Category Archives: Scheduling

Scheduling Workers in Service Operations

Why Scheduling Is Important in Services

As discussed previously. one of the main distinctions between manufacturing and service operations is the customer’s direct interaction with the service delivery process. Because of this interaction, the determination of the proper number of workers to schedule at any particular  tme is critical to the success of every service operation. On one hand, scheduling too few workers results in unnecessarily long customer waiting times. On the other hand, scheduling too many workers results in overstaffing and the incurrence of unnecessarily high labor costs, which negatively affect profits. The service manager. consequently, needs to schedule workers in a way that effectively satisfies customer demand while minimizing  unnecessary labor costs.  The cost of labor in most services is a major cost component, often running 35 percent of sales and higher. For some services, in fact, virtually all of the direct costs can be considered as labor (examples of these types of services include consulting, legal work, home care nursing, and hair salons). Thus a small but unnecessary increase in labor can have a very significant impact on a firm’s profits.

A Framework for Scheduling Service Workers

Work schedules in service operations are usually developed on a weekly basis for several reasons. First, there are state and federal laws that specify the maximum number of hours and/or days an employee can work in a given week, after which overtime premiums must  paid. Second, the distinction between full-time and part-time workers is often made on he basis of the number of hours worked in a calendar week. Full-time versus part-time status often determines the ben fits paid by the employer. and may be related to union

contracts that specify the minimum number of hours workers in each category may work. Finally. many workers, especially hourly workers. are paid on a weekly basis that is often mandated by local or state laws.  The procedure for developing a schedule for service workers can be divided into the following four major elements, as illustrated in Exhibit 12.8: (a) forecasting customer demand, (b) converting customer demand into worker requirements, (c) converting worker requirements into daily work schedule, and (d) converting daily work schedules into weekly
work schedules.

Forecasting Demand

Since the delivery of most services takes place in the presence of the customer, the customer’s arrival rate directly correlates with the demand level for the service operation. For example, the customer must be present at a restaurant to partake in the meal being served; the patient must be present in the hospital to receive treatment. In addition to the customer’s presence at the point of service, the potential for high variability in the pattern of customer demand makes it extremely important for service managers to efficiently schedule workers. The first step, therefore, in developing a schedule that will permit the service operation to meet customer demand is to accurately forecast  heat demand. There are several patterns of demand that need to be considered: variation in demand within days (or even hours), variation across days of the week. variations within a month. and seasonal variations. Because demand is often highly variable throughout a day. forecasting within-day variation is usually done in either hour or half-hour increments. Today, with the use of computers and more sophisticated point-of-sale (POS) equipment. the ability  to record customer demand in even. shorter time increments is possible (for example. IS-minute time intervals). To develop a forecast we need to collect historical data about customer demand. The actual number of customers expecting service in a given time interval (that is. half hour or hour) is the preferred data. Fortunately, there is a wide range of POS equipment available that can capture this type of data, and, in many cases. even download it onto a computer for subsequent analysis.

Control in the Job Shop

 Scheduling job priorities is just one aspect of shop-floor control (now often referred to as production activity controls.  The American Production and Inventory Control Society (APICS) Dictionary defines a shop-floor control system as
shop-floor control The major functions of shop-floor control are 1. Assigning priority to each shop order.
2. Maintaining work-in-process ("IP) quantity information. Set of procedures for maintaining and communicating the
status of orders and work cell/en. A system for utilizing data from the shop floor as well as data processing files to maintain and communicate status information on shop orders and work centers.

3. Conveying shop-order status information to the office.
4. Providing actual output data for capacity control purposes.
5. Providing quantity by location by shop order for WIP inventory and accounting purpoes.
6. Providing measures of efficiency, utilization, and productivity of labor and machines. Exhibit 12.4 illustrates more of the details related to shop-floor control.

