You are on page 1of 23

University of Houston Department of Industrial Engineering

DIGITAL SIMULATION

Final Report Simulation of the Starbucks Store at the UC (Satellite)

Dr. Suresh Khator INDE 6370 - Spring2012

Team members:

NaifAlsaadi Rohit Mishra

April10th, 2012

Table of content

Page No.

1. Problem statement. 1.1Introduction 1.2Problem Definition ... 1.3Objective and Assumption... 1.4Flow Chart ........ 2. Input Data Collection 2.1Data needed...... 2.2Analyzing Input Data ... 3. Modeling ... 3.1Basic System Definition ........ 3.2Assumptions and Simplification ... 3.3 Model Translation... 3.4Model Results............................ 4. Analysis of Results 4.1System Verifications 4.2System Improvement ... 5. Conclusion and Recommendation . 6.Appendix .

1 1 1 2 3 3 4 4 12 12 13 13 16 17 17 18 20 21

ii

Summary

The main objective of any coffee shop is to service a customer in the most cost effective and economical way by utilizing the resources and the right number of employees. This project involves the study and anlysis of a Starbucks coffee shop, situated on university of Houston at the UC (Satellite). Customer satisfaction depends on factors like, serving the coffee within a reasonable time to the customer, ensuring quality and taste of the coffee. The scope is to reduce service times and number of disappoined customers in the most cost effective way by utilizating the resources to their potential. The project team attempted to address a trade-off between system time and number of employees in Starbucks. We conducted the simulation study by changing the number of employees and creating different scenarios. By using ARENA as the simulation software and with statistical analysis, the team concluded that the number of employees are the significant factors in optimizing workflow. The model was then verified and validated. The various scenarios were simulated and the alternatives were analyzed. From the analysis, we have come up with two solutions, one being the cost effective solution and the other being the time effective solution. The two alternatives will be presented to the Starbucks store and a need based solution will be selected.

iii

1. Problem Statement
1.1 Introduction In any coffee shop, the main objective is to reduce service times in the most effective and economical way, which can be done by utilizing the resources at hand to their potential. This project involves the study and analysis of the Starbucks Coffee shop at the UC (Satellite). It is obvious that customer satisfaction depends on factors such as serving the coffee within a reasonable amount of time, the quality of the coffee, the ambience of the coffee shop, and more. The project team will address a trade-off between the system time and number of employees in the coffee shop and use ARENA as the simulation software and perform statistical analysis to find significant factors that will improve the coffee shop performance. The team concluded that the number of baristas and clerks are the significant factors. 1.2 Problem Definition This project aims to analyze the existing workflow process and provide suggestions to improve efficiency and attain a higher employee utilization. The objective is to find the bottlenecks and suggest ways to improve billing time and service time thereby leading to fewer customers in each queue, and increase the customer satisfaction ratings through an incremental approach to modeling. In short, we will try to optimize a trade-off between system cycle time and the number of employees in the coffee shop. Starbucks follows a self service system. When a customer comes to the Starbucks coffee shop, he immediately goes to the counter where a Starbucks employee greets and waits for him to choose from a menu which consists of a wide variety of coffees, they also sell their products and other snacks which are included in the menu. The customer then chooses his particular

coffee and transfers from one spot near the counter to another spot at the end of the counter. The order is given to a particular employee who makes the coffee and gives it to the customer at the receiving end of the counter. Customers then either take their orders to go and leave the shop immediately or drink their coffee there and leave when they finish. The customers who visit Starbucks during peak operation hours experience the longest waiting lines to place their orders and the longest delays to receive their orders. 1.3 Objective and Assumptions The scope of this project is to analyze the parameters which affect the overall process. Thus, we will be focusing on resource utilization of the barista and clerks to improve system cycle times. This will lead to fewer customers in each queue. This project will include the analysis of the store layout.

