustrated in Figure 4. They establish production quantities based on booked passenger loads, a forecasting tool, and the manager's judgment and experience. Information services also handle communications with individual flight controllers at the  4  airport terminal. A flight controller can make adjustments to the meal order for the flight under his or her responsibility up to one hour prior to departure. Information on order quantities is transmitted to the store room on a longer aggregate basis for inventory management and ordering from suppliers. As mentioned earlier, the kitchen receives the preliminary production sheet in order to determine production quantities.  Information Services  Store Room  Ware-wash  Kitchen  Information  !  Material Delivery & Handling  •  JAssembly w  Equipment Delivery & Handling  Checkers  Parts  Delivery  F i g u r e 4: F l o w of information and parts i n the flight kitchen  The assembly is controlled through a Variable Production Schedule (VPS), also completed by information services. This information is used for assigning assembly staff and production quantities. Employees are hired either permanently or on call. The number of permanent employee is set to a low level so that temporary employees are almost continuously needed. This results in an efficient use of staff. No set up cost is required when calling in temporary employees. Both permanent and temporary employees receive an hourly wage.  Commissary Operations Commissary operations are divided into two main tasks: assembling galleys, and delivering them to the aircraft. The assembly is done in a similar batch process flow where each flight is an entity. The equipment is stored inside carriers and trolleys (see Figure 6) that are in turn grouped by flight number. The assembled galleys are temporarily placed in a holding area of the warehouse next to the door leading to the docking bay. Delivery drivers pick-up assembled galleys, roll them onto the truck, and add the meal containers from the kitchen's refrigerated holding area (see Figure 5).  5  The caterer uses a combination of high-lift trucks and vans for delivery operations. Highlift trucks serve for delivering the full galleys and meals to flights (referred to as the main order in this document). A team of two employees is assigned to each delivery on highlift trucks. Vans are used to answer any last-minute request for catering equipment or small meal quantities prior to the departure of a flight. A single driver operates a van. If an irregular situation at a flight requires a large last-minute upload, high-lift trucks can be used at a higher cost. Holding fridge  Loading dock  Flight kitchen  Commissary Warehouse Delivery trucks  Assembled galleys  Figure 5: Main order delivery procedure Commissary Operation Scheduling  The commissary dispatcher is responsible for scheduling delivery operations. The dispatcher has access to an R G G (SABRE) terminal. This information system provides him with up-to-date flight departure and arrival times. Every morning the dispatching staff prints out the list of flights for the day and assigns trucks to each of them using their truck number. The dispatcher basically links the R G G schedule to a truck schedule. In the truck schedule, each row represents a truck/team and each cell represents an assignment. For example truck 105 is assigned the aft galley of flight CP007 at 12:00. The first dispatcher shift starts at 4:00. Three truck schedules are produced during a day: from the first flight to 10:30, 10:30 to 16:00, and 16:00 to the last flight of the day. The truck schedules are revised continuously as the day progresses. These schedules are all paper based. The dispatcher uses a highlighter colour code to designate truck assignment status. The status can be assigned (matching RGG), in progress (truck has left the kitchen dock and is loading/off-loading), or done (returned to kitchen dock). On May 6, 1999 the caterer handled 74 flights at Vancouver International Airport; 56 of them were Canadian Airlines flights.  6  Communication  Truck and van drivers are equipped with a radio transmitter linking them with the dispatcher's desk. Communication is essential to the van operations. It allows aircraft crews to send last minute requests to the vans through the dispatcher: The quality o f the transmission received i n a van can make it very hard to understand the dispatcher. Resources  The caterer owns 25 high-lift trucks. O n average, one truck is kept aside on maintenance rotation. The caterer also has 9 vans. A l l commissary galley equipment is owned by the airline. N o usage fee is charged to the caterer and inventory management and replacement is the airline's responsibility.  The Late Augmentation Plan (LAP) The Late Augmentation Plan was instated over a year ago as a means to provide meals to flights at the last minute prior to departure. It was implemented only for a subset o f Canadian Airlines flights (transcontinental and transborder). The L A P can be considered as the precursor to the meal bank system proposed in this document. The L A P is different from the regular meal provisioning system described above i n both the meal content and the delivery procedure. A L A P meal is produced for a set o f flights bearing the same characteristics in aircraft type and route sector. The L A P meal diversity is detailed i n Appendix A . These meals are produced using the same process as any other regular flight. They are stored in a separate holding area and are identified by L A P meal numbers. L A P Meal Production  The business class L A P meal casseroles are prepared in the afternoon o f the preceding day. Quantities must be established 24 hours in advance. Economy class L A P meal casseroles are pop-outs. Pop-outs are complete ready-made casserole entrees purchased from a supplier. One supplier handles all C A I kitchen requirements in Canada. W i t h such a large demand, the airline can gain economies o f scale by using pop-outs rather than having the casserole dishes made in-house. Note that pop-outs are used as the casserole for all regular Transcontinental economy-class meals. Within the next year, they should be used for economy-class meals on all routes. Pop-outs are easily accessible from any North American station as well as European stations. Pop-outs are high quality meals even i f they are less expensive than in-house produced casserole dishes. Pop-outs are ordered every month in large quantities from the supplier and stored at an intermediary supplier location where they can be retrieved on a daily basis. A s long as they are kept frozen, these pop-out dishes can last almost indefinitely. A s the final load control sheet becomes available i n the afternoon o f the day prior to departure, the necessary amount o f pop-outs for a given flight is taken out and left to thaw overnight  7  (about 18 hours). Since they arrive in plastic containers, the pop-out contents have to be transferred onto a casserole dish. The pop-outs can then be joined to the T S U just as any casserole made in the flight kitchen. The L A P T S U ' s are assembled during the graveyard shift or very early in the morning. Complete meals have a life span o f approximately 24 hours as long as they are kept refrigerated. LAP Meal Wastage It is not possible to know precisely how many L A P meals w i l l be used on any given day at the time o f production. This is why more meals are generally produced then necessary. A s long as they are kept refrigerated, meals can be used over a 24-hour period. The meals are tagged with a colour code identifying their date o f production. The unused L A P meals are pushed to the next day's usage. I f the meals are not used after the second day, they w i l l be discarded. The airline shares the cost o f wasted meals with the caterer. The non-food part o f the T S U is re-used instead o f being sent to washing thus saving T S U assembly labour costs. LAP Meal Delivery Each LAP-designated flight is assigned to a delivery van. There are usually two van drivers assigned to Canadian Airlines flights. O n top o f L A P delivery, the van drivers must respond to any request from all catered flights. For example, these requests can be for a missing coffee pot or a missing oven tray. One van is assigned to last minute flight requirements during the whole day (7:00 to 19:00). A second van is involved i n other miscellaneous operations, like handling duty-free transportation, in the morning and supports the first van for last minute flight requirements during the peak period (12:00 to 15:00). V a n drivers visit all the L A P flights to which they are assigned. They visit other flights only i f required. The van drivers' responsibilities require a quick response time. They have to move from gate to gate and often have no time to return to the flight kitchen base in between. The vans therefore carry all sorts o f equipment. This leaves little room for L A P meal containers. They can currently fit two or three carriers comfortably. The vans are not refrigerated so dry ice must be used to keep meals cool. The uploading procedure for L A P deliveries is as follows (see Figure 7). The driver shows up at the boarding gate o f a particular flight approximately 15 minutes prior to departure. The driver exits his vehicle and accesses the gate entrance inside the airport terminal using the exterior stairs that lead to the access bridge. The driver asks the flight controller about any additional meal request. I f any are needed the driver goes back to the van and pulls out the necessary quantity o f meals. L A P meals are transported i n carriers containing 7 (wide-body aircraft) or 8 (narrow-body) meals. The van driver is responsible for delivering the meals to the appropriate container inside the aircraft galley. Passenger flow disruption is an important issue in the L A P delivery operation. The driver enters from the front o f the aircraft and sometimes has to bring carriers to the center or aft galleys. A l l o f this takes place while passengers are boarding the plane and  8  getting to their seats. It can be hard for the driver to return to his truck against the incoming flow of passengers. Another issue concerns the security of passengers during the meal loading operation. Meal trays are transported in metal carriers with sharp corners. As the driver walks down the aisle, he must be careful not to hit passengers. Deadhead Equipment Each airport in the airline's network of destinations holds an inventory of commissary equipment. Each time an aircraft flies it moves part of this equipment from one station to another. Assuming all point-to-point transfer quantities are constant from day to day, inventory levels should be balanced across all stations. This is why the airline uses deadhead equipment - catering equipment without any food - i f a given flight is not full. For example, a flight with a capacity 88 economy class flies with only 50 passengers on board. Assuming that exactly 50 meals were prepared, the remaining 38 slots will be filled with deadhead tray equipment sets. Although deadheading solves the equipment balancing issue, it creates operational costs. Handling fees are charged for each deadhead tray since they need to be assembled at one end and washed at the other. Deadheading adds weight to the aircraft load and increases fuel costs. It also creates additional work for the L A P meal delivery personnel. When uploading meals inside the aircraft, the van driver has to swap a deadhead tray for a L A P meal instead of simply inserting the added meal.  Carrier  Trolley  Figure 6: Meal delivery containers  9  Airport terminal  Figure 7: LAP meal uploading procedure  Ill - Methodology and Approach The project's approach to improving catering performance is through the application of a meal bank system. The study examined the feasibility and net benefits of such a system. Using a meal bank implies reducing the main order meal quantity and uploading meals in the final minutes before departure. This way, it is possible to fill in any gap between the final passenger count and the number of meals boarded. Our approach was to first perform a preliminary analysis of potential gains. Based on the success of the preliminary analysis, we investigated current operations in order to understand the logistics of the meal bank system and define the scope of the project. We formed focus groups during the pilot study phase to understand the processes and to formulate feasible process changes. The recommendations in operational changes were then incorporated in a thorough data analysis of the meal bank model. Finally, a costing analysis was performed based on the model output, current costs, and added cost from process changes. The major steps in the methodology are detailed below. A. Preliminary Meal Bank System Analysis  The meal bank concept suggests that the main meal order should be purposely made smaller than the expected passenger load. The meal bank then fills any need for meals in the last 15 to 20 minutes. As a first analysis, we were looking for an appropriate main order size for the system such that the final upload requirements can be met. We approached the problem with forecasting techniques, using past information to predict what the future will be. The first application of the meal bank concept was done using data on final passenger loads only. Data Source  Data was obtained from the Terradata system at Canadian Airlines. This system contains information extracted from TAPS (Total Aircraft Provisioning System) and AMOS (Automated Meal Ordering System). Some Terradata tables contain the exact same information from TAPS and AMOS while some other present summary information built from cross table queries. The SABRE group in Dallas, Texas manages all Terradata systems. The tables are accessed by Canadian Airlines through ODBC drivers using MS Access. Pre-departure information was used for the purpose of preliminary analysis. This information is broken down to a flight number, city pair, date, and passenger class aggregate level. This table consists of measurements on booked passenger load and checked-in passenger quantity made at 12 different query periods prior to departure. For 11  analysis purposes, we were focusing on the last 6 query periods (6 hours prior to departure, 3 hours, 2 hours, 1 hours, 30 minutes, and post-departure).  