The Wharton School Quarter II The University of Pennsylvania Fall 1999 OPIM 631 Operations Management: Quality and Productivity Note on Process Analysis Introduction 1 In this course, we look at organizations as processes. A process can, at the most aggregate level, be thought of as a "black box" that transforms inputs (raw materials, unserved customers) into outputs (finished goods, served customers). This transformation is accomplished through resources (machines, workers, capital). For example, each year The Wharton school transforms 6000 MBA applications into 780 graduates and 5220 rejected applications. The school's resources are its faculty and administration, the buildings, the coffee machines, etc. Whereas many other courses that you might take, including finance, economics, and strategy, can live with this aggregate view, the approach of operations management is to look inside the black box of this transformation process. Specifically, the objective of this course is to enable you to: 1. analyze existing business processes along performance dimensions outlined below 2. improve business processes to achieve higher profits. Thus, this course is about understanding and improving business processes. Start with a picture The most helpful tool in analyzing business processes is the process flow diagram. It is a graphical way to describe the process and will help you to structure the information that you find. As with any management or consulting project, you first need to focus on a part of the process that you want to analyze in greater detail, i.e. you need to define the process boundaries and an appropriate level of detail. Placement of the process boundaries will depend on the project you are working on. For example, in the operation of a hospital, one project concerned with patient waiting time might look at what happens to the patient before she sees a doctor (e.g. check-in, waiting time, encounter with the nurse). In this project, the encounter with the doctor would be outside the boundaries of the analysis. Another project related to the quality of surgery, however, might look at the encounter with the doctor in great detail, while either ignoring the admissions process or treating it with less detail. A process operates on flow units, which are the entities flowing through the process (e.g., people, materials, information). A process flow diagram is a collection of boxes, triangles, and arrows. Boxes stand for process steps (which can themselves be processes). 1 This note was prepared by Professors Terwiesch, Fisher, Ulrich, and Pearson as a first introduction to process analysis. Its purpose is to help students with no experience in process analysis prepare the first couple of cases in OPIM631. OPIM 631 Note on Process Analysis 1
Triangles represent waiting areas or buffers holding inventory. Arrows represent the route of flow of the flow units. If there are different flow units that take different routes through the process flow diagram, it can be helpful to use different colors for the different routes. Simple Example Assume your learning team needs to produce 100 bagel sandwiches with ham, cheese, and veggies on them. After some experimentation, you find out that the tasks involved take approximately the following durations: - Cutting (0.5 minutes) - Spreading mayonnaise (2.5 minutes) - Cutting vegetables (3 minutes) - Putting vegetables on bagel (2 minutes) - Dressing (1 minute) - Wrapping (1 minute) You decide that each of the three of you (when it came to making bagels, the other students on your team remembered that they had to prepare for a finance exam) will take two tasks. Your process now looks like the process described in Figure 1. The triangles between process steps indicate that there may be an accumulation of incomplete sandwiches at those locations. All of the flow units within the process boundaries are called work-in-process inventory or WIP (pronounced "whip"). Raw Bagels Finished Bagels Cut/mayo Veggies Dressing/ Wrap 1 bagel / 3 min. 1 bagel / 5 min. Figure 1: Simple bagel process 1 bagel / 2 min. After operating for one hour, you realize that you are not too happy with this process (except the person doing the dressing and the wrapping, who looks the least tired of the three of you). Here is the current state of your system (kitchen): Raw Bagels Huge pile nothing 10 Bagels Cut/mayo Veggies Dressing/ Wrap 100% busy 100% busy, got yelled at Figure 2: Your kitchen after one hour of work 40% busy, was the one who yelled The first thing you notice after looking up from your work is a huge pile of bagels waiting for veggies. The "cut and mayonnaise person" has clearly worked hard and managed to prepare a total of 20 bagels in the one hour. From those 20, only 12 were picked up by the "veggie person". OPIM 631 Note on Process Analysis 2
