Capacity Management in Services Module Why do queues build up? Process attributes and Performance measures of queuing processes Safety Capacity Its effect on customer service Pooling of capacity Queuing Processes with Limited Buffer Optimal investment Specialists versus generalists Managing Customer Service SofOptics Service Operations Slide 1 Telemarketing at During some half hours, 80% of calls dialed received a busy signal. Customers getting through had to wait on average 10 minutes for an available agent. Extra telephone expense per day for waiting was $25,000. For calls abandoned because of long delays, L.L.Bean still paid for the queue time connect charges. L.L.Bean conservatively estimated that it lost $10 million of profit because of suboptimal allocation of telemarketing resources. Service Operations Slide 2 Some Questions to discuss: Why did they loose money? What are the performance measures for a call center? How model this as a process? What decisions must managers make? Service Operations Slide 3 Telemarketing: deterministic analysis it takes 8 minutes to serve a customer 6 customers call per hour – one customer every 10 minutes Flow Time = 8 min – same for every customer – histogram: → Flow Time Histogram 100% Probability 80% 60% 40% 20% Flow Time (minutes) Service Operations Slide 4 195 180 165 150 135 120 105 90 75 60 45 30 15 8 0 0% Telemarketing with variability in arrival times + activity times 0% 25% 90% 80% 70% 20% 60% 50% 40% 15% 10% 30% Service Operations Slide 5 190 More Flow Time (minutes) 170 180 140 150 160 0% 120 130 0% 90 100 110 20% 10% 70 80 5% 40 50 60 exhibit variability 100% 0 10 – 30% 20 30 In reality service times Probability Cumulative Probability Flow Time (minutes) 190 0% More 20% 170 180 5% 140 150 160 40% 120 130 10% 90 100 110 60% 70 80 15% 0 10 – exhibit variability 20% 40 50 60 In reality inter-arrival times 100% 90% 80% 20 30 Probability 25% Cumulative Probability Flow Time Histogram Telemarketing with variability: The effect of utilization Average service time = – 9 minutes 8% 100% 7% 90% Probability 80% 6% 70% 5% 60% 4% 50% 3% 40% 30% 2% 20% 190 180 170 160 150 140 130 120 110 90 100 80 70 60 50 40 30 20 Flow Time 9.5 minutes More – 0% 10 Average service time = 10% 0% 0 1% 25% 100% 90% Probability 20% 80% 70% 15% 60% 50% 10% 40% 30% 5% 20% 10% Service Operations Slide 6 190 180 170 160 150 More Flow Time 140 130 120 110 100 90 80 70 60 50 40 30 20 10 0% 0 0% Why do queues form? Call # 1. variability: – arrival times – service times – processor availability 10 9 8 7 6 5 4 3 2 1 0 0 20 40 Role of utilization: – Impact of variability increases as utilization increases! (arrival throughput or capacity ) 60 80 100 TIME Inventory (# of calls in system) 5 4 3 2 1 0 0 20 40 60 TIME Service Operations Slide 7 80 100 Flow Times in White Collar Processes Industry Process Average Flow Time Theoretical Flow Time Flow Time Efficiency Life Insurance New Policy Application 72 hrs. 7 min. 0.16% Consumer Packaging New Graphic Design Consumer Loan 18 days 2 hrs. 0.14% 24 hrs. 34 min. 2.36% Hospital Patient Billing 10 days 3 hrs. 3.75% Automobile Manufacture Financial Closing 11 days 5 hrs 5.60% Commercial Bank Service Operations Slide 8 Queuing Systems to model Service Processes: A Simple Process Order Queue “buffer” size K Sales Reps processing calls Incoming calls Calls on Hold MBPF Inc. Call Center Blocked Calls Abandoned Calls (Busy signal) (Tired of waiting) Service Operations Slide 9 Answered Calls What to manage in such a process? Inputs – InterArrival times/distribution – Service times/distribution System structure – Number of servers – Number of queues – Maximum queue length/buffer size Operating control policies – Queue discipline, priorities Service Operations Slide 10 Performance Measures Sales – Throughput R – Abandonment Ra Cost – Server utilization r – Inventory/WIP : # in queue Ii /system I Customer service – Waiting/Flow Time: time spent in queue Ti /system T – Probability of blocking Rb Service Operations Slide 11 The drivers of waiting: How reduce waiting? Queuing theory shows that waiting increases with: – variability Arrival times Service times Average Wait Time – length of avg. service time Variability – Arrival throughput Nonlinearly: “it blows up!” 100% Utilization Hence: reduce waiting by: Process Capacity – Reduction of variability – Reduction of arrivals/throughput – Add “safety” capacity Reduce length of service Increase staffing Service Operations Slide 12 Levers to reduce waiting and increase QoS: variability reduction + safety capacity How reduce system variability? Safety Capacity = capacity carried in excess of expected demand to cover for system variability – it provides a safety net against higher than expected arrivals or services and reduces waiting time Service Operations Slide 13 Example 1: MBPF Calling Center with one server, unlimited buffer. The basics of QoS Consider MBPF Inc. that has a customer service representative (CSR) taking calls. When the CSR is busy, the caller is put on hold. The calls are taken in the order received. Assume that calls arrive exponentially at the rate of one every 3 minutes. The CSR takes on average 2.5 