Upload
shalvi-shrivastava
View
77
Download
0
Embed Size (px)
Citation preview
Starbucks Simulation
Starbucks Centre
Steger Student Life Centre,
University of Cincinnati
Submitted by:
Shalvi Shrivastava
M10839782
University of Cincinnati
Table of Contents
1. Introduction and Problem Statement 3
2. Data and Model Distributions 5
3. Simulation Model in Arena 8
4. Results And Interpretations 14
5. Improving the System 16
6. Conclusion and References 18
7. Appendix 20
1. Introduction and Problem Statement 1.1 Introduction
Starbucks, based at Steger Student Life Centre, University of Cincinnati, is one of the most popular spots, specialising in beverages such as Coffee and Tea. Every day, nearly millions are served around the world.
We have used Simulation of Starbucks here to assist in evaluating operations within the coffee shop in an effort to evaluate its performance in this highly competitive industry. Data was collected from Starbucks Centre at Steger Student Life Centre after the permission of the then present shift lead named Kayla Parr. The primary aim is to ensure continued financial returns and productivity. A simulation model was built that used customer order database to ensure continued financial returns and optimal productivity. It used customer database to allow queueing methods, order taking and other service procedures to be analysed.
1.2 Description of Problem
Starbucks is one of the busiest place on campus. University of Cincinnati has classes throughout the
day and as well in the evening. This makes the coffee centre more popular. The basic flow of activity
inside the shop is shown below.
Steps involved in the process of acquiring service:
1. Customer arrives at Starbucks.
2. Customer joins the queue of order and billing.
3. Customer waits if the server is busy.
4. Customer reaches server and places the order.
5. Customer waits in the waiting area to get served.
6. Customer gets served and exits.
The above cycle continues for next customer in the queue at any server.
Standing in queues is an inconvenience for customers everywhere. Taking the example of campus
Starbucks, with classes in vicinity, waiting in line is a loss for both, students as well as professors,
who are in rush to classes.
The parameters obtained from the above procedures could be used to analyse ways to potentially
increase efficiency and decrease queue length.
1.3 Assumptions
The coffee Shop works for 12 hours a day.
There are two counters, one for Ordering and Billing having one server and Waiting counter
which has two servers.
The system is memoryless, in that each arrival is independent of the previous arrival, and
that the variables are identically distributed.
The servers are working throughout without any change in shifts.
2. Data and Model Distributions
2.1 Data Collection:
In order to collect the data needed to create simulations and analyse wait times, trips to Starbuck
shops located in Steger Student Life Centre were taken. An equal amount of time of 6 hours was
spent in taking data. Tables were set up in each location in order to collect data regarding the wait
times of customers while in line and while waiting for an ordered drink. In order to get accurate
information, tables were chosen that were in view of the entrance of the shop, the cashier’s
counter, and the waiting area.
Entities are as following:
Customers
Servers
Waiting area
In order to record the times spent in the queues within the store, a timer was used. The timer was
pressed once a customer entered the store, once a customer ordered a drink, and once a customer
received a drink. This was helpful in ensuring that the data recorded using the timers was not mixed
up between different customers who ordered different. All of the information being recorded was
sent to a spreadsheet. We have to be careful as snacks and beverages have different average
preparation times.
2.2 Variables:
Input:
Inter arrival Time of Customer
Time taken at Order and billing counter
Serving time
Response:
Waiting time – Time for which customer waits
Average Wait time for order
Maximum Wait time for order
Average Wait time for order
Maximum Wait time for order
Average Queue – Number of average people in a queue for giving orders
Maximum queue – Number of maximum people in queue for giving orders
Average wait time for delivery
Maximum wait time for delivery
Average total time Spent by Customer
Maximum Total Time spent by customer
Utilization – It is estimated by dividing the amount of time than the server is busy during a
simulation by the amount of time covered by the simulation.
2.3 Distribution:
The raw data collected from Starbucks was saved as a text file. Using Input Analyser, we can find out
the best fit for the data given. With the fit all option these are the distributions that were observed.
Inter Arrival Time:
Similarly the Input Analyser was fitted with all the other raw input data of Counter and Billing,
Beverage Service and Snacks Service distribution.
Counter and Billing Distribution:
Beverage Service time Distribution:
Snacks Service time Distribution:
3. Simulation Model in Arena
3.1 Modelling the System:
We will be using Create, Process, Decision and assign module for creating the model in Arena.
1. Arrival of Customer
2. Counter and Billing.
3. Order placing: Beverages and Snacks.
4. Service line wait
5. Customer Exit
We will go through the model step by step.
1. Arrival of Customer:
It is represented by the Create Module in Basic Process which is called ‘Arrival of Customers’.
The expression is the distribution obtained from ‘Input Analyser’ from Inter Arrival time.
2. Order and Billing Counter
The Queue for the counter and billing is given by the following process:
3. Ordering Preference? Snacks or Beverage?
Next we had observed that the Orders placed by the customers were divided in either beverages or
snacks. They have a different time distribution as snacks take lesser time to be serviced than the
beverages. It was observed that the orders were divided in 78:22 ratio, where 78% represented the
orders for beverages and 22% were for snacks.