Figure 1.1 The layout for the Starbucks at UC (Satellite)

1.4 Flow Chart


Dine In Customers

Customers Arrive

Billing Machine Process

Service Process

Customers Exit

To Go Customers

2. Input Data Collection


2.1 Data Needed To simulate a real system accurately, input data has to be collected by observing the real system. Data was collected during the peak time of the coffee shop, 11:00 pm to 2 pm over six days. The data needed to create and simulate the model are the following: Inter-arrivals for customers requesting help or service from Starbucks. That is, the time when the customer enters the Starbucks and joins the billing queue. Billing time. That is, the time that a customer spends to order and pay at the billing counter. Service time. That is, the time that a clerk spends to process an order. Percentage of customers who take "To go" orders. Travel time between stations

This data will be directly used to build the simulation model.

2.2 Data Collection and Processing Procedures Inter-arrivals

The data were collected three different days at different times. One of team members collected the time when a customer arrived. The table below shows inter-arrivals collection schedule: Time 11:00-12:00 12:00-01:00 01:00-02:00 Billing Process Times # collected 80 86 94

Similar to inter-arrivals, billing process time data were collected three different days at different times. The other team member collected the time when a customer arrived a billing machine until got a bill. The table below shows service times collection schedule: Time 11:00-12:00 12:00-01:00 01:00-02:00 # collected 85 82 81

Service Times

Similarly to inter-arrivals and billing process, service time data were collected three different days at different times. Both team member focused on customers, and collected the time when a customer got the bill and the time he/she walked away after he/she got his order (hot coffee, cold coffee, or water). The table below shows service times collection schedule: Time 11:00-12:00 12:00-01:00 01:00-02:00 #collected 88 90 79
4

Data Analysis Fitting of the Inter-arrival Time


We have collected data of the times at which the customers have entered the system. The time between each customer arrival was noted. We noticed that the customers arrive in a bunch generally accompanied by one or two others. The collected data was analyzed and the distribution was obtained which is given in the table below: Number of Data Points Min Data Value Max Data Value Expression Corresponding P Value 260 0.5 305 EXPO(41.3) .128

Fitting of the billing process Time


There is only one resource, a barista with a billing machine. We collected data of the billing times regardless of the method of payment. The collected data for the process time was analyzed using the input analyzer. The highest P value and least mean square error was observed for a BETA Distribution. The table below shows the fitting for collected data: Number of Data Points Min Data Value Max Data Value Expression Correspoinding P Value 248 15 65 15+ 50*BETA(1.2,1.12) .277

Fitting of the Service Time


The service times of the two clerks were collected along with their scheduled breaks. These service times ranged from 10 to 71 seconds. The data was analyzed and BETA Distribution fit the data using the Input Analyzer. Below is the table for the service times.

Number of Data Points Min Data Value Max Data Value Expression Correspoinding P Value

257 10 71 10+61.5*BETA(1.88,2.6) .214

Note: Please see the appendix section for further details on the curve fitting and data points entered.

3 Modeling
3.1 Basic System Definition (problem Statement) The objective of our project is to study, analyze and simulate the actual scenario of the existing processes in the store. The following were the processes observed at the store: A customer enters into the store or system and waits in the queue to place an order. When he/she reaches the counter, they place the order and make a payment. The cashier accepts the payment and gives a receipt to the customer, for his/her order. The customer step aside and wait for the coffee. When the order is placed and payment is received, the order is printed and passed to the kitchen or service area. The clerk in the service area receives the order and prepares the coffee accordingly. When the coffee is ready, the clerk in the service or kitchen area delivers the coffee to that specific customer. The customer goes and sits on a free table or leaves the store or system.

The store offers hot coffee, cold coffee, and snack. They have one counter to place an order and to process the payment and another one to prepare the order. Typically, there is one clerk is working on billing machine and two clerks are working on preparing the orders for customers. 3.2 Assumptions and Simplifications The following are the assumptions made in this project: We will not take into account the different methods of payments and the different types of food the store offers. We assume that the time that customers spend to add sugar to their orders is constant since its effect on the system is almost negligible.
7

We assume that a person sits in the dine-in section for a time period uniformly distributed between 30 to 60 minutes.