Analysis Method The first step was to extract data from the appropriate table i n M S Access. Secondly, we imported the data to a pivot table in M S Excel. The analysis was restricted to flights departing Vancouver International Airport, the largest station i n Canadian Airlines' system. Time series data on final passenger loads were used. A n initialization set was formed using data segmented by flight number and day o f the week. The average and standard deviation was computed for each time series segment o f the initialization set. The last month o f data was used as a test set or holdout set. A n ordering policy was applied on this test set and results were compared to the actual data. The performance o f the policy was evaluated by assessing the quantity o f meals from the meal bank that would have been needed in order to supply for any gap between the number o f passengers and the main meal order. 2/1/98  7/1/98  1  7/31/98  1  h — • .  J  "Y  Initialization set  Test set  Figure 8: Analysis date ranges First Ordering Policy The first proposed policy was strictly based on final passenger load data. It was a simple approach i n which the standard order was the average final passenger load, as computed on the initialization set, less one or two standard deviations. Final passenger load forecasting was done on a flight by flight basis and on a day o f the week basis. F o r example, we used past final loads for flight CP912 on Wednesdays to predict the final load on this flight on the present day (given that it is a Wednesday). Large variability i n final loads resulted in poor forecasting performance. This forced the system to heavily reduce the main order size resulting in unreasonable last minute upload quantities. See Appendix B : Preliminary Analysis for details.  Second Ordering Policy A second approach used historical pre-departure data. This data not only contains final passenger loads, but also pre-departure booked loads. Here, past information is used to understand the nature o f the volatility o f booked passenger loads to determine an appropriate policy and process.  12  From the pre-departure information an initialization set was formed on which we calculated the average and standard deviation o f the difference between the final passenger load and the number o f passengers booked six hours prior to departure. The policy calculates a main meal order quantity by taking the average difference between the 6 hours prior to departure and final passenger loads, subtracting one standard deviation from it, and adding it to the current number o f passengers booked six hours prior to the departure o f a given flight. The formula is shown below. Variations to the number o f standard deviations in the formula could have been used to obtain an optimal performance. However, this being the preliminary stage o f the analysis, one standard deviation was used i n order to prove the feasibility o f the meal bank concept. Sample data on this analysis is included in Appendix B : Preliminary Analysis.  Second Ordering Policy Let Pax be the number of passengers booked at the 6-hour pre-departure point Let Pax be the number of passengers checked-in at departure time Let D i f f be the difference between Pax and Pax Rule: Pax + [Mean(Diff _ ) - Standard Deviation(Diff )] = Main Meal Order 6  0  60  0  6  6  6  0  60  Analysis o f the second policy yielded encouraging results. These results, shown i n Figure 9, were presented to C A I management i n late January1999. The meal bank system successfully passed the initial feasibility step. The objective was to create a final upload requirement i n the order o f 10% o f total meals and reduce overage as low as possible. Meal Bank Requirement  Overage Performance  14.0% g 12.0% i  10.0%  s £  8.0%  0 6.0%  1 4.0% S  Q.  2.0% 0.0%  ^ <P <tlDate (July 1996)  <!»• <t°  -i? Date (July 1998)  Figure 9: Second ordering policy performance B. Investigation of the Meal Catering System The implementation o f a meal bank system required at least a partial re-design o f the catering process. Prior to any recommendations it was essential to conduct an investigation o f current operations. This investigation allowed for a better understanding o f the " A s - I s " environment and provided insight into opportunities for improvement. Following the investigation, focus group meetings were held with resource personnel within both the airline's and the caterer's organization. During these meetings, opportunities for improvement were described and proposed changes were evaluated.  13  The deliverables from the workgroups are listed in the Discussion and Recommendations section o f this document.  C. Scope of the Project The project scope includes the following five key areas described in detail below: 1. 2. 3. 4. 5.  M e a l Ordering Policy M e a l Production Control and Inventory Management Equipment Balancing Transportation Logistics Aircraft Operations  Meal Ordering Policy The ordering policy is a key area o f the project. A s determined i n the preliminary scenario analysis phase, substantial savings can be obtained i n reducing meal wastage by using an improved policy. A n automated database system was designed using M S Access to simulate different ordering policies and to evaluate the performance o f each scenario. This analytical tool allowed for presenting the multiple alternatives to Canadian Airlines management and determining the appropriate parameter values. The database system is described in details i n Appendix C : Costing Analysis Database.  Meal Production Control and Inventory Management The central concern i n meal production resides i n the management o f the assembled meal inventory. This inventory system has daily delivery and reorder points. The items in the meal inventory not only have a short lifetime but also are highly differentiated due to the different possible meal services across all flights. A large component o f the potential savings resides in the reduction o f meal wastage. A n alternative meal design and delivery procedure was researched i n an effort to mitigate the wastage problem.  Equipment Balancing A sub-project emerged from the aircraft operations area with regards to the equipment balancing issue. A n alternative to the current deadhead policy would yield increased uploading performance and reduced handling cost. To analyze this problem, a prototype database structure has been developed for the purpose o f simulating catering equipment transfer across the major airports in the Canadian Airlines transportation system. This analysis is currently being conducted as a separate C O E project.  14  Transportation Logistics Logistics research has focused on scheduling delivery vans to flights that required visits given the ordering policy. The constraints have been dictated by results coming from research in the aircraft operations area. A preliminary schedule has been designed for Vancouver airport flights.  Aircraft Operations This area o f the project pertains to the operations that take place at the airport terminal and inside the aircraft as bank meals are uploaded. These operations also take place in the short time window between the moment when the final passenger load is known and the flight's scheduled departure. A recommended procedure was only defined after consulting with gate agents, flight attendants, and catering delivery personnel. The objective is to maximize the meal uploading capacity within strict time limitations.  D. Model Development & Consolidation of Results The ordering policy model is an analytical tool that allows us to test different operational policies, see how they would have performed, and compare them to historical data. Note that Monte-Carlo simulation was not used. It was designed using Microsoft Access with significant input from focus groups. W e have adopted the suggestions o f user groups where possible. The source data used in the model consists o f Terradata tables downloaded from Canadian Airlines information systems. Performance was initially evaluated on the basis o f overage reduction, final upload size, and underage quantity. In the consolidation phase, the model evolved from pure policy testing to overall cost evaluation. A t this stage, performance was now evaluated on the basis o f meal cost, under-catered meal quantity and overage cost. Costing formulas were integrated into the ordering policy model. The source data was also expanded to include actual financial data for cost comparison purposes. Added costs from catering process changes were estimated and included in the final results. The costing analysis w i l l provide the necessary information to support C A I management's decision-making. Appendix C : Costing Analysis Database details the structure and content o f the ordering policy model. A l s o , please refer to this appendix for details on data source issues.  15  IV-Analysis J  This section contains findings and results from focus group meetings and interviews with resource personnel. These findings relate to implementation issues for the meal bank system. M e a l bank service options and meal bank storage and lifetime are discussed in greater detail. Another part o f the analysis detailed in this section is the choice o f parameter values for the model's ordering policy. Finally, the analysis has been categorized into different meal bank implementation scenarios. The alternative scenarios are explained and costing results are detailed.  A. Meal Bank Service Options Part o f the investigation was dedicated to defining the appropriate type o f meal to be used for final uploads. Ideally the meal bank would be composed o f one type o f meal that could be served on any flight and would last for an extended period o f time. However, i n order to be accepted, the bank meals must comply with Canadian Airline's quality standards and be as close as possible to the main order meals. Different options were considered as a solution for the meal bank service. These options are listed below.  Traditional Tray Set-Ups (Pop-Outs) This option is currently used in the L A P system. It involves the use o f pop-outs for economy meals on transcontinental and transborder flights. For business class, the meals are very similar to standard meals. L A P meals are generally chicken-based due to chicken's overall acceptance and popularity. The chicken casserole is also usually the lowest cost option. A proposed variant is to fully cater flights with Tray Set-Ups ( T S U ) and keep a portion o f the T S U ' s devoid o f a casserole dish. The uploads would simply be casseroles to be inserted on incomplete trays. This system would imply full T S U cost on all flights. This approach was tried i n the past and abandoned since it was too expensive.  Meal Pouches M e a l pouches are low quality meals that can be reused from one flight to the next. These meals are very different from the regular service meals. Quality standards would require that the portion o f such meals used on a flight be kept too low to be acceptable for meal bank system purposes.  16  Shelf-Stable Meals This new type o f meal has the advantage o f being transferable to the next flight i f not consumed. It can stay at room temperature for extended periods o f time. These meals are manufactured in Thailand and are provided in sealed individual containers. The quality o f these meals is significantly lower than that o f pop out meals. :  B. Meal Bank Storage and Lifetime Daily wastage is a significant part o f the meal bank system implementation costs. A s explained in the L A P M e a l Wastage section o f the Background, these costs currently exist with the Late Augmentation Plan. Wastage analysis was based on the use o f 48hour life cycle meal bank. B y having a 48-hour lifetime, the unused portion o f a day's production can be used on the following day, thus drastically reducing meal wastage. The appropriate meal bank inventory level can then be set using fixed daily reorder points. The nature o f a long-lasting meal is detailed below. The key consideration in having a 48-hour lasting meal is keeping a constant chain o f refrigeration. This implies a need for refrigerated delivery vans. The cost o f a refrigerated vehicle is provided in Table 1 below. Food safety is a crucial requirement in the meal process. The catering system follows H A C C P standards in order to monitor and ensure a safe product. A s described i n detail further below, there is concern that a refrigerated delivery system would not meet safety requirements. This issue w i l l be determined in a subsequent phase i f the meal bank system is to be tested. The analysis for meal bank wastage cost was performed assuming a meal that could be reused on the following day. Results show an estimated total monthly wastage cost o f $6,600. Appendix D : M e a l Bank Wastage Cost details the estimation method and provides an example. Types of Lonq-Lastinq Meals  1. Shelf stable: This type o f meal can stay at room temperature for extended periods. The quality o f such a meal is much lower. 