This leads to the second issue, which is the frustration of the veggie person. Despite doing his best, he never could keep up with the speed of the dress-and-wrapper. This resulted in no WIP between the two and in under-utilization of the "dress-and-wrapper" person (which gave her plenty of time to shout for more bagels). Third, and most importantly, all three of you are somewhat disappointed when you realize that you have only finished 11 bagels. Your preliminary analysis prior to starting went like this: it takes 10 minutes of work to finish one bagel; if three persons work for 60 minutes, this should result in 3*60/10=18 bagels. However, only 11 bagels are finished. So you fall short by almost 50%! After some discussion, you realize what happened: - It took ten minutes to "fill the pipe line" in your process, thus the first bagel was done after 10 minutes. - From minute 10 onwards, you finished one bagel every 5 minutes, despite the hard work of the cut-and-mayo person. But even then, why did the hard work not lead to more output? Was it really just a startup problem? Or, was there something more fundamental at work? The Basic Measures of Process Performance In process analysis, we focus on three fundamental performance measures: - The number of flow units contained within the process is called the inventory (I) (or WIP). Assuming we define the process boundary just before cutting and just after wrapping, this inventory includes bagels currently being worked on by any of the three of you and bagels between operations. - The time it takes a flow unit to get through the process is called the Flow time (T). An interesting question to ask is "how long does it take one bagel to move from the beginning to the end of the process?" Although this question is somewhat hypothetical in the present example, it would be an important variable if you were selling bagels made to order. - Finally, the most important measure is the rate (measured in [flow units/time]) at which the process is delivering output, which we will call the Flow rate (R). R is sometimes referred to as throughput rate. The maximum rate with which the process can generate output is also called the capacity of the process. You might be somewhat irritated that we have not talked about cost so far. However, note that any improvement in inventory, flow rate, or flow time will have a direct impact on cost, or even better, on profit. Shorter flow times will make it easier to rapidly respond to customers (especially in make-to-order environments and service operations). Typically, shorter flow time will result in additional unit sales and/or higher prices. Lower inventory results in lower working capital requirements as well as many quality advantages that we will explore in this course. Higher inventory is also directly related to longer flow times (explained below). Thus a reduction in inventory also yields a reduction in flow time. Higher flow rate translates directly into more revenues, assuming your process is OPIM 631 Note on Process Analysis 3
currently capacity constrained, i.e. there is sufficient demand that you could sell any additional output you make. Process Analysis After building the process flow diagram, the next step towards understanding and improving a business process is to perform a process analysis. The objective of the process analysis is to: Find the process step that is limiting the rate at which the process generates output. This limiting step is called the bottleneck. Find the maximum rate at which the process can generate output (capacity). If there is sufficient demand, the capacity of the process will correspond to the flow rate defined above. Compute the time it takes for a flow unit to go through the process, the flow time, including processing time and waiting time. Compute the time it takes to fulfill an order of a given size, e.g. 100 bagels. There is no precise recipe 2 for how to draw process flow diagrams and how to perform a process analysis. You will learn how to perform these tasks over the next two or three weeks as you prepare the cases for class. Figure 3 is a summary of the major steps. 1. Choose the process boundaries and the flow unit. 2. Understand how the physical process works and draw a process flow diagram. Show process steps, inventory holding points, and arrows depicting product flow. 3. Determine the capacity of each step in the process expressed as the number of flow units of product that can be processed per unit time. 4. Identify the capacity bottleneck. This is the step with least capacity. 5. Once the bottleneck is identified, think about how the bottleneck influences other process steps as well as the overall behavior of the process. Calculate different performance measures such as the process capacity, flow time, work in process inventory, and labor utilization. 6. Consider changes to improve system performance. Figure 3: Six steps for process analysis 2 Actually there are recipes for process analysis. However, in order to be applicable to all situations, the recipe is rather complex and involves many, many definitions. Thus, we prefer to give you a somewhat simplified version that will work in 80% of the cases and rely on your common sense for performing the residual 20%. OPIM 631 Note on Process Analysis 4