3.3 Model Translation 3.3.1 Selection Modeling Language The simulation software package ARENA12 Professional was used to model the system and generate a system animation. We followed the flow chart and tried to stick with the common mnemonic naming techniques. The team at all times tried to think in advance in order to make use of the limited development time. During the modeling translation phase, the system modeling and the animation layout development were conducted simultaneously. 3.3.2 Developing the Model in ARENA The system was modeled using the computer simulation package Arena 12.0 professional. Modeling components and modeling issues are addressed as follow. Generate Customers We used the create module in our model to generate customers within the system. We put a warm up period of 10 minutes since there might be customers in the system at 11 am. At time 0, we have no customer in the system and inter-arrival time for customers follows the following expression: EXPO (41.3), all time in second. Place Order Customers get into a single queue before they get served by the cashier. We used the process module here, and the resource type named is Barista. Orders are handed over to the kitchen(service area) at this point, after payment transactions are made. The process time follows
8

a BETA distribution: 15 + 50 * BETA(1.2, 1.12).Then the customers will be routed to the Servicing Process. Order Preparation and servicing After placing the orders, the customers get into a single queue to pick up their orders. We used the process module here and used the resource type as a set of resources. This set contains two resourcesClerk1 and Clerk2. The selection rule used was cyclical. The process time follows a BETA distribution :10 + 61.5 * BETA(1.88, 2.6). Customers Seating We were told that 75% of the customers who come to the Starbucks at UC Satellite give orders "to go" and the remaining 25% stay and proceed to the seating area. As UH is known for its diversity in culture. Any customer that goes to the seating area will occupy a seat with any other customer. This process is beyond the overall objective of our study , but is essential in simulating the overall work flow. Figure 1 shows the final model that contains all above process. Time taken for this process is UNIF(30,60) minutes. Figure 4.1 shows the final model that contains all the above process.

Figure 4.1 A snap shot of the final model.

3.4 Model Result The main objective of this model was to measure and improve the performance of the Starbucks store at UC Satellite. The important statistics from the model were observed and recorded in table 1. Based on the result that we got from ARENA model, we simulated the total time for the two types of customers which are Dine in and To go and also simulated the utilization results for the clerks and the barista.

10

Table 4-1: Important measurements for the system Statistics # in (Range from Half width) #of Satisfied Customers (Range from Half width) # of Disappointed Customers(Range from Half width) Cycle time for Dine In Customers Cycle time for To Go Customers Average waiting time in the Billing queue Average waiting time in service queue + Service time Utilization of Barista Utilization of Clerk 1 Utilization of Clerk 2 Results 244 - 253 232-236 12-16 38 - 40 Minutes 4 - 4.5 minutes 2.5 Minutes 40 Seconds 0.9358 0.4118 0.4117

Number of replications required: The number of replications required is calculated using the formula ( ) . Where n is the initial number of replications taken. In this

simulation we have used 20 replications. H values are the initial half width and the required half width values for the simulation. The number replications comes to 72 to get a one percent confidence interval of the number of customers serviced . We used 75 replications in our simulation since it gives us a more accurate value. Warm Up period: We are simulating a particular time frame when there are a lot of customers in Starbucks. At 11 o clock there might be people already in the system and we used the output statistic to check the warm up period. The warm up period was set to ten minutes.

11

Observations from the result 1. There are too many disappointed customers which is bad for the store. 2. The cycle time for to go customer is high. 3. A customer spends a lot of time in the billing Queue 4. Clerks are not being utilized properly

4 Analysis of Results
4.1 Model Verifications At this stage the developed model should be checked at a very basic level in order to find whether the model functions as intended. In addition, the logical flow of the system has to be verified. In order to do so, first we ensure that it includes all the components that were specified earlier, and then check to see if it runs without any errors. For this step, we initially modeled the system and we worked just with customers and clerks. Development of the real system can be started as soon as the initial model works properly with no errors. In addition, we used animation and the output files. ARENA's debugging tools was used to check if model is running without errors. Moreover, we calculate the expected values for the utilization for each resource and the number of customers. Table 2 shows the comparison between these expected values vs. the simulated values.

