Nova Southeastern University
H. Wayne Huizenga School
of Business & Entrepreneurship
Submitted to: / Dr. Phillip
Submitted by: / Nicole Mateus
Robert Edson
Abdellah Sabere
Date of Submission: November 3, 2013
Title of Assignment: ZZ Airlines Simulation Case Study
CERTIFICATION OF AUTHORSHIP: I certify that I am the author of this paper and that any assistance I received in its preparation is fully acknowledged and disclosed in the paper. I have also cited any sources from which I used data, ideas or words, either quoted directly or paraphrased. I also certify that this paper was prepared by me specifically for this course.
Student's Signature: NM, RE, AS______
*****************************************************************
Instructor's Grade on Assignment:
Instructor's Comments:
TITLE OF RUBRIC: Case Analysis (Page 1 of 2) / Course: QNT 5040LEARNING OUTCOME/S: (see syllabus) / Date: 11-03-13
PURPOSE: To facilitate effective decision making under uncertain conditions by quantifying risk. / Name of Student: NM, RE, AS
VALIDITY: Best practices in Monte Carlo simulation. / Name of Faculty:Dr. Phillip
COMPANION DOCUMENTS: Assignment and format instructions, Case
Earning maximum points in each box in ‘PROFICIENT’ column and / or
points in columns to the right of ‘PROFICIENT’ meets standard.
< less quality ...... more quality >
Performance Criteria / Basic / Developing / Proficient / Accomplished / Exemplary / Score
Identify the problem / Does not
identify the problem, or does not identify the right problem.
(0 pts) / Identifies symptoms
(5 pts) / Identifies some elements of the problem.
(10 pts) / Substantially
identifies the problem.
(12 pt) / Effectively and succinctly
identifies the problem.
(15 pts)
Describes assumptions and methods / Does not describe assumptions and methods used
(0 pts) / Does not precisely describe the
assumptions and methods used
(3 pts) / Somewhat describes assumptions and methods used
(7 pts) / Substantially
describes assumptions and methods used
(8 pts) / Effectively describes assumptions and methods used
(10 pts)
Calculate statistics using a spreadsheet / Does not calculate appropriate statistics using a spreadsheet and/or
does not provide evidence of calculations
(0 pt) / Calculates appropriate statistics using a spreadsheet (most answers are not
correct)
(13 pts) / Calculates appropriate statistics using a spreadsheet (not all answers are correct)
(21 pts) / Calculates appropriate statistics using a spreadsheet (most answers are correct)
(25 pts) / Effectively
calculates statistics using a spreadsheet (almost all answers are correct)
(30 pts)
Explain
implications of
output of statistical analysis / Does not explain
implications of
output of statistical analysis
(0 pt) / Partially
explains
implications of
output of statistical analysis
(3pts) / Somewhat explains
implications of
output of statistical analysis
(7 pts) / Substantially
explains
implications of
output of statistical analysis
(8 pts) / Effectively explains
implications of
output of statistical analysis
(10 pts)
TITLE OF RUBRIC: Case Analysis, cont. (Page 2 of 2) / Course: QNT 5040
LEARNING OUTCOME/S: (see syllabus) / Date: 11-03-13
PURPOSE: To facilitate effective decision making under uncertain conditions by quantifying risk. / Name of Student: NM, RE, AS
VALIDITY: Best practices in Monte Carlo simulation. / Name of Faculty: Dr. Phillip
COMPANION DOCUMENTS: Assignment and format instructions, Case
Earning maximum points in each box in ‘PROFICIENT’ column and / or
points in columns to the right of ‘PROFICIENT’ meets standard.
< less quality ...... more quality >
Performance Criteria / Basic / Developing / Proficient / Accomplished / Exemplary / Score
Generates solutions based on analysis and context / Does not
generate appropriate solutions based on analysis and context.
(0 pt) / Generates solutions (does not justify conclusions).
(7 pts) / Partially: *generates and justifies solutions based on analysis and context; and *justifies conclusions.
(15 pts) / Substantially: *generates and justifies solutions based on analysis and context; and *justifies conclusions.
(17 pts) / Effectively: *generates and justifies solutions based on analysis and context; and *justifies conclusions.