2. Resistant: The meal, i f it is kept refrigerated, w i l l remain at an acceptable quality and safety level. Small modifications to the economy class meals would convert it to a resistant meal. The only part o f the T S U that is not resistant for 48 hours is the tossed salad. It could be replaced by any type o f marinated salad. This change would result i n a small incremental cost o f less than 10% o f the meal cost. With business class meals, the casserole portion can last 48 hours i f kept refrigerated. However, there exists a problem with the appetizer and T S U . They are made with higher quality ingredients with very limited lifetime. The FlexMeal is a cold T S U meal  17  available as an alternative to the aisle service menu. Those meals are cold but due to the nature of their ingredients they are not designed to last more than 24 hours. Changing the business class meals tofitthe safety requirements of a long-lasting meal would require lowering their quality and appearance. Refrigerated van fleet  To make the long-lasting meal alternative possible, the transportation system must be refrigerated. The two possibilities are to retrofit existing vans or to purchase new ones altogether. Retrofitting Insulation Installation of refrigerating unit  $ 3,900.00 $ 6,185.00 $10,085.00  New refrigerated van Standard new van Warranty Insulation Installation of refrigerating unit Exterior graphics package  $27,400.00 $ 1,850.00 $ 3,900.00 $ 6,185.00 $ 621.00 $39,956.00  Table 1: Delivery vehicle cost HACCP standards  The Hazard Analysis Critical Control Point is an international standard inspection system. The key element of a HACCP system is its preventative nature and the exercising of control throughout the manufacturing process, at critical steps called Critical Control Points (CCP). This way, defects which could impact on the safety of the food being processed can be readily detected and corrected at these points before the product is completely processed and packaged. A ready-to-eat meal such as the bank meals is considered to be a Critical Control Point in itself. HACCP requires that critical limits be established as well as an adequate monitoring procedure. For example, the meals have to be kept at the adequate refrigeration temperature. This emphasizes the need for refrigerated trucks in a longlasting meal system. There is concern that the monitoring procedure during the delivery process would not be adequate. The human factor is affecting the monitoring of critical limits outside the kitchen fridge holding area. According to a Food and Beverage Analyst at CAJ, re-using meals over more than a one-day span through a refrigerated van system is risky due to unacceptable monitoring and human intervention. The current process would have to be modified in order to satisfy food safety requirements.  18  C. Aircraft Operations This section o f the process analysis concerns the last-minute meal uploading procedure or final upload. The most important aspects o f aircraft operations are defining the exact procedure, the time required to perform the upload, and the maximum meal upload size. A focus group was formed to study this particular part o f the catering operations. The recommendations from the aircraft operation focus group are included in the Discussions and Recommendations section. Important facts and issues to consider for the uploading procedure are detailed here.  Uploading Procedure and Timing In order to define an appropriate delivery procedure we must understand the issues, and people involved. It is essential to determine the exact required time window for the upload operations. This time window w i l l be used in designing the delivery van schedules. The minimum time window is dependent on the following requirements: • the final passenger load must be known, • sufficient time must be allowed for the uploading operation to take place without incurring any delays to the flight. Important Facts • • •  Passengers cannot show up later than the cut-off time in order to be allowed boarding. The cut-off times are shown in Table 2. The door o f the aircraft must be closed 5 minutes before departure at the latest. The driver usually takes carriers to the aircraft one at a time. Although some drivers w i l l handle two i n rushed situations.  A t the Check-In Counter A t the Gate  Domestic 20 minutes 15 minutes  Transborder & International 30 minutes 20 minutes  Table 2: Passenger boarding cut-off times Description of Uploading Operations The van driver is responsible for delivering the meals to the appropriate location inside the aircraft. Dropping carriers at the door is not an acceptable option according to the flight attendants' employment safety standards. If the quantity o f meals is reasonably small (5 or less) the flight attendants can store the meals in the front and move them later. The presence o f deadhead trays requires a swapping procedure as discussed in the Deadhead Equipment section o f the Background. This time consuming procedure could  19  be fully or partially eliminated i f deadhead equipment is reduced. This issue is dependent on the on-going equipment balancing project. A signing procedure is currently in place to ensure that the airline gets charged for the exact amount ordered by the flight controller. The van driver must walk up to the gate to obtain the required upload quantity and get the controller's signature for it. Once the final upload quantity is known the time requirements are defined by the following factors: • • • • •  Passenger status (boarding, sitting, still at the gate): I f passengers are still boarding this means that the longest process times are to be considered. Aircraft type: For a B737, it is possible to simply swap carriers. For others, all meals are in trolleys. This implies that trays must be inserted individually. Receiving galley: (front, center, or aft) Except for the B737 and A 3 2 0 aircraft, a space should be made available in the center galley for meal bank increase purposes. Deadhead equipment: i f deadhead trays are present they must be swapped. Equipment requests: whether or not the driver is responsible for requests for other missing equipment, i.e. coffee pots, oven trays, etc.  Time estimates The time taken to get to the gate for order confirmation is not included ( i f so add another 2 to 5 minutes). It is assumed that the upload is delivered to the center galley on widebody aircraft. N o time is included for other equipment requests. The variation is dependent on passenger status. These estimates are based on an interview with a delivery van driver. Note that time estimates w i l l vary from one driver to the next. The values listed below account for worst case scenarios.  Upload Quantity Required Time 1 carrier, J-class 1 carrier, Y-class 2 carriers, Y-class 1J+ I Y 1J + 2 Y or 3 Y  2 to 4:30 minutes 5 to 10 minutes 10 to 15 minutes 7 to 12 minutes 12 to 17 minutes  Table 3: Uploading time estimates Maximum Upload Size The number o f meals that can be uploaded directly affects the model's performance. In fact the maximum upload is one o f the parameters that define a test policy. W e want to have as large a maximum as possible but the model must respect logistic constraints. The constraints are dictated by the combination o f limited meal carrying capacity and limited available time.  20  The final uploads are delivered by vans. There is a limit to the upload quantity that can be reasonably handled by a van driver. Here are the facts to consider: •  The maximum quantity that can be carried in one trip up the stairs is limited to 1 or 2 carriers.  •  The driver can make an additional trip i f multiple carriers have to be uploaded.  •  Due to the injury risk, transporting two carriers at a time could become a safety issue and be completely ruled out.  For analysis purposes, we assume a maximum upload quantity o f 3 carriers, 24 meals (narrow-body) or 21 meals (wide-body). A l s o note that for international flights the limit is set to 10 passengers since two meals are served. A procedure has to be defined for cases where the required upload goes beyond the limit. The alternatives are the following: • •  Use shelf-stable meals to be boarded as a standard for meal bank flights. Use a high-lift truck for the final upload delivery instead o f a van. This implies that a high-lift truck is assigned on a regular basis to the final upload o f the flight in question since 20 minutes is too late for calling on a unplanned high-lift truck delivery. High-lift trucks are much more expensive than vans and require double the manpower costs. This makes using the high-lift truck for regular operation highly non-practical.  In the analysis o f the meal bank model, we assumed that i f the upload requirement goes beyond the maximum quantity, the flight is under-catered.  D . Ordering  Policy  Analysis  The ordering policy is a function o f different user specified parameters. The database model, detailed in Appendix C : Costing Analysis Database, was designed so that different parameter values could be tested. There are two areas for parameter analysis: the main order reduction factor and the intermediary upload decision rule.  Main Order The parameter determining the main order policy is the reduction factor, a number multiplied by the standard deviation in the formula as used i n the second policy o f the preliminary analysis phase. Note that the main order is made at the 3-hour pre-departure point in accordance with the caterer's contractual agreement.  Main Order Formula Let Pax be the number of passengers booked at the 3-hour pre-departure point Let Pax be the number of passengers checked-in at departure time Let Diff _ be the difference between Pax and Pax Rule: Pax + [Mean(Diff _ ) - Reduction_Factor*StdDev(Diff )] = Main Meal Order 3  0  3  0  0  3  3  0  3  3  0  21  The average and standard deviation are trimmed in order to be unaffected by past outlier events. This was achieved by screening out Diff values that where outside o f an absolute maximum value o f 30 meals. 3 0  Different reduction factors were tested in order to determine an appropriate policy. The sample data in Figure 10 was obtained using long sector transcontinental flights in June 1999. The use o f one standard deviation is justified as an adequate compromise between cost savings and under-catered instances. A s we see on the left hand side graph, the total monthly meal cost start to rise more significantly as the reduction factor becomes smaller than 1. Although not ideal, the underage level at one standard deviation (155 meals) is significantly lower than the actual under-catering for June 1999 (272 meals). The problem is that a small number o f exceptional instances drive the under-catering figures. This is explained i n the Results section and detailed in Appendix F : Costing Analysis Results. Lowering the reduction factor is not an adequate solution to reduce meal underprovisioning since it drives the costs higher and over-caterers most o f the flights. This is similar to a situation where excess inventory is used to cover productivity problems. Total Monthly Meal Cost  Total Monthly Underage  $220,000 $215,000 $210,000  s 100  $205,000 $200,000 $195,000 0  0.5  1  1.5  2  Reduction Factor  0  0.5  1  Reduction Factor  Figure 10: Reduction factor sensitivity Intermediary Upload The meal bank system includes a delivery van visit between the main order delivery and the final upload. This visit, referred to as an intermediary upload, allows for making meal quantity adjustments when an unexpected shift in passenger load occurs or the flight controller is aware o f special situations such as a delayed connection or a mechanical problem on a competitor's aircraft. The intermediary upload can be a simple adjustment to the main order only i f it is still i n the holding fridge. This is very unlikely for the 1-hour period, but possible on certain flights (short sector - narrow bodies) at the 2 hour count. Experience shows that generally, at the 1-hour count the main order cannot be adjusted and a separate delivery is required. The ordering policy database model uses expected meal shortage to determine i f the intermediary upload is necessary. The upload size can either be 2 carriers (14 meals) or one carrier (7 meals). The policy is a function o f two parameters: M a n d m. These  22  parameters are respectively referred to as Minimum trigger for uploading 14 meals, and Minimum trigger for uploading 7 meals i n the costing database form.  Intermediary Upload Formula Let Pax! be the number of passengers booked at the 1-hour pre-departure point Let Pax be the number of passengers checked-in at departure time Let Diff-,_o be the difference between Pax and Pax! Let Expected Final Upload = Pax! + A v g ( D i f f ) - Main Meal Order Rule: If Expected Final Upload >= M, then order 14 meals If Expected Final Upload between M and m, then order 7 meals If Expected Final Upload < m, then make no order at this point 0  0  10  Analysis on the intermediary upload was also conducted to determine appropriate (M, m) values. The sample data shown i n Figure 11 below was obtained using the same flights as above with a reduction factor o f 1 standard deviation. Four different (M, m) pairs were tested. The chosen pair was M= 18 and m = 10. This parameter choice is an adequate balance between meal cost and underage. The choice o f intermediary upload parameters does not have a significant impact on cost except maybe for the smallest pair (14,7) that is slightly higher than others. A s expected, the underage decreases as the parameters are reduced. The true benefits are achieved by simply having an intermediary upload available. The choice o f these parameter values is not crucial. Total Monthly Meal Cost $220,000  Underage  ,  200  $215,000 $210,000 $205,000  »  —  $200,000 $195,000 J  , 14 7  , , 18 20 10 12 Intermediary Parameters  =  Oj  28 14  ,  14 7  ,  18 10  ,  20 12  28 14  Intermediary Parameters  Figure 1 1 : Intermediary upload parameter sensitivity E. Scenario Analysis This section details the different scenarios that were studied i n the costing analysis. Note that the analysis was performed exclusively on economy class service. Assumptions on parameter values, added costs and procedures are detailed as well as final costing results.  