Hence we placed a 2-way decision module after the billing and counter queue.
4. Beverage Counter
Approximately 78% of customers prefer to purchase beverages from Starbucks. Hence an ‘Assign’
module is placed after the decision module with the distribution obtained from ‘Input Analyser’ for
Beverage Preference distribution. The attribute here is assigned ‘Prep_Time’ which will converge
into Service Module after it.
5. Snacks and Cookies Counter
Snacks are preferred by approximately 22% of incoming customers. The time taken for reheating the
already prepared snacks is lower than preparing the beverage. The distribution is as obtained from
the ‘Input Analyser’. The attribute here is also assigned ‘Prep_Time’ which will converge into Service
Module after it.
6. Servicing Queue
The Servicing Module, refers to the total time a customer will wait for his order to be ready, this will
be the addition of the ‘prep_time’ obtained from the previous assign modules ‘Beverages’ and
‘Snacks’.
7. Customer Exits
The final one was the dispose module as shown. It represents exiting of Customer after collecting his
order.
8. Run Setup and Replications
With 100 replications I have run my model for 12 hours (720 minutes ), with my base time as
minutes.
4. Results and Interpretations
4.1 Output Result:
We get a detailed report after we run the model as described above. It has a pre-defined Number
Out. This gives us the Number Out i.e. those who have successfully left the system. Here I have run
the model for 12 Hours, which gives me the Number Out as 453, for 100 replications.
The detailed report produced by Arena is given below. We can view the following categories:
Entity
Queue
Resource
1. Entity
Under Entity we have Time. It gives us output regarding Average Value, Minimum, and
Maximum. The Output is as follows:
The average waiting time for a customer is 6.42 min and the average service time for customer is
4.46 min. We have to reduce the overall system time.
1. Queue:
We obtain the Waiting Time and Number of Entities waiting from the queue report.
We see that the waiting time is .63 min for Order and Billing Counter while 5.7 min for Servicing
Queue.
2. Resource:
Under resource, we look at the Scheduled Utilization of the Resources.
This gives the utilization of all the resources. The output is as follows:
It can be seen that ‘Resource 2’ for Service Queue is used maximum while the other resource for
Order and Billing Queue is less.
5. Improving the System
5.1 Suggestions:
To reduce the Average Total Time, we increase the capacity of
Order and Billing Counter
Servicing Counter
This would reduce the waiting time of queues.
Changes:
The Resource Capacity of Order and Billing Counter was increased to 2.
The Resource Capacity of Servicing Counter was increased to 3.
5.2 Comparison of Results:
Rockwell Software provides us with a complimentary tool with Arena called Process Analyser. We
can use it to compare different scenarios when we increment the resource at each counter. We used
it for two scenarios:
1. Increasing Order and Billing Resource Scenario: Here we increment Resource_1 at Billing and
Order Counter from ‘1’ to ‘2’.
2. Increasing Serving Resource Scenario: Here we increment Resource_2 at Serving Counter
from ‘2’ to ‘3’.
It is seen that the Customer Total Time in system is reduced from 10.895 min to 9.865 min and Time
average of number of people in system (WIP) from 7 to 6 in Scenario 1 (Increasing Order and Billing
Resource Scenario). This is further reduced in Scenario 2 (Increasing Serving Resource Scenario) with
Customer Total Time in System to be 5.143 min and Time Average of number of people in system to
be 3.
Here, the best scenario is marked in red. Both the charts depict that the improved model is
the best when we increment the Serving Resource from ‘2’ to ‘3’.
5.3 Alternate Scenario
We observe from the Process Analyser Output that the best case scenario would be if we
increment server at the Serving Counter.
Hence once we increase the resource in our current model with 100 replications, we observe
the following result.
We see the Serving Queue Average Waiting Time decreases from 5.79 min to 0.036 min. This
is an incredible reduction in time, making it convenient for customers. There is also
reduction in the Average Number of Customers waiting in queue, from 4 (approx. value 3.7)
to 0.
6. Conclusion and References
6.1 Conclusion
It has been observed from the Simulation modelled in Arena Software that the waiting time
at the Serving Queue was large. Once we take that counter into consideration, we get the
best scenario i.e. when we increment the resources at Service Counter we get better result
where Customers have lower system time. Changing the amounts of resources can result in
changes within the queue times due to resources being employed in different areas. By
testing multiple variations of resources, an optimal model for efficiency and profitability for
Starbucks was hence found.
