The Holborn Airways meet a revenue problem right now, which wants to get the advise on a most profitable routes, the trans-Kingsway route between the City and the West End. For this route, it has two different fares, full fare and discount fare. Full fare targets to business people which company can earn £100; discount fare targets to student which company can earn £35. As well, the company also set up the protection level for the full fare ticket that is 20 right now. Based on the different revenue between these two different flight fares, the purpose for the operational research team to maximize the profit for Holborn Airways.
This report will present the process to build the stimulation model, data analysis and recommendations. Based on data analysis and the simulation model built, the recommendation for the optimal protection level is 30 with the original controlling capacity strategy; it is 70 for the new idea that suggested by the Head of Marketing. By keeping protection level at 20 seats, compared with the Head of Marketing’s idea, the original protection level mechanism is better because it has higher revenue. However, this report has limitations because the reservation tickets’ data does not include the tickets that customers want to cancel and the sample generated in the model is a little different from the real data that includes all unpredictable factors. Therefore, when the CEO of Holborn Airways made final decision, the CEO should also consider other limitations for the models, uncertainties and unpredictable factors in the real world.
Data Analysis
Processes for Modeling
Data Collection and Transformation
The first step is to collect available data and transform them to the data can be used. The data the Holborn provided are 5 flights which was collected manually by the reservation staff, which is the potential passengers who want to book a tickets that is also reflect the potential demand.
However, the original data includes the 5 different flights’ fare type and booking time. The first step is to sort both discount fare and full fare and then transform the booking time to hours in order to find the interval time for both different fare type in different flights. As the above process, the interval can be found. In appendix A, it shows the lambda (λ) for each flight’s fare types can also be calculated as (1/ mean of interval time). The purpose for this is to generate two averages Poisson processes for discount fare and full fare with rates λ1 and λ2. As well, it was also used to find the average interval rate for Calls for the module build, which is 5.679 hours. Based on the rates λ1 and λ2, the probability for full fare and discount fare can be calculated as λ1 (Full Fare) / λ1 (Full Fare) + λ2 (Discount Fare) and λ1 (Discount Fare) / λ1 (Full Fare) + λ2 (Discount Fare); which are equal to 0.533 and 0.467.
The purpose for this process it to collect valuable data to generate Monte Carlo sampling for the simulation model. The value calculated from the original data is robust enough because it has 5 different samples (5 different flights) with at least 120 trails (Potential Customers), which is more than 30. Therefore, the T-distribution can be used and it is robust enough to generate the sampling.
Assumption and Model Building
After collected and revised the received data, the next step is to develop the assumptions that are used to set up two different stimulation models because it has two different scenarios for the protection level strategy that is the new idea from the head of marketing. In Appendix B and C, they show different model structures developed by the operational research team, which includes cumulative arrival time, total ticket count, number of discount ticket sold, response for the ticket, type of call, ticket number for both under protection level and tickets without protection, potential revenue, and real revenue.
Scenario 1 (Original Strategy for Protection Level): The protection level’s ticket has to be sold first with full fare. If not, the discount ticket cannot be sold.
Scenario 2 (Head of Marketing Idea’s): The discount tickets can sell with full fare at same time, but it cannot sell more than the tickets outside of protection level.
Cumulative arrival time: This column is based on the exponential distribution with the mean as average interval rate found in Appendix A because it is random. Meanwhile, the cumulative arrival time cannot above 720 hours because 1 month has 720 hours (24 hours*30 Days).
Type of Call: This column is type of tickets that are based on the discrete distribution with the probability in the Appendix A. “1” represents full fare and “2” represents discount fare.
Total Ticket Sold: This column is based on how many tickets will be accepted however, the maximum seat is 100, which means the maximize ticket is 100.
Discount ticket sold: This column counts for discount ticket; the limitation is discount ticket cannot sale more than the Total seat (100) – Protection Level.
Response: This column will shows the whether the potential customers’ tickets are decline or accept based on cumulative arrival time that has to be before 720 hours.
Tickets under Protection Level: It is the available tickets for protection level.
Tickets outside of Protection Level: It is the available tickets outside of protection level.
Potential revenue: It is the potential revenue that Holborn can make based on the different class of potential customers.