Full Meal Bank System Implementation Scenario In the F u l l M e a l Bank System Implementation all flights are included i n the meal bank program. •  A van is scheduled to visit each flight every day before departure time.  23  • •  The schedule allows for intermediary uploads for wide body aircraft only. The main order is purposely reduced for all flights, based on a reduction factor o f 1.  Reduced Meal Bank System Implementation Scenario Here, only international and long-sector transcontinental flights are included in the meal bank program. The catering procedure for other flights (short-sector transcontinental and transborder flights) is unchanged from the present situation. • A van is scheduled to visit each international and transcontinental-long flight every day both at final and intermediary upload periods. • Flights departing later than 16:00 are excluded from the system. This only excludes flights 133, 996 and 976 from the meal bank system and greatly reduces the van driver shift requirements. • Transborder and Transcontinental-short flights currently in the L A P program remain as is. •  The main order is reduced for international and transcon-long flights only, based on a reduction factor o f 1.  Assumptions Main Order Reduction M a i n order reduction for scenario analysis purposes is defined as using a reduction factor o f one standard deviation. Provisioning Objective It is assumed that the objective is to provide one meal for each passenger on board. Note that in fact, Canadian Airlines designates some o f its flights as partially provisioned. These flights are catered to 90 percent o f the passenger load since a significant portion o f passengers usually decline meal service. These are usually short-sector flights with passengers connecting from a previous flight with a meal service. Maximum Final Upload N o documentation or set rules determines what could be the maximum upload quantity handled by L S G commissary personnel. The performance o f the meal bank system is directly affected by the maximum upload limit. Costing analysis results are shown for a maximum upload quantity o f 3 carriers. This implies a maximum meal quantity o f 21 meals for transcontinental flights and 10 for international flights. Additional Vans It was assumed, for analysis purposes, that refrigerated vans would be used. This is necessary for allowing the re-use o f meals over a 24-hour period. T w o vans are already available but would need to be retrofitted with a refrigeration system. T w o new  24  additional vehicles would be purchased. Costs for additional vans are discounted over a 5-year period using lease calculator estimation. Two new refrigerated vans Purchase price Down payment Lease amount Term Interest Rate Buyout value Monthly payment Total (annual)  Retrofitting two existing vans $ 40,000.00 /van $ $  $ $ $ $  40,000.00 5 6% 10,000.00 625.00 1,250.00 15,000.00  Purchase price Down payment Lease amount Term Interest Rate Buyout value Monthly payment  years  /van for 2 vans for 2 vans  Total (annual)  $ 10,000.00 /van $ $ 10,000.00 5 6% $ 25,000.00 $ 156.25 $ 312.50 $ 3,750.00  years  /van for 2 vans for 2 vans  Fuel, Maintenance, Registration, and Insurance Two tanks/week  $  $ Annual Maint. (annual)  $ $  $ Regist & Ins. (annu$  $ Total (annual)  $  60.00 120.00 6,240.00 3,000.00 6,000.00 1,200.00 2,400.00 -14,640.00  /van for 2 vans /van for 2 vans /van for 2 vans for 2 vans  Total Annual Added Cost:| $33,390.00 |  Table 4: Total added vehicle costs Increased Man-Hours (Van Schedule)  V a n schedules were built for evaluating the requirements for additional vehicles and drivers for each scenario o f the meal bank system. The detailed results are included i n Appendix E : Delivery V a n Driver Schedules. Summary results for added man-hour costs are shown here.  Current Reduced System Implementation Full System Implementation  Driver Wages $ 136,500.00 $ 250,250.00 $ 404,950.00  Added Cost $ $  113,750.00 268,450.00 All figures annual  Table 5: Added man-hour costs Meal Bank Wastage  Cost analysis was performed assuming meals can be re-used on the following day. This implies that meal wastage cost would be incurred i f meals are not used on the second day, The estimated amount as shown in Appendix D : M e a l Bank Wastage Cost is included in added meal bank costs.  25  Scenario Analysis Results The cost analysis was performed on each scenario. The full system implementation was rejected at an early stage since projected meal cost savings did not justify added costs. A reduced system concentrates on high meal service cost flights. The opportunity for improvement is thus higher. Flights with a similar meal service are selected to reduce diversity problems and allow pooling o f safety stocks. W e present the costing results for the reduced meal bank system (see Table 7). Results are divided between meal cost and meal underage quantities. The former shows the savings provided by the meal bank system and the latter shows the service level improvement. The savings are the difference between current billed costs and meal bank model output costs for flights included i n the proposed meal bank program. Net savings are obtained by subtracting the added cost summarized in Table 6 to the savings. Service level is measured by the total number o f meal shortages for a given month. I f this amount increases, service level decreases and a negative figure is displayed. Service level is shown for revenue passengers only as well as for all passengers. Vehicles  $  2,782.50  Man-Hours  $  9,479.17  Meal Wastage  $  3,288.35  Total Cost  $ 15,550.02  All c o s t s p e r m o n t h  T a b l e 6: A d d e d cost s u m m a r y Detailed costing analysis results for each flight on each month is provided in Appendix F : Costing Analysis Results. Note that irregular events are taken out o f the results. These are events where an unusual change in the booked load occurs between the main order and the final upload. A meal bank model based on average behaviour cannot handle such events. In the model, these flights w i l l suffer large meal underage. The method o f identification o f the outliers is detailed i n Appendix F : Costing Analysis Results. The net savings and percent underage reduction results vary significantly across test months. This variation is mostly due to the actual performance o f the caterer on each o f these months. If over-provisioning was frequent over a given month, the underage was usually low. In such a case, the cost savings are high and the underage reduction is low (e.g. February). The reverse situation occurs i f flights were actually provisioned at a lower level with more under-provisioning (e.g. March). The schedule change i n A p r i l resulted in high potential savings. During that month, all flights were over-catered as a safety measure for adjusting to the new schedule. W e also observe that the opportunity for improvement in both cost and service level decreases as the load factor increases. For example, June was the first month o f the high season for traveling, where most flights were full (high load factor). For that month, the net savings and percent underage reduction were low.  26  Meal Cost Month Feb-99 Mar-99 Apr-99 May-99 Jun-99  Savings (Loss) Net Savings (Loss) 29,690.50 $ 14,140.48 $ 11,941.15 $ (3,608.87) $ • 76,406.84 60,856.82 $ $ 21,832.09 $ 6,282.07 $ 17,050.00 $ 1,499.98 $  Underage - Revenue Passengers Only Month Current Feb-99 Mar-99 Apr-99 May-99 Jun-99  Meal Bank 0 16 43 66 33  Reduction 11 0 33 6 19  (11) 16 10 60 14  Percent Reduction 0.0% 100.0% 23.3% 90.9% 42.4%  1 140 95 149 2  Percent Reduction 2.8% 100.0% 49.0% 70.3% 1.9%  Underage - All Passengers Month Current Feb-99 Mar-99 Apr-99 May-99 Jun-99  Meal Bank 36 140 194 212 104  Reduction 35 0 99 63 102  Table 7: Reduced system analysis results  27  V - Discussion and Recommendations A. Meal Ordering The main order should be done using the 3-hour booked load, as the caterer is responsible to adjust the order up to this period. The main order is set to be lower than the current booked load using the following formula: Main Order = Booked Load . + Mean(Difference f i i - 3-hour) - Standard Deviation(Difference f i i - 3-hour) 3  hour  na  n a  The intermediary upload w i l l be done at the 1-hour pre-departure mark. The upload size is either 7 or 14 meals. The decision is made according to the following decision rule: Let Expected Final Upload = Booked Load i. + Mean(Difference If Expected Final Upload >= 18, then order 14 meals. If Expected Final Upload between 18 and 10, then order 7 meals. If Expected Final Upload < 10, then make no order at this point. hour  nn  ai-i-hour)  - MainOrder  B. Meal Production Control and Inventory Management  Main Order N o changes are proposed for the production o f meals for the main order. The main order w i l l be automatically generated. The kitchen w i l l simply use the output number and build the order accordingly. Use o f passenger load information prior to the 3-hour point can be incorporated into the model to provide the kitchen with production quantity forecasts.  Intermediary Upload The intermediary upload could be either a decision coming from the flight controller, from the kitchen, or automatically generated. Generally, the latter would be the preferred option since it has the advantage o f reducing the risk o f confusion and could reduce communication requirements, especially during busy periods o f the day. Automation also reduces the responsibilities o f flight controllers. However, flight controller intervention should be possible, and is required, to cope with irregular operations caused by canceled flights or missed connections.  Meal Bank Production The recommended meal bank w i l l be composed o f traditional T S U ' s with pop-out casseroles. The production process w i l l remain unchanged.  28  Four types o f meals w i l l be required: one for each service for international flights, one breakfast service, and one lunch/dinner service for transcontinental flights.  Meal Bank Inventory The bank w i l l be held in the same refrigerating area and tagged using the current system. Significant cost savings can be achieved through the use o f a long-lasting meal with a refrigerated van fleet. However, according to Food & Beverage Standards, the risk involved in the delivery process is a potential problem even i f vehicles are refrigerated.  C. Equipment Balancing The issue o f deadhead equipment reduction is i n progress. Aircraft operations w i l l be eased i f it is determined from the parallel equipment balancing simulation project that we do not require a full deadhead policy.  D. Transportation Logistics Scheduling for the reduced system scenario requires the use o f four delivery vans with five shifts. This requires the addition o f two vans and more driver shifts. Refer to Appendix E : Delivery V a n Driver Schedules for the base schedule.  E. Aircraft Operations The visit time for final uploads w i l l be 20 minutes prior to departure for all flights included i n the program. A n alternative way to communicate the upload quantity to the drivers is required. The final upload cannot be larger than 3 carriers.  Final Upload Order Communication The focus group team recommended the use o f gate telephone communication. It was also suggested to provide the van drivers with cellular phones. This w i l l reduce communication time by allowing the drivers to get the order quantity information from their vans before getting to the aircraft. Busy signals on the gate telephone could be avoided by equipping the line with callwaiting. The distinct ring option could also be added to the lines such that controllers would know that the call is coming from catering and give it a higher priority. The Vancouver International Airport Authority manages these gate telephone lines. W e could not get a final answer on the cost and feasibility o f such changes to the airport lines.  29  F. Scenario Selection W e recommend the implementation o f the reduced meal bank system as presented in Table 7. Testing o f such a system on five months from February 1999 to June 1999 at Vancouver International Airport confirmed the potential for average monthly cost savings in the order o f $15,800 and service level improvement o f approximately 50% less shortages for revenue passengers. The system should be targeted at economy class meals only. The production o f a meal bank for business class would not be practical unless quality standards are lowered. W e also observed that with new upgrade policies i n place, the business class load is close to capacity on most flights. A special procedure must be in place in the event o f a flight requiring a final upload exceeding the maximum amount. M e a l pouches or shelf-stable meals could be used to provide all passengers with a meal. Space should be dedicated for these meals on all flights as a precautionary measure. These meals may be especially useful for nonrevenue passengers on long-haul flights. Special attention should be given to irregular operation as a cause for poor service level performance. Outliers drive the number o f under-catered instances in the meal bank model results. The ordering policy tool cannot predict irregular behavior in booked passenger patterns. Several factors can explain such a behavior. Although the flight controller may be aware o f irregular operations as they are taking place, it is not clear how to accurately model the response without more data or anecdotal information. The month o f A p r i l causes problem due to the flight schedule change. The system makes decisions on meal orders based on past data. If new flights are added, no past data is available to calculate the passenger count difference distribution. W e propose a buffer period during which these flights are safely catered and data can be accumulated. N e w flights added to the schedule in A p r i l were excluded from the cost savings and service improvement estimates for A p r i l data and then inserted in the analysis for M a y .  