(20 pts)
Uses prescribed format(including cover sheet and grading rubric)and writing style (language, grammar, punctuation, and spelling) / Does not use prescribed format and writing style
(0 pt) / May use prescribed format OR writing style (only one)
(3 pts) / Generally uses prescribed format and writing style
(7 pts) / Substantially uses prescribed format and writing style
(8 pts) / Effectively
uses prescribed format and writing style
(10 pts)
Uses APA format
(APA Style Manual 6.0) / Does not provide references.
(0 pt) / Does not apply APA style to references.
(1pts) / Partially applies APA style to references.
(3 pts) / Substantially applies APA style to references.
(4 pts) / Effectively applies APA style to all references.
Optimal quality and quantity of citations.
(5 pts)
OVERALL GRADE (100 total possible points): / %
Comments: ______
1
Management Report for ZZ Airlines
Executive Summary
ZZ Airlines is trying to determine whether or not to proceed with their advertising campaign to increase phone inquiries. If they decide to proceed, they want to determine if hiring a second agent to help with its toll-free reservation system is recommended. A sample of data was taken based on similar campaigns in the past.
After completing several types of analysis using @RISK, it can be concluded that adding an additional agent to help the toll-free system is a must but proceeding with the advertising campaign is determinant on the owner’s willingness to exceed his low wait-time expectations.
Background
ZZ airlines recently opened its doors as a commuter service. They have seen business grow rapidly and thus identified the need to improve their customer service system in order to meet the demands of their growing customer base. Initially ZZ airlines started a toll-free reservation system that operated between 12:00A.M. and 6:00 A.M. with only one agent on duty. In addition to the toll-free system, they are planning to roll out an advertising campaign. With this campaign and growing customer base, management is expecting higher than normal call volumes, and needs to prepare so they are able to meet the owners demand for no more than a three or four minute customer hold time. High-quality customer service is crucial to the success of any business and it is important for ZZ airlines to make the right adjustments in order to keep their customers satisfied.
Problem
The main problem we are facing is how we ensure our recommendation to hire a second agentis justified from an operational cost. Our dilemma is to maintain customer satisfaction while ensuring an efficient agent utilization with the least amount of agent idle time. One of the criteria set-forth by management for customer satisfaction is to maintain the wait time between 3 and 4 minutes.
Analysis
Our analysis started by looking at the original call distribution and identifying the key performance indicators that would yield a stable condition for the queue. The main key performance indicators we measured and analyzed are indicated in the following order:
- Average wait time
- Maximum wait time
- % Agent idle time
- % Agent utilization
Before we proceeded with our analysis and comparison of the above key indicators, we looked at the mean service time for each case scenario and compared it to the mean inter-arrival time of the calls. If the mean service time was less than the mean inter-arrival time we simply concluded that the probability that the queue is stable would be higher. This means that the agents were able to take calls and service customers gracefully without impacting customer satisfaction and resume to take the next call, which did not sit in the queue for more than three minutes. A simple glance at the data shows the mean service time of a sample of incoming calls from the original call distribution was slightly higher than the mean service times. This led us to the assumption that the queue was at higher capacity and that the one agent servicing customers prior to the advertising campaign did not have too much idle time available between calls.
Data analysis prior to the Advertising campaign
Using the sample data of current incoming calls from our homework, with one agent servicing customers, the average wait time was fairly high at 46.88 minutes. The agent total idle time for the entire shift was 2.55 minutes which represented 0.71% idle time. This data showcased that the queue in this case scenario was far more stretched and quite unstable since customers had to wait for longer than the desired 3-4 minutes that management set out as threshold for customer satisfaction.
Table 1. Stats for 1 Agent
Total wait time / 4688.484Average wait time / 46.88484
Max wait time / 91.50561
Total idle time / 2.552547
Agent Idle time % / 0.71%
Further analysis of the data using the @RISK tool, we generated 1 simulation of 1,000 iterations for the case scenario of one agent prior to the advertising campaign.According to Palisade Corporation, at each iteration formulated, @RISK draws a new set of random numbers for the @RISK distribution functions which in this case are exponential distribution for current incoming calls with a mean of 3.356 and triangle distribution for service times with a mean of 3.2292. The results given on the output reports helped us validate some of our assumptions and gave us a clearer picture on the agent utilization.