Tools of Shop-Floor Control

There are a variety of written forms that can help the supervisor maintain control in the job shop; these are easily generated by the appropriate software and updated frequently through
a normal interaction of supervisor and software.
• The dispatch list (usually generated on a daily basis) tells the shop foreman what jobs need to be accomplished that day, what priority each has. and how long each
will take. Exhibit 12.5A presents an example of a dispatch list.
• Exception reports provide the supervisor with the information needed to handle special cases and problems. An example of this is the anticipated delay report shown in Exhibit 12.5B. Typically made out once or twice a week. these reports are  reviewed to determine if any of the delays are serious enough to warrant revision of the master production schedule.
• The input/output control report. or simply the 110 report. iu ed by the -or to monitor the relationship between the workload and the station.  If these relationships are significantly out of balance, then the -supervisor can identify where adjustments are needed. An example of such a report is shown ill Exhibit 12.SC.

• Status reports give the supervisor summaries on the performance of the operation, and usually include the number and percentage of jobs completed on time, the lateness of jobs not yet completed, the volume of output. and so forth. Two examples  of status reports are the scrap report and the rework report.Input/output control is a major feature of a control system.control is that the total workload accepted (the input) should never exceed the capacity to perform job (the output). When the input exceeds the output. then backlogs occur. This has several negative consequences: Jobs are completed late. making customers unhappy, and sub: frequent or related jobs incur a delay before they can be started. This delay also result un: satisfied customers. Moreover, when jobs pileup at a work center. congestion occurs,  r sing becomes inefficient, and the flow of work to downstream work centers becomes radical. All analogy of this phenomenon to the flow of water is 1 in Exhibit 12.6. A -ripple but effective (antral device is the Gantt chart. It i~ used to help plan and jobs, again, using software, A Gantt chart is a type of bar chart that plots ta ks to bee against time. It also helps show relationships between jobs. Exhibit  shows a Igantt chart for a job shop attempting to complete three jobs (A, B. and C). This chart.

Shop-floor control systems in most modern plants are now computerized, with job status information entered directly into a computer as the job enters and leaves a work center. Some plants have gone heavily into bar coding and optical scanners to speed up the reporting process and to cut down on data-entry errors." As you might guess. the key problems  in shop-floor control are data inaccuracy and lack of timeliness. When these occur, the information fed back to the overall planning system is wrong and incorrect production decisions are made. Typical results are excess inventory and/or stockpot problems, missed  due dates, and inaccuracies in job costing.Of course, maintaining data integrity requires that a sound data-gathering system be in place; but more important, it requires adherence to the system by everybody interacting with it. Most firms recognize this, but maintaining what is variously referred to as shop  discipline, data integrity, or data responsibility is not always easy. And despite periodic drives to publicize the importance of careful shop-floor reporting by creating data-integrity task forces, inaccuracies still can creep into the system in many ways: A line worker drops a part under the workbench and pulls a replacement from stock without recording either .transaction. An inventory clerk makes an error in a cycle count. A manufacturing engineer fails to note a change in the routing of a part. A department supervisor decides to work jobs in a different order than specified in the dispatch list.

OPT is a software system that contains a proprietary algorithm for production scheduling. From OPT (optimized production technology) has evolved the managerial concept of theory of constraints lTOC). OPTffOC is a production planning and control (PPC) method that attempts to optimize scheduling by maximizing the utilization of the  bottleneck in the process. Traditional production planning and control is considered today to be no more than a production tool. Integrated PPC is a concept that marries an underlying philosophy of planning  tools to implement that philosophy to optimize the process. There are three major approaches to integrated PPC: push systems. pull systems, and bottleneck sy terns. The forerunner to the push systems is a tool developed by Joseph Orlicky at IB~t in I974-material requirements planning (MRP). In the late 196Os,Taichi