G. Markov Decision Process Model A s explained in the background section, a possible solution for a new ordering policy is to use the Markov Decision Process model developed by Jason Goto. Using a simpler decision rule, the database approach was more flexible and adaptable to real data. Although it is not providing an optimal solution, the database model is a more practical and efficient approach to obtaining cost savings and service improvement estimates for the meal bank system. The M D P model was implemented using S A S . One o f the basic requirements for running the model on a given flight number is that the aircraft capacity would be constant across all dates in the test data range. This is essential since the M D P model calculates  30  optimal decisions based on past information for that flight where the capacity is constant. Unfortunately, the data on aircraft capacity contains a lot o f variability for most o f the flights. This results in incomplete output data where only flight dates where the aircraft capacity is equal to that entered as a run-time parameter are kept. This problem would require that the data be cleaned to have uniform aircraft capacity. Another issue with the M D P model involved the time required to run a test policy for each flight. The run time o f the M D P model is more reasonable for aggregate models o f the state space. This may introduce some error due to rounding. The use o f the M S A c c e s s database model developed in the current project presented in this document did not cause such problems.  H. Confirmation Study The purpose o f this section is to sketch some o f the major issues that the implementation plan must address as w e l l as some suggested approaches that may be used as a guide to building the plan. The following is not intended to be a substitute for detailed implementation planning. Prior to any implementation, a minimum two-week confirmation study should be undertaken in order to confirm the results o f this study. There are two broad objectives of the confirmation study. The first is to ensure that the recommended approach is considered feasible by users. The approach needs to be completely transparent to users in order to identify any deficiencies prior to implementation. The second objective o f the confirmation study w i l l be to provide an opportunity to get user feedback prior to the implementation. User feedback may provide opportunities for improvements that have not been identified i n the analysis. The confirmation study should use recent historical flight data and passenger load data. The month o f September 1999 could be used for this purpose since it is within the summer flight schedule but was not included i n the analysis. The goal is to develop an understanding o f what would have happened i f we had implemented the findings o f the study. In particular, we would like to ensure that the flights that required uploads would have received the necessary upload amount. Irregular operations need to be examined to ensure that, in such cases, human intervention could improve on the performance o f the model. W e w i l l also identify the benefits that would have accrued to Canadian i f the recommendations o f the study had been applied. Certain areas o f the proposed process improvements would need further analysis and discussion. These areas w i l l be addressed in the confirmation study. The first area is the flight schedule change at the end o f October. The delivery vehicle assignments should be revised according to the winter schedule to ensure adequate coverage o f meal bank flights. Second, the new communication procedure between the delivery personnel and flight controller requires testing. Third, we must confirm that sufficient time and meal  31  quantity is allocated for performing meal uploads. Fourth, we must ensure that the vehicle refrigeration system for reusability o f meals meets safety standards.  /. Implementation The implementation should be done in progressive steps. A first phase could be planned with a few flights within the same meal group using a single van. This step would provide additional consultation to ensure that personnel support the changes. Once this first step is confirmed another vehicle and additional flights could be added to the meal bank system. This progressive approach would eventually lead to a full roll-out o f the proposed changes.  32  VI - Conclusion In this document, we demonstrated the potential for service improvement and meal cost savings through a revision o f the catering process at Canadian Airlines. The study covered flight operations at Vancouver International Airport. The proposed system is restricted to economy class service on a number o f flights i n the international and longsector transcontinental groups. Through the present study, Canadian Airlines management gained valuable insight regarding the hard constraints o f the meal process. The study allowed for a better understanding o f the current operations and the costs associated with them. It also proposed improvements where appropriate. Following the completion and acceptance o f the proposed process changes, a detailed implementation plan is required i f the findings o f this study are to be implemented.  33  Bibliography Canadian Airlines International  web site, http://www.cdnair.ca  Canadian Food Inspection Agency (1998). Food Safety Enhancement Program (FSEP) Manual - Volume One, http://www.cfia-acia.agr.ca/english/ppc/haccp/manual/home.html. Goto, Jason H . (1999). A Markov Decision Process Model For Airline Provisioning.  Meal  Richard B. Chase (1998). Production and operations management: manufacturing services - 8 ed., M c G r a w - H i l l .  and  th  Makridakis, Spyros G . (1998). Forecasting: methods and applications - 3 W i l l e y & Sons Inc.  rd  ed., John  34  Appendix A: LAP Meal Diversity and Quality Ideally one type o f L A P meal would fit all flights. Unfortunately, differences in aircraft configuration do not permit this and as a result 11 different L A P meals are required.  Variable  Values  Possibilities  Class o f passenger Type o f meal  Business (J) or Economy ( Y ) Breakfast, cold lunch/dinner (J only), hot lunch/dinner 331,358, 1011, 1611 Royal Doulton, White  2 3  Tray dimensions China type (J only)  4 2  Table 8 : Meal type variables The class o f passenger, tray size, and china type variables are inter-dependent. They are related to aircraft type also. If it were not for the tray size and china type variables, 5 types o f meal would do; one breakfast meal for each passenger class, the same for hot lunch/diner, and one J-class cold lunch. Trays are necessary for trolley size limitations. China is an aesthetics issue, but deemed important to the airline. China is an issue only for J class meals. Eliminating china diversity could save significant costs in inventory stock reduction and employee time, on top o f simplifying L A P meal operations. Using a single type o f china could eliminate two o f the current 11 L A P meal varieties.  LAP LAP LAP LAP LAP LAP LAP LAP LAP LAP LAP  1 2 3 4 5 6 7 8 9 10 11  Type o f meal Breakfast Breakfast Breakfast Breakfast Breakfast Hot lunch/diner Hot lunch/diner Hot lunch/diner C o l d lunch/diner Hot lunch/diner Hot lunch/diner  Passenger class J J J Y Y J J J J Y Y  Tray size 331 1611 1611 331 1011 331 1611 1611 1611 331 1011  China Royal Doulton White  Royal Doulton White  Table 9: Current LAP meals If the L A P program were to be extended to all routes, the number o f L A P meals would increase for the following reasons: second meal on certain international flights, different meals on different route sectors, picnic baskets, short-sector flight meals on a 358 tray.  35  Appendix B: Preliminary Analysis  Winnipeg and Ottawa  p)  C  C  Edmonton Flights  S S § r? 5 £  §5  a aS ~ =: a —  = s a s  c  PI  Figure 12: Sample final passenger load variability  Capacity 88 leaving YVR on Mondays Flight 1 2 3 4 5 6 7 e 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44  Test Set 7/6/98 85 77 72 68 74 82 81 74 67 86 87 82 88 77 59 48 44 70 50 88 77 88 82 88 84 87 77 88 85 68 67 83 77 71 75 52 34 77 83 77 64 67 66 22  7/13/98 7/20/98 7/27/98 85 82 75 77 84 71 79 59 64 81 80 68 77 76 69 81 87 79 79 81 88 80 64 68 40 45 49 75 81 76 81 76 84 86 84 80 71 69 83 81 77 71 51 56 33 58 83 42 41 43 51 51 53 43 48 33 39 79 79 68 76 68 66 87 69 60 84 84 81 88 88 88 88 84 88 56 75 66 47 42 73 70 76 48 84 78 73 71 86 76 68 ' 81 86 73 85 73 68 71 70 70 88 75 53 59 58 54 76 37 36 46 41 75 76 82 61 48 54 73 50 67 80 88 67 85 75 54 66 77 68 28 39 43  Initialization Set Avg StdDev Min 66-176 16.727 55.514 15.888 45.868 20.294 49.763 14.615 53.343 15.324 73.735 8.0653 78 13.134 62.813 13.299 27.167 22.146 57.483 23.828 67.344 14.302 71.7 7.6891 65.531 13.078 64.667 16.883 69.375 15764 62.73 18.133 58.526 13.669 62.316 18.695 47 1.4142 76.974 8.1589 72.605 13.814 73 12.789 71.179 15.302 72.229 15458 80.921 10.607 64.361 16.701 56.514 18.625 71.8 12.776 71.622 16.892 57429 25.913 71.441 15.812 67.657 19.873 69444 15.522 70.769 14.141 62.147 17.921 67.313 16.772 63.306 18 448 81 459 7.4297 54.658 20.062 61.722 18.064 37.182 25.339 59.636 16.794 45.25 15.295 47.286 10.688  28 31 16 22 29 52 45 30 2 16 34 56 41 26 40 0 41 29 46 55 26 42 28 40 50 33 29 39 33 14 29 32 39 45 29 41 0 57 20 23 7 34 24 38  Overage (Final Upload) Main Order Quantity Supplying mean - 1 stdev Supplying mean - 2 stdev mean - 1 stdev mean-2stdev 7/6/98 7/13/98 7/20/98 7/27/98 7/6/98 7/13/98 7/20/98 7/27/98 49 33 (36) (36) (33) (26) (52) (52) (49) (42) 40 24 (37) (37) (44) (53) (53) (47) (60) (31) 26 5 (46) (53) (33) (38) (74) (59) (67) (54) 35 21 (33) (46) (45) (33) (47) (60) (59) (47) 38 23 (36) (39) (38) (54) (53) (46) (31) (51) 66 58 (13) (24) (23) (29) (16) (15) (21) (21) 65 52 (14) (23) (29) (27) (16) (36) (16) (29) 50 36 (24) (30) (38) (44) (28) (32) (14) (18) 5 -17 (62) (35) (40) (44) (84) (57) (66) (62) 34 10 (52) (41) (47) (42) (76) (65) (66) (71) 53 39 (34) (28) (23) (48) (42) (37) (45) (31) 64 56 (18) (22) (16) (20) (26) (30) (24) (28) 52 39 (36) (19) (49) (32) (30) (44) (17) (31) 48 31 (29) (33) (29) (23) (46) (50) (46) (40) 54 38 3 21 (13) 5 (5) (2) (21) (18) 45 26 (13) (38) 3 (22) (32) (57) (16) (3) 45 31 1 4 2 (20) (12) (8) (13) (10) 44 25 (26) 1 (45) (26) (28) (18) (7) (9) 46 44 13 7 11 5 (4) (2) (6) (4) 69 61 (19) (10) 1 (27) (10) (18) (18) (7) 59 45 (18) (23) (32) (17) (9) (7) (31) (21) 60 47 (28) (27) (41) (40) (22) (9) (13) 56 41 (26) (28) (28) (25) (43) (43) (40) (41) 57 41 (31) (31) (47) (47) (47) (47) (31) (31) 70 60 (18) (14) (18) (24) (28) (24) (28) (14) 48 31 (39) (27) (56) (25) (35) (44) (8) (18) 38 19 (39) (35) (58) (28) (23) (54) (4) (9) 59 46 (29) 11 (42) (24) (30) (2) (11) (17) 55 38 (30) (23) (29) (47) (40) (46) (35) (18) 32 6 (36) (39) (54) (44) (62) (80) (70) (65) 56 40 (12) (25) (28) (41) (30) (27) (46) (11) 48 28 (35) (25) (37) (25) (55) (45) (45) (57) 54 38 (23) (14) (39) (30) (33) (32) (17) (16) 57 42 (14) (31) (29) (46) (28) (33) (13) (18) 44 26 (15) (49) (27) (33) (32) (31) (14) (9) 51 34 (25) 14 (18) (20) (3) (42) (3) (1) 45 26 11 9 4 (10) (20) (8) (15) (D 74 67 (3) (2) (8) (10) (8) (9) (15) (D 35 15 (48) (13) (68) (46) (39) (33) (26) (19) 44 26 (33) (29) (23) (47) (24) (41) (6) (51) 12 -13 (52) (68) (76) (55) (77) (93) (101) (80) 43 26 (24) (42) (32) (41) (28) (59) (49) (11) 30 15 (36) (36) (38) (47) (51) (51) (53) (62) 37 26 15 9 4 (2) (6) (2) (13) (17)  Initialization Set Range: 10/13/97 - 6/29/99  Average Sum  (25) (1.108)  (21) (956)  (22) (993)  (18) (819)  (41) (1,824)  (37) (1.672)  (38) (1.709)  (34) (1.535)  Table 10 : Sample first policy results  36  Flights From YVR on Mondays  Flight 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56  Test Set Pax6 7/6/98 7/13/98 7/20/98 7/27/98 70 75 74 91 83 78 70 68 93 71 70 77 91 101 97 92 69 44 36 60 80 83 82 72 85 94 101 94 83 46 47 75 78 55 53 41 80 77 55 76 93 94 72 68 91 75 84 49 243 249 212 208 88 84 93 72 159 190 196 171 96 87 89 85 212 173 160 192 91 85 87 84 91 89 96 87 84 98 69 89 89 93 96 90 76 74 76 91 189 156 129 153 70 91 80 88 84 101 103 91 96 107 96 93 119 94 77 82 94 104 78 118 93 100 78 106 95 101 88 91 87 87 88 83 72 99 91 92 91 94 77 92 82 188 156 134 102 175 185 191 196 92 88 81 92 319 194 186 189 80 68 82 71 103 99 95 106 76 87 94 86 65 45 74 53 92 78 82 80 93 95 92 93 84 72 65 74 75 74 73 79 94 85 89 95 80 58 63 51 56 92 81 71 85 85 77 71 100 63 48 56 56 52 54 32 240 255 226 230 46 50 45 48 70 42 42 48 180 173 197  Init. Set Diff6 0 Mean StdDev -6.2857 4.8902 -3.8095 4.5893 0.6667 10.758 -5.3182 3.5908 1.5909 6.5441 -5.1905 3.7232 -11.857 10.155 0.5263 9.8959 -1.4545 4.5327 0.1579 6.9862 -1.0455 7.2274 -1.5455 6.724 7.1429 18.575 -0.3333 8.3506 -9.619 30.331 -11.455 16.919 -5.0476 12.048 -0.2857 8.2592 -4.3636 4.4137 -9.1818 9.8833 -8.55 8.3001 3.6 8.0092 8.6818 29.806 -4 9.3808 -5.0455 4.8352 -3.6364 11.253 -7.6364 6.6372 -1.7273 7.2518 -4.3636 4.2375 -22 20.474 -5.9524 6.3834 -2.7619 3.9485 -5.1111 7.3396 13.15 27.639 -0.0952 15.446 -46.182 32.408 -5.7368 7.8233 -2.3333 7.2134 1.6818 9.6973 -8.2857 7.0295 2.1905 11.197 3.35 5.7057 -6.2727 6.212 -2.6818 7.4219 1.4545 12.872 -6.5 10.071 -5.0476 4.5987 3.2 9.9398 -3.1429 5.6505 -4.8636 7.4405 0.8571 6.6429 -1.4286 4.0196 -14.4 13.801 -3.5238 7.6917 -2.25 8.9788 -13.9 10.182  Main Order* 7/6/98 7/13/98 7/20/98 7/27/98 59 64 63 80 75 70 62 60 83 61 60 67 82 92 88 83 64 39 31 55 71 74 73 63 63 72 79 72 74 37 38 66 72 49 47 35 73 70 48 69 85 86 64 60 83 67 76 41 232 238 201 197 79 75 84 63 119 150 156 131 68 59 61 57 195 156 143 175 82 76 78 75 82 80 87 78 65 79 