6.2: References
1. Simulation with Arena (W.David Kelton, Randall P. Sadowski, Nancy B. Zupick, Rockwell
Automation)
2. Data – Starbucks (University Of Cincinnati)
3. Image : https://www.uc.edu/mainstreet/sslc.html
Appendix
Category Overview19:04:54 December 4, 2016
Starbucks Model
Time Units:Replications: 100 Minutes
Values Across All Replications
Key Performance Indicators
AverageSystemNumber Out 453
Model Filename: Page of1 5C:\Users\Shalvi\Desktop\Simulation\Shrivastava_Shalvi_Project2
Category Overview19:04:54 December 4, 2016
Starbucks Model
Time Units:Replications: 100 Minutes
Values Across All Replications
Entity
Time
VA Time MaximumAverage
MinimumAverageHalf WidthAverage
MinimumValue
MaximumValue
4.4663Entity 1 0.02 4.2452 4.6758 0.8458 6.9708
NVA Time MaximumAverage
MinimumAverageHalf WidthAverage
MinimumValue
MaximumValue
0.00Entity 1 0.00 0.00 0.00 0.00 0.00
Wait Time MaximumAverage
MinimumAverageHalf WidthAverage
MinimumValue
MaximumValue
6.4284Entity 1 0.76 1.5219 25.8201 0.00 44.2353
Transfer Time MaximumAverage
MinimumAverageHalf WidthAverage
MinimumValue
MaximumValue
0.00Entity 1 0.00 0.00 0.00 0.00 0.00
Other Time MaximumAverage
MinimumAverageHalf WidthAverage
MinimumValue
MaximumValue
0.00Entity 1 0.00 0.00 0.00 0.00 0.00
Total Time MaximumAverage
MinimumAverageHalf WidthAverage
MinimumValue
MaximumValue
10.8947Entity 1 0.78 5.7671 30.4334 0.9395 49.6569
Other
Number In MaximumAverage
MinimumAverageHalf WidthAverage
Entity 1 1.15 461.69 478.00 448.00
Number Out MaximumAverage
MinimumAverageHalf WidthAverage
Entity 1 1.30 453.20 470.00 437.00
WIP MaximumAverage
MinimumAverageHalf WidthAverage
MinimumValue
MaximumValue
6.9508Entity 1 0.49 3.6005 19.2671 0.00 31.0000
Model Filename: Page of2 5C:\Users\Shalvi\Desktop\Simulation\Shrivastava_Shalvi_Project2
Category Overview19:04:54 December 4, 2016
Starbucks Model
Time Units:Replications: 100 Minutes
Values Across All Replications
Queue
Time
Waiting Time MaximumAverage
MinimumAverageHalf WidthAverage
MinimumValue
MaximumValue
0.6393Order and Billing Counter.Queue
0.04 0.3760 1.4167 0.00 9.0468
5.7980Servicing Queue.Queue 0.76 1.0793 25.3153 0.00 43.8955
Other
Number Waiting MaximumAverage
MinimumAverageHalf WidthAverage
MinimumValue
MaximumValue
0.4107Order and Billing Counter.Queue
0.02 0.2363 0.9194 0.00 6.0000
3.7093Servicing Queue.Queue 0.48 0.6745 16.0505 0.00 28.0000
Model Filename: Page of3 5C:\Users\Shalvi\Desktop\Simulation\Shrivastava_Shalvi_Project2
Category Overview19:04:54 December 4, 2016
Starbucks Model
Time Units:Replications: 100 Minutes
Values Across All Replications
Resource
Usage
Instantaneous Utilization MaximumAverage
MinimumAverageHalf WidthAverage
MinimumValue
MaximumValue
0.9022Resource 1 0.00 0.8690 0.9322 0.00 1.0000
0.9643Resource 2 0.00 0.8838 0.9971 0.00 1.0000
Number Busy MaximumAverage
MinimumAverageHalf WidthAverage
MinimumValue
MaximumValue
0.9022Resource 1 0.00 0.8690 0.9322 0.00 1.0000
1.9286Resource 2 0.01 1.7677 1.9942 0.00 2.0000
Number Scheduled MaximumAverage
MinimumAverageHalf WidthAverage
MinimumValue
MaximumValue
1.0000Resource 1 0.00 1.0000 1.0000 1.0000 1.0000
2.0000Resource 2 0.00 2.0000 2.0000 2.0000 2.0000
Scheduled Utilization MaximumAverage
MinimumAverageHalf WidthAverage
Resource 1 0.00 0.9022 0.9322 0.8690
Resource 2 0.00 0.9643 0.9971 0.8838
0.900
0.910
0.920
0.930
0.940
0.950
0.960
0.970
Resource 1Resource 2
Model Filename: Page of4 5C:\Users\Shalvi\Desktop\Simulation\Shrivastava_Shalvi_Project2
Category Overview19:04:54 December 4, 2016
Starbucks Model
Time Units:Replications: 100 Minutes
Values Across All Replications
Resource
Usage
Total Number Seized MaximumAverage
MinimumAverageHalf WidthAverage
Resource 1 1.14 461.28 477.00 447.00
Resource 2 1.30 455.18 472.00 439.00
455.000
456.000
457.000
458.000
459.000
460.000
461.000
462.000
Resource 1Resource 2
Model Filename: Page of5 5C:\Users\Shalvi\Desktop\Simulation\Shrivastava_Shalvi_Project2