How to determine capacity? As discussed above, each process step in a process flow diagram can be thought of a process in itself. Therefore, the concept of capacity applies to both an individual process step and to an entire process. Let's revisit the bagel example. Remember that the first step (cutting and mayo spreading) took 3 minutes per bagel, the second step (cutting veggies and putting them on the bagel) took 5 minutes and the third step (dressing and wrapping) took 2 minutes. The duration of the activities that comprise a process step is called the activity time. To determine the capacity of an individual process step, we write: capacity=1/activity-time. E.g. for the first step, we write: capacity 1 =1/(3 minutes/bagel)=0.333 bagels/minute, which we can rewrite as 0.333 bagels/minute = 0.333 bagels/minute * 60 minutes/hour = 20 bagels/hour. (These computations using measurement units might remind you of your high school science class, and YES, your physics teacher was right after all: keep track of the measurement units!) Similarly, we can compute capacities of the second step to be 12 bagels/hour and of the third step to be 30 bagels/hour. If there is more than one person (or machine) carrying out a process step, the above formula changes to: capacity=number-of-workers/activity-time. This is intuitive, as the capacity grows proportionally with the number of workers. In simple processes with just one type of product (flow unit), we call the process step with the least capacity the bottleneck. The capacity of the overall process is equal to the capacity of the bottleneck. In the bagel example, this is the veggie process step, thus the overall process capacity is 12 bagels/hour. In processes with multiple product types, the analysis is a little more complicated. Why can't we just do the same analysis as above? Consider the process flow diagram shown in Figure 4, which describes a process where multiple variants of bagels get produced; e.g. cream cheese, veggie bagels and bagels with grilled bacon and veggies. This product mix complicates the process analysis. It is important to understand that the capacity of the process crucially depends on the product mix. For example, the process step "cream cheese" might have a very long activity time, resulting in a low capacity of this activity. However, if only one out of a hundred customers requires cream cheese, this low capacity would not be a problem. Thus, to find the bottleneck and to determine capacity in a multi-product situation we need to compare the available capacity with the requested capacity. The analysis is given in Table 1. Numbers that are raw data (i.e. that you would find in the case or by observing the real process) are printed in bold. Numbers that are derived by analysis are printed in italics. We assume the demand is 180 bagels a day, of which there are 30 OPIM 631 Note on Process Analysis 5
grilled-veggie, 110 veggie only, and 40 cream cheese. Assuming that the working day is 10 hours, demand is 3 grilled-veggie bagels/hour, 11 veggie bagels/hour, and 7 cream cheese bagels/hour. Raw Bagels Put Grilled Stuff on B. Grilled-Veggie Veggie Cream Cheese Cut Veggies on Bagel Wrap Finished Bagels Cream Cheese Figure 4: Process analysis for three bagel types Cut Grilled Stuff Veggies Cream Cheese Wrap Activity time 3 [min/bagel] 10 [min/bagel] 5 [min/bagel] 4 [min/bagel] 2 [min/bagel] Available capacity 1/3 [bagel/min] =20 [bagel/hour] 1/10 [bagel/min] =6 [bagel/hour] 1/5 [bagel/min] =12 [bagel/hour] 1/4 [bagel/min] =15 [bagel/hour] 1/2 [bagel/min] =30 [bagel/hour] Requested capacity grilled -veggie + veggie + cream cheese 3 [bagel/hour] 11 [bagel/hour] 4 [bagel/hour] 3 [bagel/hour] 0 0 3 [bagel/hour] 11 [bagel/hour] 0 0 0 4 [bagel/hour] 3 [bagel/hour] 11 [bagel/hour] 4 [bagel/hour] = total 18 [bagel/hour] 3 [bagel/hour] 14 [bagel/hour] 4 [bagel/hour] 18 [bagel/hour] Implied Utilization = requested Cap/ available Cap 18/20=90% 3/6=50% 14/12=117% 4/15=27% 18/30=60% Table 1: Finding the bottleneck in the multi product case When computing the requested capacity, it is important to remember that some activities (e.g. cutting) will be requested by all product types, whereas others (e.g. grilled stuff) will only be requested by one product type. This will (hopefully) become clear by looking at the process flow diagram. By comparing the ratio of requested capacity and available capacity, which is also called the implied utilization of the activity, we can now find the "busiest" activity, in this case the veggie operation. As this ratio is above 100%, the process is capacity constrained and, unless we can work overtime (i.e. add extra hours at the end of the day, in which case our available capacity would go up), we will not be able to meet demand. Little s Law Flow Time, Flow Rate, and Inventory are related by the following identity (known as Little s Law): I = R * T. OPIM 631 Note on Process Analysis 6