50 70 72 76 79 73 72 70 72 87 168 135 108 132 57 81 70 78 74 86 88 76 81 93 82 79 105 85 68 73 85 95 69 109 84 58 36 64 53 89 76 79 75 81 80 76 65 87 79 80 79 80 63 78 68 172 140 86 118 96 106 112 117 78 74 67 78 309 184 176 179 72 60 74 63 88 84 80 91 67 78 85 77 43 63 72 51 80 66 70 68 83 85 82 83 73 61 54 63 58 57 56 62 84 75 79 85 73 51 56 44 83 47 72 62 73 73 65 59 94 57 42 50 51 47 49 27 212 227 198 202 35 39 34 37 59 31 31 37 156 149 173  Test Set PaxO Meal bank requirement" 7/6/98 7/13/98 7/20/98 7/27/98 7/6/98 7/13/98 7/20/98 7/27/98 66 70 72 77 7 6 9 -3 78 79 70 67 3 9 8 7 92 68 78 79 9 7 18 12 82 85 87 79 0 -7 -1 -4 67 45 32 67 3 6 1 12 78 81 71 77 7 7 4 8 72 81 81 70 9 9 2 -2 78 49 46 78 4 12 8 12 78 52 55 47 6 3 8 12 81 75 59 74 8 5 11 5 88 87 76 66 3 1 12 6 88 77 78 57 5 10 2 16 236 221 223 200 4 -17 3 22 85 79 87 75 6 4 3 12 166 178 172 179 47 28 16 48 77 79 86 73 9 20 25 16 103 189 -40 14 83 86 88 86 1 10 10 11 86 82 75 6 -5 -3 74 79 60 64 9 0 10 -6 74 77 76 70 2 1 -3 -3 77 78 88 86 5 8 16 -1 179 175 139 156 11 40 31 24 67 10 80 86 79 88 -1 16 1 14 86 98 86 88 0 10 10 7 108 100 102 100 15 18 23 -5 81 87 83 88 -4 19 10 3 99 71 108 79 4 2 -1 -5 89 61 81 79 31 25 17 26 87 83 81 84 -2 7 2 9 88 80 79 70 7 0 3 5 88 88 88 88 1 9 8 9 83 80 88 75 3 17 10 7 177 140 153 110 5 0 67 -8 171 164 178 196 75 58 66 79 86 78 86 77 8 4 19 -1 317 195 179 176 8 11 3 -3 77 70 74 76 5 10 0 13 96 103 103 95 8 19 23 4 64 80 88 69 -3 2 3 -8 49 59 84 44 6 -4 12 -7 88 71 73 85 8 5 3 17 88 84 88 3 2 5 82 76 88 21 22 25 75 74 62 18 6 12 88 85 80 13 6 -5 55 64 58 4 8 14 59 78 69 12 6 7 80 65 68 7 0 9 64 55 58 7 13 8 53 56 33 6 7 6 219 199 196 -8 1 -6 51 44 42 12 8 7 42 48 49 11 17 12 174 164 172 18 15 -1 4080 4911 4986 4929 * Pax6 + [Avg(Diff6_0) - Stdev(Diff6_0)] Average 8.3415 9.6296 10.127 8.0909 Sum of positive 352 556 607 516 Sum of negative -10 -36 -50 -71 Max 75 58 67 79 Min -4 -17 -40 -8 * If positive, values are final upload. If negative, values are meal overage  Table 11: Sample second policy results  37  Appendix C: Costing Analysis Database This appendix describes the structure and content o f the Costing Analysis Database. The appendix is divided into two sections. The first section describes data source issues. The second section explains how the database functions. This is done by describing every form, report and query in the database detailing its purpose and how data is filtered and calculated.  A. Data Source  Issues  Issue of full data set vs. test subset The final recommendations are mainly based on the comparison o f cost and quantities obtained from the ordering policy testing model against current cost figures obtained from Canadian Airlines data archives. Comparison is done on a monthly total basis. A n important aspect o f the analysis is to ensure that both the actual and model values are obtained from the same set o f flight/dates. A s it occurs, the source data used in the model is not complete. Holes in pre-departure data force the model to exclude certain flights/dates. A s a result the model output set is slightly smaller than the total actual set for a given month. W e therefore ensure a valid comparison by reducing the actual flight data set to be equal to the one used for the model. This reduced set o f flight w i l l be referred to as the test subset i n this document. Note that the test subset is approximately 97% o f the real set for the flights included in the final scenario analysis.  Current meal cost Canadian Airlines information systems keep track o f all caterer invoice figures. The information contains invoice details down to the item number and description. This information can be retrieved from a table that lists all items that were charged by the caterer for a given flight on a given date, along with the actual expensed amount, the corresponding accounting code, and the class o f service code. A query was designed using the accounting code as a filter to strictly obtain food item costs. In order to compare invoiced amounts to model-generated costs certain items should be screened out. These include special meal surcharges, deadhead equipment charges, L A P meal charges, and reduced tray charges. A surcharge is added to the invoice on top o f the regular price o f a meal i f it is changed to a special meal. The fee for deadhead equipment is introduced for the handling and washing o f the empty trays used to fill up the galleys. L A P fees are added i f a flight requires an upload from this program. A reduced tray charge is added i f flights, for which a meal service was prepared, are canceled on short notice. A l l the above-described charges would neither be captured nor affected by the model costing.  38  Note that the F C P V 1 0 7 t table dynamically grows as invoices are received. A small portion o f the real cost could be missing due to late invoices, rejected invoices, or retroactive price adjustments billed later.  Current Meal Quantity The number o f meals catered is obtained from invoice data just as the meal cost described above. In this case though, a different table containing summary invoice information is used instead o f the one containing information at the item level.  B. Database  Structure  The policy tool architecture consists o f layers o f tables, queries, reports, and forms. What the user sees is a form with text boxes, and action buttons. Changing the values o f parameters inside the form allows for testing different policies. To see the results o f a test policy, the user presses a button that prints a report containing performance measures. The reports extract information from queries and perform statistical calculations on them. There are multiple layers o f queries. The top layer assembles the appropriate information in one summary table. The intermediary layers create fields calculated from Terradata table values. The bottom query layer extracts necessary fields from the table layer. The base tables are directly imported from the Terradata information system.  Main Form (Forml) This form is the main menu for the whole costing database. It serves two functions: entering the parameters and data ranges for running a single test policy, and launching the reports for the given parameter set. Policy Parameters M i n i m u m trigger for uploading 14 meals (M) M i n i m u m trigger for uploading 7 meals (m) Under-catering Factor (Reduction Factor) Underage Cost per Passenger M a x i m u m Final Upload  Parameter used in the intermediary upload formula explained below. Parameter used in the intermediary upload formula explained below. Parameter used in the main order formula. This fraction is multiplied by the standard deviation o f the 3hr-todeparture difference distribution. Dollar value cost for each instance o f a passenger without a meal. I f set to 1, the result is simple underage meal quantity. M a x i m u m number o f meals that can be added to a flight at the final upload.  39  EMEU  SU Costing Analysis Database  ( . l l l . l l l i >u A i i i i l i e s  Intermediary Upload -  v \  Under-calering Factor  Minimum trigger for uploading 14 meals Minimum trigger (or uploading 7 meals  Underage Cost per Passenger Maximum Final Upload Data Range Start date for historical data  2/1/99  End date for historical data  3/31/99  Start date for test period  4/1/99  End date for test period  4/30/99  Current System Report  Actual Billed Report  From Flight To Flight  AMOS report  Meal Bank Report  Figure 13: Costing database user interface  Data Range Parameters  From Flight To Flight Start date for historical data Stop date for historical data Start date for test period Start date for test period  Lower bound of the range of flights to be tested. Coded by flt_num. Upper bound of the range of flights to be tested. Coded by fltnum. Lower bound of the date range for the data sample period (initialization set). Upper bound of the date range for the data sample period (initialization set). Lower bound of the date range for the data test set (holdout set). Lower bound of the date range for the data test set (holdout set).  Buttons  Current System Report Meal Bank Report  Prints the report containing the actual cost results based on meal quantity. Prints the report containing the model cost results for the specified ordering policy.  40  A M O S Report  Actual B i l l e d Report  Prints the report containing the A M O S model cost results using the specified delivery parameters (max upload, minforl4, minfor7). Prints the report containing the actual cost results based on invoice history.  Actual Billed Cost Billed cost Report  Presents data from b i l l e d c o s t query and calculates totals for each column. Billed cost Query  Extracts records from Detailed_bill_yvrJantoAug99 table that match records in HST_01_subset (test subset) on the basis o f date, flight number, departure station, arrival station, and class o f service code. This record matching step is necessary to ensure that both the actual and model values are obtained from the same set o f flight/dates.  Filters BRD FLT NR COS_CD CTGCD3 ITMNR BRDFLTDT  Inside flight number range from M a i n Form. Y class only Exclude code for Special Meals Exclude codes for Reduced Trays, Deadhead charge, and L A P meal delivery charges. Inside data test set  Current Cost (Actual Meal Quantity) Cost real Report  Presents data from cost_real query and calculates totals for each column. Cost real Query  Joins records from c o s t r e a l m e a l s query and cost_real_underage query with matching flight numbers.  Filters FLT N R  Inside flight number range from M a i n Form.  Formulas UnderageCost_rev  IIf([MealCost]=0,0,[SumOfUnderageCost_rev]) If a flight has no meal service (cost) default revenue passenger underage to zero.  41  UnderageCost_all  IIf([MealCost]=0,0,[SumOfUnderageCost_all]) If a flight has no meal service (cost) default total underage to zero.  Cost real meals Query  This query multiplies the number o f meals catered by the unit price o f each item scheduled for any flight. •  M e a l quantity taken from H S T _ O l _ o v e r _ m l _ l v C H l S T query.  •  Scheduled meal service items taken from f l i g h t _ i t e m s _ Y V R _ Y _ J a n _ J u n 9 9  •  M e a l service item prices taken from i t e m _ p r i c e s _ Y V R _ u p d a t e d 2  Filters ITMPRCEFFDT ITM_PRC_DIS_DT CTG_CD_3 dptstacd FLT DT COS_CD  Price effective date must be before lower bound o f data test set Price discontinued date must be after upper bound o f data test set Only extract records with codes for food items Y V R only Inside data test set Y class only  Formulas There are two different pricing methods: one using a provisioning ratio, and one using a set quantity per passenger rule. The provisioning ratio simply indicates what percent fraction o f the total load should be provided with the item, e.g. 60% chicken casserole and 40% beef. The set quantity figure indicates to supply one item per X passengers, e.g. supply 1 B B Q pork bun per 9 passengers. Note that the great majority o f meal items are supplied based on a ratio function. Multiplying the meals catered quantity by a ratio yields fractional meal quantities. The caterer's ordering system rounds up these ratio quantities before multiplying them by their unit price. This is replicated in the costing formula by using the standard CintO function. MealCost  OverCost  Sum(IIf([PVN R A T C D ] = " P " , [ I T M P R C A M T ] * C I n t ( [ P V N R A T Q T Y ]/100*[CTR P S G Q T Y ] ) , [ I T M P R C A M T ] * C i n t ( [ C T R P S G Q T Y ] / [ P V N_RAT_QTY]))) Calculation o f cost based on meal quantity ( C T R P S G Q T Y ) from H S T _ 0 1 _ o v e r _ m l _ I V C H I S T query. Sum(Hf([PVN R A T C D ] = " P " , [ I T M P R C A M T ] * C I n t ( [ P V N R A T Q T Y ]/T00*[over m l qty]),[ITM P R C AMT]*CInt([over m l q t y ] / [ P V N R A T QTY]))) Calculation o f overage cost based on over-catered meal quantity ( o v e r m l q t y ) from H S T O l o v e r m l J V C H I S T query.  42  HST 01 over ml IVCHIST Query  Extracts records from l n v o i c e _ h i s t _ Y V R _ J a n t o J u n 9 9 that match records in H S T _ O l _ s u b s e t (test subset) on the basis of date, flight number, departure station, arrival station, and class of service code. This record matching step is necessary to ensure that both the actual and model values are obtained from the same set of flight/dates. Filters  BRD FLT NR COS CD CTG_CD_3 ITM_NR BRDFLTDT CTRIVCOPCD  Inside flight number range from Main Form. Y class only Exclude code for Special Meals Exclude codes for Reduced Trays, Deadhead charge, and LAP meal delivery charges. Inside data test set from Main Form Only use accounting code 00 (Routine scheduled service)  Formulas  over ml qty  IIf([CTR_PSG_QTY]-[ob_rev_psg_qty]-[ob_nrv_psg_qty]>0, [CTR_P S G Q T Y] - [ob_rev_psg_qty] - [ob_nrv_psg_qty] ,0) Number of meals over-catered is the difference between meals catered minus total on-board passengers. If negative, default to zero.  Cost real underage Query  Calculates totals grouped by flight number on fields in c o s t _ r e a l _ u n d e r a g e _ a l l . Filters  Dpt_sta_cd FLTDT  YVR only Inside data test set from Main Form  Cost real underage all Query  Extracts records from H S T _ 0 1 _ s u b s e t in order to calculate underage figures. Filters  COS CD  Y class only  Formulas  Underage_nrv_incl  [ob_rev_psg_qty]+[ob_nrv_psg_qty]-[act_ctr_ml_qty] Total number of meals under catered (both revenue and nonrevenue passengers)  UnderageCost_all IIf([Underage_nrv_incl]>0,[Underage_nrv_incl] * [Forms]! [Form 1 ] .[undercost],0) Total under-catering cost (revenue &non-revevue) based on  43  UnderageCostrev  Underage Cost per Passenger in Main Form. IIf([Underage]>0,[Underage] * [Forms]! [Form 1 ]. [undercost],0) Revenue passengers under-catering cost based on Underage Cost per Passenger in Main Form.  