12

Table 4-2: The comparison between the expected and simulated value Statistics Simulated Results Number in Number of Satisfied Customers Number of Disappointed Customers 244 - 253 232-236 12-16 Expected Results 247 232 Reasonable Reasonable
Since customers see a queue and exit we cannot calculate the expected value

Comment

Cycle time for Dine In Customers

38 - 40 minutes

46 minutes

Reasonable because the distribution values vary per replication

Cycle time for To Go Customers

4 - 4.5 minutes

2 Minutes

Reasonable because the distribution values vary per replication

Average waiting time in the Billing queue Average waiting time in service queue + Service time Utilization of Barista Utilization of Clerk 1 Utilization of Clerk 2

2.5 minutes 40 Seconds

Time spent in queue cannot be calculated Time spent in queue cannot be calculated

0.9358 0.4118 0.4117

0.917 0.410 0.410

Almost same Almost same Almost same

Note: Calculation sheet attached at the end of the report 3.

4.2 System Improvement Our main objective in the experimentation and analysis phase of our project is to minimize the system time and also the number of disappointed customers. The main factors that can help improve system time were found to be the number of baristas and clerks.

13

Recently the coffee shop manager hired one more employee to improve the coffee shop's performance. In order to find whether hiring this employee was the right thing to do, we conduct process analysis using the ARENA process analyzer which enable us to build various scenarios and see which of them has the most significant effect on the total time for customers. Table 4.2 show these scenarios associated with the number of disappointed customers and cycle times. It is very clear that the second alternative which is adding one more barista who is responsible for taking order is the most efficient alternative which will reduce the number of disappointed customer to zero and improve cycle time of "to go" customers. In the table below you can see that the third scenario is the best one since it reduces the billing queue and gets the cycle time for to go customers down to 2 minutes, but the resources are not being utilized properly .
Table 4.2 The different scenarios from ARENA process analyzer

EMBELLISHMENT MODEL

We noticed that the when the manager hired another employee, the utilization of his employees was very low. So we simulated an embellishment model which uses one of the clerks as a cross trained employee who can do the billing work if the other barista is busy and the queue is long. We used Barista and clerk as a set in the billing order giving preference to the Barista. The results for the embellishment model are as follows:

14

# of Disappointed Customers(Range from Half width) Cycle time for Dine In Customers Cycle time for To Go Customers Average waiting time in the Billing queue Average waiting time in service queue + Service time Utilization of Barista Utilization of Clerk 1 Utilization of Clerk 2

1-2 38 - 40 Minutes 2.5 minutes 35 - 40 Seconds 20 - 25 Seconds 0.67 0.60 0.73

Observations from the result 1. The disappointed customers have reduced drastically. 2. The cycle time for to go customer is reduced. 3. A customer now spends a reasonable time in the billing Queue 4. Clerks are being utilized properly

5 Conclusion and Recommendations


For work flow optimization, we need to choose an alternative which has minimum possible cost with less number of disappointed customers. When number of employees increases, the average system time will decrease. If we just focus on decreasing system time, we can implement Scenario (2) from our Pan analyzer outputs, in which we have two baristas and two clerks.

15

If we take cost into consideration then the embellishment model will save a lot of money for the store as one extra person will not be required. We also achieve better cycle times and less number of disappointed customers through this embellishment. At the end of this simulation study we have come up with two solutions out which Starbucks will decide which will be better for their operations. The first being the time saving solution where there are four employees and the second being the cost effective solution where are three cross trained employees. The team thinks that the embellishment model should be implemented as it saves the store a lot of money in the long run.

16

6 Appendix 6.1: Customers Arrival Time


Table 6-1The observed data during the peak time for arrival customers.

Final Fitting

17

6.2: order time (billing time)


Table 6-2 The observed data during the peak time for order time.

Final Fitting

18

6-3: service time


Table 6-3The observed data during the peak time for service time.

19

Final Fitting

20

You might also like