You might think this relationship is trivial, however, it is not and its proof is rather complex for the general case (which includes stochastic variability) and by mathematical standards is very recent. Little's Law is useful in finding any third variable when the other two are known. For example, if you want to find out how long patients - on average - spend in the waiting room for a certain hospital operation, e.g. X-ray, you could do the following: - observe the queue at a couple of random points during the day, giving you an average inventory I. Let's say this number is 7 patients. - look in your computer to see how many patients came through X-ray on that day, giving you the average flow rate R. Let's say there were 100 patients over a period of 8 hours, yielding R=100/8=12.5 patients/hour - use Little's Law to compute T=I/R=7/12.5=0.56 hours=33.6 minutes This formula will be helpful in computing T in many applications you will encounter. Obviously, if you see how to compute T directly, then it s easier not to use the formula. Little's Law can also be used to find I given R and T or to find R given I and T. When does Little's Law hold? The short answer is always. Little's Law does not depend in which sequence the flow units (e.g. patients) are served 3 (remember FIFO and LIFO from your accounting class?). The only caveat is that if I, R or T vary over time, then Little s Law is still valid, but only if used with average values of I, R and T. Total Time x to process a given quantity of work Q Remember the disappointment when finding out that you had produced only 11 bagels when operating in the process of Figure 1? How could we have computed this before starting the work? We now understand that the veggie process step was the bottleneck, because it has the least capacity. As we tried to push as many bagels through the system as possible, we were capacity constrained and the flow rate of the system, once it got going, was equal to the capacity: 12 bagels/hour. So how long does it take to produce ten bagels, starting with an empty system? It will take ten minutes (the sum of the three activity times) until the first bagel is completed. This is the time to "fill the pipeline". From then onwards, we get an additional bagel every five minutes. (This is because the Veggie operation can produce no more than one bagel every five minutes.) Thus bagel 1 is produced after 10 minutes, bagel 2 is produced after 15 minutes and bagel 10 is produced after 55 minutes. More formally, we can write the following formula. The time x to finish Q units is: x = T + (Q-1)/R 3 Note however, that changing the sequence will impact a given flow unit (e.g. the patient coming in first in the morning). But Little s Law deals with averages, i.e. if patient A will wait longer, one or more other patients will have less waiting time. OPIM 631 Note on Process Analysis 7
For a continuous flow process, this time is: x = T + Q/R. Types of Processes The decision to arrange your process as depicted in Figure 1 was not dictated by the product (the recipe of the ham-cheese bagel) but it was your managerial choice. Consider the two alternative process layouts in Figure 5. Conveyor Belt Raw Bagels Finished Bagels Cut/mayo Veggies Wrap Machine-Paced Line Flow Raw Bagels Cut/mayo Veggies Wrap Cut/mayo Veggies Wrap Finished Bagels Cut/mayo Veggies Wrap Same Person Three Parallel Work Cells Figure 5: Two alternative process lay-outs The first process is called a machine-paced line flow. A single conveyor belt carries the bagels between workers and moves at a constant speed. Such a process is different from the example we had in Figure 1, called a worker-paced line flow, as it does not allow for a build up of inventory between the process steps. The pace with which the workers must complete activities is dictated by the speed of the conveyor belt. The second process corresponds to three work cells. In this example, each work cell consists of one person who completes the entire set of activities to produce a completed unit. As a result of this process layout, there is no need for inventory between the activities. Let's go and see some real organizations in action. Make sure to visit the following webpages, each of which offer a virtual tour through their (manufacturing) process: Buell Motorcycles: http://www.buell.com/tour/factour.html Monitor Sugar: http://www.monitorsugar.com/htmtext/pictflow.htm Peavey Guitars: http://www.peavey.com/wolfgang/index.html Statton Furniture: http://www.statton.com/tourpics.htm OPIM 631 Note on Process Analysis 8