Graph 1. Average Wait Time for 1 Agent Graph 2. Max Wait Time for 1 Agent
As shown above (graph 1), using our critical value at 95%, we see the average wait time for our simulation to fall between 13.8 and 67.6 minutes. These numbers fall within 2 standard deviations of our mean of 39.1 minutes. Our data is slightly skewed to the right for both graphs. The max wait time falls between 34.7 and 127.9 minutes within 2 standard deviations of the mean 80.21 minutes. This clearly shows 1 agent is not enough to support the great customer satisfaction ZZ Airlines is trying to achieve. The longer in the morning it goes, the longer the wait time will be as more customers start to call. Agent utilization will be high but at the expense of longer wait times for the customers which is not what they want.
Adding a second agent prior to the advertising campaign yielded greater results that were aligned with improving customer satisfaction. In fact, the maximum wait time was reduced by 98% to 2.66 minutes instead of 91.50 minutes for 1 agent. The average wait time was null while the % agent idle time was increased to 38.85%. This would allow agents to ensure proper entering of the data and documentation of any cases they logged. Since the agent idle time is higher, we would even recommend for agentsto create knowledge base articles that customers can leverage online to deflect some of the simple incoming calls.
Table 2. Stats for 2 Agents
Total wait time / 16Average wait time / 0
Max wait time / 2.6651
Total idle time / 280
Agent 1 Idle time % / 36.674%
Agent 2 Idle time % / 41.029%
Total idle time % / 38.85%
Graph 3. Average Wait Time for 2 Agents Graph 4. Max Wait Time for 2 Agents
Graph 3 above clearly shows the better service that is received for the customers in regards to wait time. For our simulation of 1,000 iterations, the average wait time is 1.15 minutes which is a short reasonable time to wait and falls way ahead of the 3 minute mark they were aiming for. The max wait time would be 25.5 minutes which falls outside of 2 standard deviations from the mean of 7.75 so there is a very small probability that will happen.
Graph 5. Total Idle Time % for 2 Agents Graph 6. Idle Time % for 1 Agent
Looking further at the agent % utilization in this case scenario validated our thought process that having two agents prior to the advertising campaign is highly recommended. Based on the mean, the % agent utilization was at more than 96% with one agent versus 62% with both agents. This gives each agent some time to breathe and take a moment to relax before taking the next caller. The graph for both agents is skewed to the left while the graph for 1 agent is skewed to the right. We believe the 34% difference in agent utilization greatly outweighs the shorter wait time the customer would be receiving.
Data Analysis After the advertising campaign
Looking further at the data after prior advertising campaign shows similar trends. As table 4 shows, the average wait time increased from 28.76 to 46.37 minutes after the campaign was done. This illustrates that the advertising has contributed to an increase in the long wait time. The total idle time has doubled after the campaign. Since the wait time has surpassed the 4 minute max, customer satisfaction has indeed been eroded and this alone constitutes a reason to adjust the staffing number to two agents instead of one. The data below shows the comparison between one agent before and after the advertising campaign:
Table 3. Stats Before Campaign 1 Agent Table 4. Stats After Campaign 1 Agent
Total wait time / 2876.38 / Total wait time / 4637.746726Average wait time / 28.7638 / Average wait time / 46.37746726
Max wait time / 63.75539 / Max wait time / 80.45607433
Total idle time / 2.552547 / Total idle time / 5.457470206
Agent Idle time % / 0.71% / Agent Idle time % / 1.52%
Graph 7. Average Wait Time Graph 8. Average Wait Time
After Campaign for 1 Agent After Campaign for 2 Agents
Looking at our simulation of 1,000 iterations, the average wait time with 2 agents brings down the mean from 56.35 to 1.77 minutes. The 95% critical value for 2 agents has average wait times between .72 and 3.62 minutes both falling inside 2 standard deviations from the mean. Those numbers show that 2 agents even after the campaign, can still produce high volume of calls and still maintain the 3 to 4 minutes of wait time for the customers.
Graph 9. Max Wait Time Graph 10. Max Wait Time
After Campaign for 1 Agent After Campaign for 2 Agents
Adding two agents in our staffing model and running the @RISK simulation provided a great deal of insight on the improvements of the KPI’s. In fact, the maximum wait time customers would have to endure with 2 agentswould be 28 minutes and that’s falling way outside 2 standard deviations from the mean. Based on the mean of 9.4 minutes with 2 agents, it still falls past the 4 minute max the airline wants a customer to wait but these numbers are worst case scenarios which have some possibility of happening but the best case scenarios still outweigh the worst case scenarios.