Meal Bank Cost (Model) Cost Report  Presents data from c o s t query and calculates totals for each column. Cost Query  Joins records from c o s t _ m e a l s query and c o s t _ u n d e r query with matching flight number, departure station, and arrival station. Filter FLT NR  Exclude flight 8  Cost meals Query  This query multiplies the number of meals catered by the unit price of each item scheduled for any flight. • Meal quantity taken from Ext_MBDemand query. • Scheduled meal service items taken from f l i g h t _ i t e m s _ Y V R _ Y _ J a n _ J u n 9 9 • Meal service item prices taken from i t e m _ p r i c e s _ Y V R _ u p d a t e d 2 Filters Same as Cost_real_meals Query Formulas MealCost  Sum(IIf([PVN R A T CD]="P",[ITM P R C AMT]*CInt([PVN R A T Q T Y ]/100*[MealsCatered]),[ITM P R C AMT]*CInt([MealsCatered]/[PVN R A T_QTY]))) Calculation of cost based on meal quantity (MealsCatered) from Ext_MBDemand query.  OverCost  Sum(IIf([PVN_RAT CD]="P",[ITM P R C AMT]*CInt([PVN R A T Q T Y ]/100*[over ml qty]),[ITM P R C AMT]*CInt([over ml qty]/[PVN R A T QTY]))) Calculation of overage cost based on over-catered meal quantity (over_ml_qty) from Ext_MBDemand query.  Cost under Query  Calculates underage cost from Ext_MBDemand query. 44  Filters FLT DT Formulas UnderCost  UnderCostnrv  Inside data test set from Main Form  Sum([under_ml_qty]* [Forms]! [Forml].[undercost]) Multiply the number of under-catered meals for revenue passengers only by the Underage Cost per Passenger in Main Form. Sum([under_ml_qty2] * [Forms]! [Form 1 ]. [undercost]) Multiply the number of under-catered meals for all passengers by the Underage Cost per Passenger in Main Form.  Ext MBDemand Query  Two options: 1) Extract records from MBDemand 2) Extract records from MBDemandnonrv In both cases this query calculates the amount of meals catered for each flight, the total passenger quantity, the number of meals under-catered (total), the number of meals under-catered (revenue only), and the final upload quantity. Formulas MealsCatered Paxqty  Under_ml_qty  Under_ml_qty2  Final_Upload  [lhrUpload]+[Final_Upload]+[MainOrder] The total number of meals catered is the sum of all three deliveries. [ob_rev_psg_qty]+[ob_nrv_psg_qty] Total passenger quantity is the sum of on-board revenue and nonrevenue passengers. IIf([ob_rev_psg_qty]-[MealsCatered]>0,[ob_rev_psg_qty][MealsCatered],0) The number of under-catered meals for revenue passengers only. If negative, default to zero. IIf([Pax_qty]-[MealsCatered]>0,[Pax_qty]-[MealsCatered],0) The number of under-catered meals for all passengers. If negative, default to zero. IIf([FinalUpload]>0,IIf([FinalUpload]<Int([Forms]![Forml].[maxupl oad]), [FinalUpload], [Forms]! [Form 1 ]. [maxupload]),0) This constrains the final upload quantity from MBDemand query to the maximum upload set in the Main Form.  MDDemand Query  This query models the ordering policy. It calculates the order quantities for the main, intermediary, and final uploads. These decisions are made on descriptive statistics of  45  inter-period passenger counts, taken from YVRf l i g h t s M e a n D i f f e r e n c e s query, and current multi-period passenger counts, taken from D i f f e r e n c e T a b l e . Filters FLTJDT FLTNR Dpt sta cd COS_CD 4 2 0 Formulas MainOrder  lhrUpload  lhrUpload_pr e  FinalUpload Overcatering  Inside data test set from M a i n Form Inside flight number range from M a i n Form Y V R only Y class only Not null. Eliminate a record i f data for this query period is missing. Not null. Eliminate a record i f data for this query period is missing. Not null. Eliminate a record i f data for this query period is missing.  IIf(CInt([DifferenceTable3]. [4]+[YVRflightsMeanDifferences] . [ A v g O f Diff4_0][Forms]![Forml].[Fraction]*[StDevOfDiff4 0])<=[avl st qty],CInt([D ifferenceTable3].[4]+[YVRflightsMeanDifferences].[AvgOfDiff4_0][Forms]! [Form 1 ]. [Fraction] * [StDevOfDiff4_0]), [ a v l s t q t y ] ) M e a l order = booked load at 3-hour pre-departure + average difference between 3-hour and final - reduction factor * standard deviation from average difference distribution. A n If statement prevents the main order from being higher than the aircraft seat capacity IIf( [ 1 hrUpload_pre]+[MainOrder]<=[avl_st_qty], [ 1 hrUpload_pre], [avl _st_qty] - [MainOrder]) This If statement prevents the sum o f the main order and intermediary upload from exceeding the aircraft seat capacity IIf([DifferenceTable3].[2]+[YVRflightsMeanDifferences].[AvgOfDiff 2_0]-[MainOrder]>=[Forms]! [Forml ]. [Minfor 14], 14, IIf([DifferenceTable3].[2]+[YVRflightsMeanDifferences].[AvgOfDiff 2_0]-[MainOrder]>=[Forms]! [Forml].[Minfor7],7,0)) Let Expected Final Upload be: booked load at 1-hour pre-departure + average difference between 1 -hour and final - MainOrder . If Expected Final Upload >= minimum for 14, then order 14 meals. If Expected Final Upload between minimum for 14 and minimum for 7, then order 7 meals. If Expected Final Upload < minimum for 7, then make no order at this point. [ob_rev_p sg_qty]+[ob_nrv_psg_qty ] - [MainOrder] - [ 1 hrUplo ad] Final upload= final passenger count - meals already boarded. Iif(([MainOrder]+[lhrUpload]+IIf([FinalUpload]>0,[FinalUpload],0)[DifferenceTable3].[0])>0,[MainOrder]+[lhrUpload]+IIf([FinalUpload ]>0,[FinalUpload],0)-[DifferenceTable3].[0],0) Let total catered meals be: main order + intermediary upload + final upload (zero i f negative). Number o f meals over catered = total catered meals - final passenger  46  Visit  count. If negative, default to zero. IIf([FinalUpload]>0,l,0) Boolean value for the necessity o f a delivery van visit at the final point (true i f final upload > 0).  MBDemand no nrv Query  Same as M B D e m a n d except that no meals are uploaded for non-revenue passengers. Formulas FinalUpload  [ob_rev_p sg_qty] - [MainOrder] - [ 1 hrUplo ad] Final upload = final revenue passenger count - meals already boarded.  AMOS System Cost Cost AMOS Report  Presents data from c o s t A M O S a l l query and calculates totals for each column. Cost AMOS all Query  Joins records from cost_AMOS_quantities and cost_AMOS_meals query with matching flight number, date, departure station, and arrival station combinations. Filters FLT N R Inside flight number range from M a i n Form. Formulas Underage_rev  Underage_all  Sum([cost_AMOS_quantities]. [underage_rev] * [Forms]! [Form 1 ]. [u ndercost]) M u l t i p l y the number o f under-catered meals for revenue passengers only by the Underage Cost per Passenger in M a i n Form. Sum([cost_AMOS_quantities]. [underage_all] * [Forms]! [Form 1 ]. [un dercost]) M u l t i p l y the number o f under-catered meals for all passengers by the Underage Cost per Passenger in M a i n Form.  Cost AMOS quantities Query  This query uses the historical data (HST_01_subset) and A M O S forecasts (cost_AMOS_forecast) to calculate the final upload size, the overage quantity, and the underage for revenue passengers only and for all passengers. Formulas Upload_raw  [ob_nrv_p sg_qty]+[objre v_p s g q t y ] - [drv_fcs_psg_qty ]  47  Upload  total_meals overage  underageall  underage_rev  -  Final upload = final passenger count (rev+nrv) - meals already boarded ( A M O S forecast). IIf([Upload_raw] Between 0 A n d [Forms]! [Forml].[maxupload],[Upload_raw],IIf([Upload_raw]<0,0 , [Forms]! [Form 1 ]. [maxupload])) This constrains the final upload quantity from MBDemand query to the maximum upload set in the M a i n Form. If negative, default to zero. [drv_fcs_psg_qty]+[Upload] The total final number o f meals catered. IIf(( [totalmeals] - [ob_rev_psg_qty ] [ob_nrv_psg_qty])>0, [totalmeals] - [ob_rev_psg_qty][ob_nrv_psg_qty] ,0) Number o f meals over catered = total catered meals - final passenger count. If negative, default to zero. IIf(([ob_rev_psg_qty]+[ob_nrv_psg_qty][total_meals])>0,[ob_rev_psg_qty]+[ob_nrv_psg_qty][total_meals],0) The number o f under-catered meals for all passengers = final passenger count - total catered meals. If negative, default to zero. Ilf(([ob_rev_psg_qty]-[total_meals])>0,[ob_rev_psg_qty][total_meals],0) The number o f under-catered meals for revenue passengers only. If negative, default to zero.  Cost AMOS forecast Query  Extracts the 3-hour pre-departure forecasts for flights within the data test range from the pre-departure information table ( A M O S _ 0 1 ) .  Filters FLTJDT dpt_sta_cd els qry_pd cd COS_CD  Inside data test set from M a i n Form Y V R only = 4 (the 3-hour query period) Y class only  Cost AMOS meals Query  This query multiplies the number o f meals catered by the unit price o f each item scheduled for any flight. •  M e a l quantity taken from C o s t _ A M O S _ q u a n t i t i e s query.  •  Scheduled meal service items taken from f l i g h t _ i t e m s _ Y V R _ Y _ J a n _ J u n 9 9  48  •  Meal service item prices taken from i t e m _ p r i c e s _ Y V R _ u p d a t e d 2  Filters Same as Cost_real_meals Query  Formulas MealCost  OverCost  Sum(IIf([PVN R A T CD]="P",[ITM P R C AMT]*CInt([PVN R A T Q T Y ]/100*[total meals]),[ITM P R C AMT]*CInt([total meals]/[PVN R A T Q TY]))) Calculation of cost based on meal quantity (total_meals) from C o s t _ A M O S _ q u a n t i t i e s query. Sum(IIf([PVN_RAT CD]="P",[ITM PRC AMT]*CInt([PVN R A T Q T Y ]/100*[overage]),[ITM_PRC_AMT]*CInt([overage]/[PVN_RAT_QTY]))) Calculation of overage cost based on over-catered meal quantity (overage) from C o s t _ A M O S _ q u a n t i t i e s query.  C. Source Data Tables  The following table describes source data tables referenced by the queries detailed above. Database T a b l e  Description  Station  Date R a n g e  HST01 PD_1998 AMOS_01  Final loads Pre-departure data Pre-departure data (costing tables) Flight scheduled items Item prices Invoice (detailed to item level) Invoice history  YVR All YVR  Nov98 - Jun99 Feb98 - Jan99 Sep98 - Jun99  YVR YVR YVR YVR  Jan99 - Jun99  Flight_items Item_list Detailed_bill_yvrJantoAug99 Invoice_hist_YVR_JantoJun99  Jan99 - Aug99 Jan99 - Aug99  Table 12: Source data tables  49  Appendix D: Meal Bank Wastage Cost The following is an example of the method used to estimate meal bank wastage for the reduced scenario. Trans Continental Flights - Lunch/Dinner  Date 2/1/99 2/2/99 2/3/99 2/4/99 2/5/99 2/6/99 2/7/99 2/8/99 2/9/99 2/10/99 2/11/99 2/12/99 2/13/99 2/14/99 2/15/99  Daily Demand 42 52 46 19 22 27 49 17 48 29 29 39 12 35 13  Target Daily Inventory Production 77 64 77 23 77 54 77 46 77 31 77 46 77 31 77 49 77 28 77 49 77 29 77 48 77 39 77 38 77 39  Unused Meals 54 25 31 58 55 50 28 60 29 48 48 38 65 42 64 '  Wasted Meals 0 2 0 12 24 4 0 11 1 0 19 0 26 4 25  End of Day Inventory 54 23 31 46 31 46 28 49 28 48 29 38 39 38 39  14 0 7 0 0 0 0 10 0 0 0  32 44 33 39 36 8 46 31 31 25 20  (break in table for summarization) 6/19/99 6/20/99 6/21/99 6/22/99 6/23/99 6/24/99 6/25/99 6/26/99 6/27/99 6/28/99 6/30/99 Mean Demand Maximum Std Dev Inventory Level  31 33 37 38 41 69 31 36 46 52 57 32.3 91 14.9 77  77 77 77 77 77 77 77 77 77 77 77  32 45 33 44 38 41 69 31 46 46 52  46 44 40 39 36 8 46 41 31 25 20  Mean Wastage 9.2 Total Wastage 1358 Unit Meal Cost $ 5.00 Total Wastage Cost $ 6,790.00 Monthly Wastage Cost| $ 1,358.00  Table 13: Meal bank wastage cost estimation  50  Flights in the meal bank program are pooled into three categories based on their meal service: Transcontinental breakfast, Transcontinental lunch/dinner, and International meals (double service). Note that flights CP900 and CP916 are isolated from the Transcontinental breakfast service group since they use different tray sizes. Daily demand for both intermediary and final uploads from the model output is summed up across all flights i n each pooled category. The daily target inventory level is calculated using the mean plus three standard deviations o f the daily demand distribution. Daily production is the difference between the target inventory level and the previous day's final inventory. The unused part o f the day's inventory is then divided between wasted (unused meals from the previous day's production) and the end o f day inventory (unused meals from the current day's production). The wastage figure is calculated from the sum o f meals wasted and the unit meal cost. According to the current agreement Canadian Airlines' share would be half o f the cost.  TransCon - Breakfast CP900 (Breakfast) CP916 (Breakfast) TransCon - Lunch/Dinner International Total Monthly Cost CAI Share  Target Inventory Mean Daily Wastage 52 11.5 18 5.3 16 3.0 77 9.2 46 10.2  Monthly Cost 1,124.90 493.50 120.40 1,358.00 3,479.90 6,576.70 3,288.35  $ $ $ $ $ $ $  T a b l e 14: M e a l b a n k wastage cost s u m m a r y  51  Appendix E: Delivery Van Driver Schedules V a n driver schedules were built for evaluating man-hour requirements for each scenario. The costs are calculated using an estimated hourly wage o f $25 for drivers. The added cost shown below is obtained by subtracting the current wages from the scenario total cost. In the final costing results, the current costs were not subtracted since we assumed that existing delivery van service would be kept as is. Current Driver  Van 1 2  Start 1 2  End 7:30 11:45  Total  19:20 14:45 Daily Weekly Yearly  Man-Hours Salary 12 3 15 $ 375.00 105 $ 2,625.00 5460 $136,500.00  Reduced System Implementation  Driver  Van 1 2 3 3 4  Start 1 2 3 3 4  End 7:05 7:05 7:00 11:10 11:15  16:00 19:20 8:00 14:55 12:15  Total  Daily Weekly Yearly Added Cost (annual)  Man-Hours Salary 9 12.5 1 4 1 27.5 $ 687.50 192.5 $ 4,812.50 10010 $250,250.00 $ 113,750.00  Full System Implementation  Driver  Van 1 2 3 4 5 6  Start 1 1 2 3 4 5  End 6:20 17:10 7:20 7:55 11:45 11:15  Total  16:00 22:45 18:30 17:55 14:10 14:05 Daily Weekly Yearly  Added Cost (annual)  Man-Hours Salary 10 6 11.5 10.5 3 3.5 44.5 $ 1,112.50 311.5 $ 7,787.50 16198 $ 34,950.00 $ 268,450.00  Table 15: Van driver schedules  52  A. Scheduling Methodology Above results are based on the summer 1999 schedule for Vancouver International Airport flights. A base van schedule was designed for each scenario using June 2, 1999 scheduled departure times.  