How are the processes different from each other? Some produce a large variety of products (e.g. Statton furniture), while others basically produce one single product (Monitor). Some processes are highly automated, while others are largely manual. Some processes resemble the legendary Ford assembly line, while others resemble more the workshop in your local bike store There are many other dimensions along which processes can differ. Empirical research in operations management, which has looked at thousands of processes, has identified five clusters or types of processes. Within each o the five clusters, processes are very similar concerning variables such as product variety or production volume, as is described in Table 2. Job Shop Batch process Apparel sewing Bakery Semiconductor wafers Auto assembly Computer assembly Large Auto assembly Workerpaced line flow Machinepaced line flow Continuous Process Examples Design company Consulting Paper mill Oil refinery Food processing Product variety ( variants) High (100+) Medium (10-100) Medium (1-50) Product volume (units / year) Low (1-100) Medium (100-100k) High (10k-1M) 1-10 High (10k 1M) Low (1-10) Very High Specificity of Process Technologies General purpose General purpose Specialized Specialized Highly specialized, custom designed Process Flow / Layout Similar process equipment grouped together. Different types of jobs proceed along different paths through facility Similar process equipment grouped together. Most batches follow similar path Equipment laid out in order of process steps. Linear flow, identical across products. Similar to worker paced line flow. However, the speed of the operation is dictated by an automated line. Straight line, direct connection between process steps Table 2: Process types and their characteristics Identifying the process type is more than an academic exercise in definitions. Specifically, there are two benefits to this exercise: 1. When creating the process flow diagram, you will know whether your diagram will look more like Figure 1 or one of the two cases described in Figure 5. 2. Similar process types tend to have similar problems. For example, as in the bagel process described in Figure 1, worker paced lines tend to have the problem of work-balancing and inventory build-up. Job-shops tend to suffer from long flow times. Thus, once you know the process type, you can determine fairly quickly the important issues in the case (or what problem your consulting customer is facing). OPIM 631 Note on Process Analysis 9
We will discuss these issues further in class, when we talk about the virtual plant tours. How to Improve a Process? Now that we have all the tools available to understand what is going on in a process, the logical question becomes "How can we improve the performance of a process?" Again, there is no single recipe, but at least a short list of generic answers: Add capacity to the bottleneck. This is an improvement if the value of the extra capacity exceeds the cost of the extra capacity. For example, if you could get one additional person to help out on the veggies, it would improve the flow rate through your process. Improve balance by moving work from the bottleneck activity to a non bottleneck activity. For example, your process in Figure 1 would improve if the person who is in charge of the dressing / wrapping would take over the task of putting the vegetables on the bagel. This would bring her activity time up to 4 minutes/bagel (and thus her capacity up to 15 bagels / hour), but the veggie activity would go to 3 minutes/bagel (capacity of 20 bagels/hour). As the process capacity is determined by the activity with the least capacity, the process capacity would move from 12 to 15 bagels/hour. Determine the best span of control for a worker; e.g. do we have a cell with a single worker who builds the entire product or do we have an assembly line on which each worker performs a narrow, short task. Assuming that there is no benefit of specializing in one activity (thus the times for performing each task would remain unchanged if a worker carries out several activities), the work cell layout in Figure 3 is extremely attractive. The activity time of completing one bagel is 10 minutes/bagel. Thus an individual can produce 6 bagels/hour. Multiplied by 3 workers, this would give 18 bagels/hour. This improvement arises because no worker is forced to be idle. If a process is demand rather than capacity constrained, stimulate demand by offering additional value to the customer; e.g. higher quality, more product variety or shorter lead time. OPIM 631 Note on Process Analysis 10