Real revenue: It is the real revenue that Holborn can make based on whether the customers’ requests for tickets are accepted or not.
Limitation for the models
Latin Hypercube Sampling
These models are built on Latin Hypercube Sampling, which means it generated by near random samples stratifies the input probability. Although it improved the accuracy from the Monte Carlo sampling, it still has some difference from the real data that includes uncertainties and unpredictable factors, such as weather, policy issues, illness for customers, etc. As a result, the sample created by operational research cannot make it exactly same as real world. However, it is robust enough for data analysis; because each scenario will run at least 1000 times for these random samples to find the result to reduce the bias and improve the accuracy.
Limitation and improvement for provided data
Meanwhile, for the data provided by Holborn are not complete because it does not includes other tickets that customers booked from other sources, such as website, airways’ front desks, and etc. As well, it does not show any information about how many customers will cancel the ticket before flight departure. Other factors also should be considered in this analysis, which are the plenty for customers to cancel the ticket, the cost for overbooking and empty seat. Therefore, the data and model can be improved by providing these data in order to make it more accurate.
Protection Level Analysis
The protection level is one of the most essential factor to affect the revenue and the aim for the operational research team is to maximize the revenue and minimize the cost to improve the profit. Therefore, it is the key to solve this revenue problem.
Optimal Protection Level
In this comparison, the operational research team set up seven different protection levels in their models, that are 20, 30, 40 ,50 ,60 ,70, and 80, to find the optimal protection level. After they got the optimal protection level, 30, they also used Risk Optimizer to compare with their result, which shows the same value, 29, is the best seat to protect for full fare passengers.
Trend Analysis
In graph 1, it shows the trend for Scenario 1 with 7 different simulations that represents 7 different protection levels that are from 20 to 80. It shows the sim#2 is the peak for this trend, and after Sim #2 the trend is decreasing, which proof the optimal protection level should be kept as Sim #2 that is 30.
Total Revenue Analysis
In graph 2, it shows the total revenue for these 7 different simulations and the Sim#2 is the best choice. In this graph, the Sim#1 and Sim#2 has the highest level of mean and similar level of lower-quartile but Sim#2 has a greater higher-quartile for total revenue that is around 2% higher than Sim#1’s mean. Therefore, Sim#2 is better than Sim#1.
As well, compared Sim#3 and Sim#2, Sim#3 has a greater higher-quartile for revenue, but the mean is slightly lower than Sim#2. Meanwhile, Sim#3’s lower-quartile is around 6% lower than Sim#2’s mean. Therefore, by considering the risks, Sim #2 is better.
Therefore, after a series of analysis based on the model made by the operational team, the best number of seats to protect for full fare passengers is 29, which means the protection level should keep around 30 in the Scenario (Original Protection Level Strategy).
New Idea’s compare with the protection level mechanism
The Head of Marketing suggests a different way of controlling capacity but the original strategy for protection level is better to use for protection level 20. However, the new control capacity is better to use with a high protection level. To be more specific, if the protection level is above 40-45, the new controlling capacity is better; if not, the original strategy for protection level is better.
In the graph 3, it shows two different scenarios with seven different protection level. Based on the same protection level, 20, the original strategy for protection level is better that is around 6% than the new controlling capacity (Scenario 2). Therefore, right now, the original strategy has higher revenue for them.
However, in the graph 3 also expresses if the protection level changed to more than 45, the situation changes, which means the new strategy for controlling capacity has higher revenue. As well, the new strategy will bring the maximize profit for Holborn Airways. The optimal protection level, 70, can increase 1.5% of revenue compared with the original strategy for controlling capacity.
Therefore, the idea recommend by the Head of Marketing should be used with a high volume of protection level, and the low volume of protection level should use the original strategy.
Conclusion
All in all, after a series of analysis based on the different scenarios, the best number for the seat to protect for full fare is 70 with the new idea suggests by the Head of Marketing. However, for the original controlling capacity strategy, the best seat to keep is 29, which means the protection level should be around 30. In fact, all forecast is wrong, but it is better than nothing. Therefore, the purpose for this report is to provide more valued information for decision making to reduce biases. Thus, the limitations for the models, uncertainties and unpredictable factors in the real world should be considered when the CEO makes final decision.
alexsupernb
grammar_fix.docx
0