Ohno at Toyota developed the Kantian system, the first of the pull systems. Eli Goldratt's OPT, developed in the late 1970s, is considered the genesis of bottleneck systems." OPT distinguishes between bottlenecks and capacity constrained resources. A bottleneck applies to the case in which a stage or a number of stages in a system cannot process the good or service quickly enough to prevent backlogs (both in terms of work-in-progress and demand). A capacity-constrained resource (CCR) is a good or service necessary for the creation of the-final product that is exhausted before the final product is delivered. To il-: lustrate the difference between a bottleneck and a capacity-constrained resource, let's consider a retail laundry that specializes in cleaning shirts. Claude's Cleaners is noted for their quality and inexpensive service, so much so that there is always a line of customers waiting for the sole clerk, Clark, to process each transaction. Unfortunately, they are not noted for speedy service; it usually takes a week just to' get a shirt cleaned and pressed. Claude's has three presses, but they are used only three or four days each week because the starch for the shirts usually runs out. (In an effort to minimize costs, Claude's uses the just-in-time concept for materials management and it takes three days for the starch to be delivered after an order is placed.)
In this service process. Clark is a bottleneck. The starch is a capacity-constrained resource (CCR)-as its inventory is increased, the flow through the process component (the presses) is increased. The inventory of starch acts like a temporary bottleneck. The effects of CCRs typically can be reduced in the short term by relatively simple adjustments. Improvements to bottlenecks. on the other hand, are usually expensive and time-consuming. To optimize the flow through a bottleneck in the system, the bottleneck must operate continuously and at full capacity. A planning/control communication mechanism, known as Drum-Buffer-Rope, is used to accomplish this objective. Since the bottleneck is the slowest component of the process, it sets the pace or tempo for the system-much like a
drum beat sets the pace for a marching band. With the output of the process limited to the output of the bottleneck, decreases in output at the bottleneck cannot be recovered. Therefore. an inventory of goods or services, or "buffer inventory," is necessary before the bottleneck so it will always be operating at maximum capacity. To assure that the buffer is maintained at an optimal level (that is, just enough to keep the bottleneck operating), the rate at which the bottleneck is processing (the drum beat) must be communicated to the source of the goods or services that the bottleneck is processing. Since this communication is in one direction-from the bottleneck to the source of the goods or services (input)-and it pulls the input to the buffer, it is referred to as the "rope." Returning to Claude's Cleaners, Clark is the bottleneck or drum. The buffer would be the inventory of cleaned clothing waiting for the customer. What is the rope? Is there a rope?

The quantity of goods or services sent to the buffer is called the transfer batch. Because the purpose of the transfer batch is to maintain the buffer at its optimal level, the quantity will be dependent on the processing rate of the bottleneck. The quantity of goods  r services produced by the bottleneck's input source at one time is referred to as the process batch. In bottleneck systems it is critical to recognize that to optimize the entire system transfer batches and process batches may not be of the same quantity. At Claude's Cleaners. Bettie and Bert box the shirts after they have been pressed. If we consider Bettie and Bert together as a single input source for Clark. then the process batch size from boxing to the pick-up area is two. But. since Clark can only process one transaction  t a time the ansfer batch quantity pone.

Scheduling n Jobs on m Machines Complex Job Shops

Complex job shops are characterized by multiple machine centers processing a variety of  different jobs arriving at the machine centers in an intermittent fashion throughout the day. If there are n jobs to be processed on an machines and all jobs are processed on all machines. then there are (n !) alternative schedules for this job set. Because of the large number of schedules that exist for even small job shops, simulation is the only practical way to determi ne the relative merits of different priority rules in such situations. As in then job on one machine case, the 10 priority rules (and more) have been compared relative to their performance on the evaluation criteria previously mentioned.  which priority rule should be used? We believe that the needs of most manufacturers are reasonably satisfied by a relatively simple priority scheme that embodies the following principles:

1. It should be dynamic, that is, computed frequently during the course of a job to reflect changing conditions.
2.  It should be based in one way or another on slack time (the difference between the work remaining to be done on                the time remaining to do it in).

 

Scheduling n Jobs on Two Machines

The next step up in complexity of job shop types is the nl2 case, where two or more jobs must be processed on two machines in a common sequence. As in then] I case, there is an approach that leads to an optimal solution according to certain criteria. Also, as in the nil case, we assume it is a static scheduling situation. The objective of this approach, termed Johnson s rule not method (after its developer), is to minimize th.t flow time, from the be’ ginning of the  first job until the completion of the last. Johnson’s rule consists of the following steps:

1. List the operation time for each job on both machines.
2. Select the job with the shortest operation time.
3. If the shortest time is for the first machine, do that job first; if the shortest time is for the second machine, do that job last.
4. Repeat Steps 2 and 3 for each remaining job until the schedule is complete

solution procedures leading to optimality are not available. The reason for this is that even though the jobs may arrive in static fashion at the first machine, the scheduling problem becomes dynamic, and a series of waiting lines start to form in front of machines downstream.