Graph 11. Agent Idle % for Graph 12. Agent Idle % for
1 Agent after Campaign 2 Agents after Campaign
According to graph 12, our total idle time for both agents after the campaign had a mean idle % of27.5%. That’s roughly almost 75% agent utilization which still looks like a high percentage for the increasing amount of calls coming in. This illustrates great improvement in the quality of service the agents are providing since they are not rushing to pick up calls that have been waiting on hold for more than 3 minutes. This staffing model will reduce from agent burn out and low morale compared to the one agent model where he/she is being utilized almost 98% of the time. It is also a fault tolerant system in case one of the agents gets stuck on a call longer than normal or has to use the restroom or get a sip of water.
Based on tables 5 and 6, it shows both agents have similar idle times meaning they more or less spend the same amount of time on calls. We can only recommend that the agents have some degree of experience with airline customer service to be able to expedite the calls quicker.
Table 5. Summary Statistics for Table 6. Summary Statistics for
Agent 1 Idle Time % after Campaign Agent 2 Idle Time % after Campaign
Statistics / Percentile / Statistics / PercentileMinimum / 8.276% / 5% / 14.037% / Minimum / 7.662% / 5% / 14.917%
Maximum / 54.843% / 10% / 16.928% / Maximum / 55.102% / 10% / 17.732%
Mean / 27.388% / 15% / 18.779% / Mean / 27.702% / 15% / 19.319%
Std Dev / 8.152% / 20% / 20.371% / Std Dev / 8.031% / 20% / 20.835%
Variance / 0.006645135 / 25% / 21.531% / Variance / 0.006450188 / 25% / 22.098%
Skewness / 0.197034985 / 30% / 22.458% / Skewness / 0.251267417 / 30% / 23.185%
Kurtosis / 2.865965601 / 35% / 24.047% / Kurtosis / 2.927817806 / 35% / 24.191%
Median / 27.325% / 40% / 25.019% / Median / 27.291% / 40% / 25.377%
Mode / 29.752% / 45% / 26.242% / Mode / 27.325% / 45% / 26.433%
Left X / 14.037% / 50% / 27.325% / Left X / 14.917% / 50% / 27.291%
Left P / 5% / 55% / 28.392% / Left P / 5% / 55% / 28.373%
Right X / 41.106% / 60% / 29.504% / Right X / 41.366% / 60% / 29.292%
Right P / 95% / 65% / 30.459% / Right P / 95% / 65% / 30.586%
Diff X / 27.069% / 70% / 31.403% / Diff X / 26.450% / 70% / 31.684%
Diff P / 90% / 75% / 32.475% / Diff P / 90% / 75% / 32.851%
#Errors / 0 / 80% / 34.267% / #Errors / 0 / 80% / 34.517%
Filter Min / Off / 85% / 35.595% / Filter Min / Off / 85% / 36.275%
Filter Max / Off / 90% / 37.839% / Filter Max / Off / 90% / 38.357%
#Filtered / 0 / 95% / 41.106% / #Filtered / 0 / 95% / 41.366%
Conclusion and Recommendation
When reviewing the data outputs above, it is apparent that ZZ Airlines is better off adopting a two agent customer service model instead of staying with their current one agent model. In order to lower customer wait times and meet the expectation set forth by the owner of having no more than a three to four minutes hold time, this should be put into action right away. Not only will it improve customer experience but it will also improve the agents overall quality of work life. It is no secret that customer service plays a key role in the sustainability and profitablity of a company. In order to ensure positive customer service it is also important to have happy and motivated employees. Overworked and unhappy employees are not able to provide the top-notch service that is required. It is our recommendation that having two agents in place will strengthen the business and keep customers satisfied. Regarding whether or not they proceed with the campaign, if the owners of the airline feel they do not want to risk customers waiting no more than 4 minutes maximum, then they should not proceed with the campaign because based on our simulation with 1,000 iterations, there may be instances where 2 agents might get extended calls and wont be able to control the wait time below their requested mark. The agent idle % might also pass 50% in certain scenarios so that number might be too high for some owners. Definitley they have to have 2 agents even if they don’t want to proceed with the campaign. If they feel the small probabilities of both agents getting tied up in extended calls are worth the reward of higher call volumes with low waiting times, then doing both campaign and getting a second agent is the way to go. An added recommendation would be to screen the customer as soon as they call so the agent can know what the customer is calling about before they are passed on to them. That information would save some time and the customer will not feel like that time was considered waiting time.
1
1