Full System Implementation • • • •  Visit all flights with a meal service A l l o w minimum 5 minutes for transit time between one flight to the next Intermediary upload visits for wide-body aircraft only Visit time: 10 minutes for narrow-body, 15 minutes for wide-body, and 10 minutes for intermediary uploads  Reduced System Implementation • • • • •  Include international (1-99) and transcontinental long-sector (900-999) flights only (excluding late flights) A l l o w minimum 5 minutes for transit time between one flight to the next Intermediary upload visits for all included flights Visit time: 15 minutes for final uploads, and 10 minutes for intermediary uploads Other flights are serviced according to current procedure  A reduced implementation scenario schedule was designed for each month i n the analysis (February to June 1999). The schedules were also tested with actual departure times on three different sample weeks in A p r i l , M a y , and June. In all cases, the planned schedule could be adjusted within the set fleet and driver shift limitations from the base schedule. The van driver schedule for June 1999 is shown on the next pages. Coloured flights are those included in the meal bank system. Other flights are either visited as part o f the current L A P program (two blocks, no colour) or visited on demand for special missing equipment requests (one block, no colour). The vans are assigned to flights that are part o f the same meal type pool as best as possible. This way, meals can be loaded on any flight and less replenishment is required. Van  Meal Type  1  Breakfast Lunch/dinner International Breakfast Lunch/dinner Breakfast Lunch/dinner International International  2 3  4  Flights 902, 984 988, 992, 994 9, 29, 3 916, 986 910, 990, 912, 982 900 920 29, 17  1  Table 16: F l i g h t assignments  53  CO  CD  CD CO  ro ro ro o  < < < <  N  o  cr o in T > X o  a C; CO Ik  _, :•.  cn  £  CD a co ro ro CO  o O o CD CO co ro cn O cn o o o o < X < < m < Z < O O O  a  a  6  Cn •o cr CO Cn £ .•  c ro U  cc  c'  CD CD Cn a ro ro • -  (.:"•  Cn ro ro CO CO Cn cn 8 o  g  co CD cn Cn Jk. u •U co cn -.; Cn O c O o O c o cn cn Ui C" O o (? o 5 -< < < -< < Z o < -C < < P* O CO O rjj •< > > Tl < •< -< C m X m XJ X O o N Q O D Co O  i  1  i  a  O  cn Cn CD ro CD CO CO o CO it  CD  IO ro ro ro ro ro o cn CO Cn o o o cn Cn Cn cn o cn O O O cn o o TJ < Z X -< < < -< O < < a TJ < X < m m XJ C m o -< n D O a  CD  CD  Cn CO fo  CD CD  81Qo  o  —i O  .':  co o  <  3  O) Cn CD  SB  CD  4-  s•c. 8  CO CD ro Cn cn CD CO Cn *• a ro o O -C  co Co CO Co Co a' o O O o o  a  > X  •<  <  O C  s  1  <  s2 o  < X O a  K Flight #  ^1 S Scheduled cn Departure  < < -< X < -< > < o a X  Destination 4:30 4:3b 4:40 4:45 1:50 4:55 5:00 505 5:10 5:15 5:20 5:25 5:30 5:35 5:40 5:45 5:50 5:55 6:00 6:05 6:10 6:15 6:20 6:25 6:30 6:35 6:40 6 45 6:50 5:55 7:00  w  7:05  w  7:10 7:15 7:20 7:25  „  7:30  ho  7 35 7:40  CJ IO  to  7:45  u  7:50 7:55  -I  8 00 8:05 3:10 8:15 8:20 B:25  1  8:30 B:35 8:40 8:45 8:50 8:55 9'00 9:05  M H U  9:10 9:15 9:20 9:25 9:30 9:35 3:40 9:45 9:50 3:55 10:00 10:05 10:10 10:15 10:20 10:25 10:30 10:35 10:40 10:45 10:50  54  3  -' ..  ;• D I •0 -. .7: O) C 3 C DW O 7, DUifc.' C O CT C O C D O S O1 O C D C O C O .: O fc. | C Je -tOC: C OI > > > i 1 T|  2  ; j> t.  OB co X N)  2245  •o o 3 ~ • o cn -< •< •< in -< •< X < -< -< -n m < < o O X O o z O X O  .c 0 u  o Cn o < m <  •'  •  • O < < < m o O  cn ro no o o O  D  D  ro  •  •J  1  cn .••  Z o 33 -H  -fc; cn o -< < < -< < c •< Kl x o  •fc=•  Oi •• . < < CO 0 < D Ci D ./•  c  3  cn  co  D  o  no no no •g Co no o o • . cn < z X m < Q rn o X a O  o> co cc ; tB  -.  J> no  ro  C 85 -t-  :  light #  OC O§ Cn O ( C DC OC OO O < -<~- :• < < .< i < < > i no r\ o c  X < o  O  1  „  D I3C O  7'• 0•.t-; C M  cn -J no ct  _• o  ?  o ... Co O o O -<  cn  §  -  m O  O  co W  3  1  3 c  :.  <  o  o  Scheduled -g -J :•• departure cn cn < < O -c 33 X < O Z) X < Destination 0: 55 1: 30  ISO  1: 35 1 I 10 11 15 11 20 11 25  1  1: 30  fl  1. 35 11 40  *>  CO w  •  *•  n- 1 J3  1 145 1 150  CO  1 155  w  12 00  to  12 05 12 10 12 15  CO  12 20  to M  1? 25 12 30  -*  12 35  1  12 40 12 45 12 50 12 55 13 00 13 05 13 10  i  13 15 13 20 13 25  co  I-i  M  13 30  M  13 35  Co  13 40  CO  13 45 13 50  CO  13 65 14 00 14 05 14 10 14 15 14 20  _l  14 25 14 30  to  1  14 35  CO  14 40 14 45 14 50 14 55 15 00  *  15 05 15 10 i : 15  Ii  20  15 25 H  30  II 35 l b 40  i  15 45  L  15 50 15 55 IE :00 1( :05 11 :10 ' i :15 11 :20 If :25 If :30 11 :35 It 4 0 H :45  II  :50  11 :55 1 :00 1 :05 1 :10 i  55  :15  |YYZ  I •i C -o D C C DD t cnOiCT!s cnC1/ roon 3 O 7.Co 3D r s Dfc. XI C T >IfiC c£ C O0 D c s fc. O8fc.i C C OrD t o.D8 X -si noroXiD D C OC O O -J O -o ~o no V O OO O nD J -i cnS cn fefe*. C OOC C O C C O08 C iiC O O -< < o C O< < S : TITJUJ TJC m C > DC3 •< S 1 Tt> sn I > C - TJ > X C -1 i X TJ i N I X X X •  o  t. o -< -< O  o 5  O  a; E>  .0  -•• a < -< -< Z o  V)  •••  cn J-I  • o :: :• o -< < < < < < • m O O  •  •  •  <  -J  '•• • -.. fco  .  -fc. Ik  o o •< < < z a •< TJ O  -< < O  co JD  0  ro o  n ;  <  < c  o  O O CT)  m  cn o  •••  < < o  :  0  •••  Co  •o  o ro no o o o no cn • o n cn Z X -i < < < < •g O m Q O Q O <  o  O  a  • o  m O  Co  -  o a  Co o -< < -< ni o  o  VI  co co o o o 5 -< <  •< O o  -n ":  "light #  ;  cn  -fc-  < < -< X < -< < o o  Scheduled  Departure  Destination 7:20 7:25 7:30 7:35 7:40 7:45  O f ro  7:50 7:55 8:00  ro  18:05 18:10 18:15 18:20 18:25 18:30  ro  18:35 18:40 18:45 18:50 18:55 19:00 19:05 19:10  M  19:15 19:20 19:25 19:30 19:35 19:40 19:45 19:50 19:55 20:00 20:05 20:10 20:15 20:20 20:25 20:30 20:35 20:40 ?0:45 20:50 20:55 21:00 21:05 21:10 21:15 21:20 21:25 21:30 21:35 21:40 21:45 21:50 21:55 22:00 22:05 22:10 22:15 22:20 22:25  1  22:30  1  56  Appendix F: Costing Analysis Results Detailed results for each test month are presented below. The first table shows total monthly cost savings per flight. The second table contains the shortage totals per flight. Month  Flight  Savings  2 2 2 2 2 2 2 2 2 2 2 2 2 2 2  3 7 9 17 29 900 902 910 912 984 986 988 990 992 994 Flight  3 3 3 3 3 3 3 3 3 3 3 3 3 3 3  3 $ (605.86) 7 $ (2,236.37) 9 $ 111.53 17 $ 423.16 29 $ 3,317.37 900 $ 634.22 902 $ 670.68 1,307.32 910 $ 912 $ 1,109.30 984 $ 676.80 1,268.84 986 $ 988 $ 1,514.96 990 $ 945.72 992 $ 1,281.61 994 $ 1,521.87 Flight Savings  Month  Month 4 4 4 4 4 4 4 4 4 4  3 7 9 17 29 900 902 910 912 984  $ $ $ $ $ $ $ $ $ $ $ $ $ $ $  2,650.13 2,810.58 6,495.38 2,656.85 3,500.87 412.46 633.90 1,050.20 794.64 1,010.63 584.86 1,906.80 2,638.72 1,150.66 1,393.82 Savings  $ 10,848.38 $ 17,566.41 $ 7,961.91 $ 24,870.20 $ 8,966.73 332.59 $ 468.39 $ 40.95 $ 1,207.48 $ 516.52 $  57  4 4 4 4 4  986 988 990 992 994 Flight  $ $ $ $ $  622.42 363.27 643.39 1,125.82 872.38 Savings  5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5  3 7 9 17 29 900 902 910 912 916 920 984 986 988 990 992 994 Flight  $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $  2,155.63 3,513.43 416.81 4,366.92 2,069.22 619.05 168.52 620.42 1,107.92 297.17 1,809.29 257.84 430.61 1,295.97 297.23 1,207.86 1,198.20 Savings  6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6  3 7 9 17 29 900 902 910 912 916 920 982 984 986 988 990 992 994  $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $  1,622.96 1,494.04 1,080.55 1,567.16 2,688.41 222.58 346.30 479.86 1,278.88 480.52 683.88 1,385.87 588.38 543.37 524.90 368.55 587.51 1,106.28  Month  Month  Table 17: Detailed costing results  Month  Flight  2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 Month  3 7 9 17 29 900 902 910 912 984 986 988 990 992 994 Flight  3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 Month 4 4 4 4 4 4 4 4 4  3 7 9 17 29 900 902 910 912 984 986 988 990 992 994 Flight 3 7 9 17 29 900 902 910 912  Current Underage Revenue All Passengers Passengers  Meal Bank Underage Revenue All Passengers Passengers  0 0 1 4 0 0 0 0 0 0 0 0 0 0 0 0 7 0 0 0 0 9 0 0 0 11 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 2 0 0 0 0 0 0 0 0 0 5 10 29 0 0 0 0 Current Underage Meal Bank Underage Revenue All Revenue All Passengers Passengers Passengers Passengers 1 6 0 0 0 20 0 0 2 4 0 0 4 18 0 0 0 7 0 0 0 4 0 0 0 2 0 0 0 0 0 0 0 1 0 0 24 8 0 0 12 0 0 0 0 6 0 0 0 23 0 0 1 8 0 0 0 5 0 0 Current Underage Meal Bank Underage All Revenue Revenue All Passengers Passengers Passengers Passengers 6 7 3 23 0 0 0 0 0  8 54 10 28 1 2 4 0 0  9 0 0 0 0 6 0 0 0  10 36 0 0 0 17 0 0 0  4 4 4 4 4 4 Month  984 986 988 990 992 994 Flight  5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 Month 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6  3 7 9 17 29 900 902 910 912 916 920 984 986 988 990 992 994 Flight 3 7 9 17 29 900 902 910 912 916 920 982 984 986 988 990 992 994  0 16 0 0 0 16 0 0 0 13 0 0 0 13 0 0 0 12 0 0 4 17 18 36 Current Underage Meal Bank Underage All Revenue Revenue All Passengers Passengers Passengers Passengers 3 3 3 5 0 5 14 0 8 13 0 0 13 18 0 0 11 16 0 5 5 11 0 0 6 19 0 0 0 3 0 0 0 2 0 0 0 3 0 0 5 9 0 0 0 5 0 1 0 5 0 0 0 30 3 28 0 45 0 . 10 0 1 0 0 15 24 0 0 Current Underage Meal Bank Underage All Revenue Revenue All Passengers Passengers Passengers Passengers 1 0 15 0 7 1 0 3 0 0 0 2 0 0 0 0 0 4  4 0 16 0 7 9 2 18 0 0 1 4 8 8 0 7 8 12  Table 18: Detailed service level results  0 0 6 4 0 0 0 0 0 0 0 0 0 0 0 0 3 6  0 0 6 4 0 0 0 0 0 0 0 2 38 24 0 0 16 12  Outlier events were excluded from the results. These events are shown in Table 19 below. Outliers were extracted from both meal bank and current cost and underage totals. We observe that for most of these flights, no shortage occurred in fact. We can conclude that, although the booked load changed dramatically in the last 3 hours preceeding the flight's departure, the flight controllers were aware of the situation and adjusted the meal order appropriately. The method of identification for these events is explained below. Current Underage Flight 988 992 984 984 984 984 986 992 992 990 990 994 994 984  Date 2/3/99 2/6/99 2/11/99 3/3/99 3/25/99 4/8/99 4/8/99 4/9/99 4/17/99 4/19/99 4/25/99 6/22/99 6/24/99 6/12/99  Departure YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR  Arrival YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ  Meal Bank Underage  All Revenue Revenue All Passengers Passengers Passengers Passengers 0 0 25 38 0 32 0 33 0 0 9 28 0 0 29 53 0 9 24 41 0 26 0 23 0 0 21 23 0 0 20 48 0 0 42 25 0 22 0 29 0 55 22 0 19 25 27 33 0 0 20 30 170 170 0 0  Table 19: Outlier events  A. Identification of Outliers, an Example  To identify outliers, a distribution of underages from the meal bank output was compiled. Events in the top quartile were identified as outliers or irregular events. The table below shows an example of a flight affected by an irregular event. The data is taken from the meal bank model output for flight CP988 in the test period of February 1999. This flight is never under-catered except for one day where a large shortage occurs.  61  Date 2/1/99 2/2/99 2/3/99 2/4/99 2/5/99 2/6/99 2/7/99 2/8/99 2/9/99 2/10/99 2/11/99 2/12/99 2/13/99 2/14/99 2/15/99 2/16/99 2/17/99 2/18/99 2/19/99 2/20/99 2/21/99 2/22/99 2/23/99 2/24/99 2/25/99 2/26/99  flt_nr 988 988 988 988 988 988 988 988 988 988 988 988 988 988 988 988 988 988 988 988 988 988 988 988 988 988  dpt_sta YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR YVR  arr sta YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ YYZ  2/28/99  988  YVR  YYZ  MealsCatered Final_Pax_Qty Underage_Rev Underage_AII 145 145 0 0 165 165 0 0 121 83 25 38 118 116 0 0 108 108 0 0 0 0 0 0 170 161 . 0 0 111 107 0 0 159 135 0 0 113 112 0 0 151 151 0 0 155 155 0 0 153 132 0 0 168 157 0 0 112 106 0 0 116 104 0 0 134 134 0 0 119 119 0 0 180 180 0 0 218 218 0 0 180 180 0 0 143 143 0 0 128 115 0 0 125 125 0 0 164 143 0 0 180 169 0 0 180  180  0  0  Table 20: Identification of an outlier  The explanation for the shortage of flight CP988 on February 3 becomes clear when we look at the change in pre-departure booked loads in the last 6 hours prior to departure. Date  flt_nr  QueryPeriod  Pax_Qty  Capacity  2/3/99  988  5 (6-hour)  57  180  2/3/99  988  4 (3-hour)  58  180  2/3/99  988  3 (2-hour)  107  180  2/3/99  988  2 (1-hour)  112  180  2/3/99  988  1 (30-min)  112  180  2/3/99  988  0 (final)  121  180  Table 21: Booked load history for an outlier  At the decision point for the main order, the booked load was only 58 passengers. By the 1-hour mark, the load had increased to 112. As the ordering policy model is set-up, intermediary uploads can be 0, 7, or 14 meals. We can safely assume that a particular event has taken place and that the flight controller is aware of it. Human intervention could rectify the situation at that point with a special delivery or a modification to the main order.  62  B. Ordering Policy  Parameters  The following is the list o f scenario tests that were run to obtain the M e a l Bank cost figures presented i n Table 17 and Table 18 above. Each parameter corresponds to one o f the text boxes i n the M a i n F o r m o f the Costing Analysis Database (see p.37 i n Appendix B). ID 1 2 3 4 5 6 7 8 9 10  M  m Reduc. Mnxupload  18 10 18 10 18 18 18 18 18 18 18 18  10 10 10 10 10 10 10 10  1 1 1 1 1 1 1 1 1 1  10 10 10 10 10 21 21 21 21 21  1 1 1 1 900 900 900 900 900  StopFlight starthist stophist startrest 99 1 '15/99 1/31/99 2/1/99 99 1/15/99 2/28/99 3/1/99 99 1/15/99 3/31/99 4/1/99 99 4/1/99 4/30/99 5/1/99 4/1/99 99 5/31/99 6/1/99 999 1/15/99 1/31/99 2/1/99 999 1/15/99 2/28/99 3/1/99 999 1/15/99 3/31/99 4/1/99 999 4/1/99 4/30/99 5/1/99 999 4/1/99 5/31/99 6/1/99  2/28/99 3/31/99 4/30/99 5/31/99 6/30/99 2/28/99 3/31/99 4/30/99 5/31/99 6/30/99  Table 22: Ordering policy parameters  63  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0099498/manifest

Comment

Related Items