minutes to complete the reservation. The time for service is also assumed to be exponentially distributed. The CSR is paid $20 per hour. It has been estimated that each minute that a customer spends in queue costs MBPF $2 due to customer dissatisfaction and loss of future business. – Holding cost H = – Average number waiting in buffer Ii = – MBPF’s waiting cost = H Ii = Service Operations Slide 14 Example 2: MBPF Calling Center with limited buffer size. Impact of blocking In reality only a limited number of people can be put on hold (this depends on the phone system in place) after which a caller receives busy signal. Assume that at most 5 people can be put on hold. Any caller receiving a busy signal simply calls a competitor resulting in a loss of $100 in revenue. – # of servers c = – buffer size K = What is the hourly loss because of callers not being able to get through? Service Operations Slide 15 THE BAT Case = Managing the operations of a customer service department Putting Tech Support on The Fast Track Handouts to be distributed in class Service Operations Slide 16 Example 3: MBPF Calling Center with 1 or 2 queues. Impact of Resource Pooling 2 phone numbers – MBPF hires a second CSR who is assigned a new telephone number. Customers are now free to call 50% either of the two numbers. Once they are put on hold customers tend to stay on line since the other may be Ri = 1/3min worse.. Tp = 2.5min Queue Server Tp = 2.5min 50% Queue Server 1 phone number: pooling – both CSRs share the same telephone number and the customers on hold are in a single queue Tp = 2.5min Ri = 1/3min Which system is “better?” – – – In which sense? When? Why? Queue Servers Service Operations Slide 17 Example 4: MBPF Calling Center with 2 service tasks. The impact of process structure & resource capabilities: Specialization Vs. Flexibility A second service task is added. Two possibilities to structure the process: Tp = 2.5min Specialization Tp = 2.5min Ri = 1/3min – Each service task is performed by a specialized agent – Average flow time T = Queue Server Tp = 5min Flexibility – The entire service is performed by one of two flexible agents = generalists. – Agerage flow time T = Ri = 1/3min Queue Servers Which system is “better?” – In which sense? – When? – Why? Service Operations Slide 18 Queue Server Increase quality of service: 1. reduce variability Two types of variability: – Predictable – Unpredictable = “Stochastic” Two sources of variability: – Arrivals – Length of service Predictable variability is reduced by: – Proper triage: differentiated treatment – Proper scheduling & appointments – Standardization of service (not always an option) Key = synchronize arrivals with end of service Service Operations Slide 19 How increase quality of service with stochastic variability 2. reducing utilization is your only option How reduce utilization? 1. Reduce throughput – Not typically desired b/c of social, ethical, or financial concerns … 2. Increase capacity Recall section 3!: bottleneck management Key: when one cannot perfectly synchronize flows so that there is remaining, irreducible stochastic process variability then one must build in a capacity cushion. One cannot provide high quality of service at high utilization Service Operations Slide 20 Increase quality of service: anticipate predictable variability + build safety-capacity for stochastic variability. e.g. smart staffing – Average walk-ins often are fairly predictable Keep data (use IT!): find average trend (predictable) + stochastic variations Staff accordingly: use time-buckets + build safety-capacity staffing 16 Number of patients 12 +1 s 8 Maximum # Patient arrivals/hr Average # Patient arrivals/hr 4 -1 s 0 5AM Service Operations 11AM 5PM Time of day Slide 21 11PM 4AM Source: McKinsey Quarterly 2001 Smart Staffing/Capacity Management at Sof-Optics 100 90 80 70 60 50 40 Demand (# Calls/30min) 30 20 Current Supply/Capacity (# Calls/30min) 10 Optimized Supply w/o demand mgt or capital investment Service Operations Slide 22 17:30 16:30 15:30 14:30 13:30 12:30 11:30 10:30 9:30 8:30 7:30 6:30 0 Call Centers In U.S.: $10B, > 70,000 centers, > 3M people (>3% of workforce) Most cost-effective channel to serve customers Strategic Alignment – accounting: 90% are cost centers, 10% are revenue centers – role: 60% are viewed as cost, 40% as revenue generators – staffing: 60% are generalists, 40% specialists – Trend: more towards profit centers & revenue generators Trade-off: low cost (service) vs. high revenue (sales) Source: O. Zeynep Aksin 1997 Service Operations Slide 23 Framework for Analysis and Improvement of Service Systems Divide day into blocks based on arrival rates: – Separate “peaks” from “valleys” For each block evaluate performance measures given current staffing Quantify financial impact of each action – Workforce training: reduces mean and variability of service time – Work flexibility from workforce: pools available capacity – Time flexibility from workforce: better synchronization Supply – Retain experienced employees: increased safety capacity mgt – Additional workforce: Increases safety capacity – Improved Scheduling: better synchronization – Incentives to affect arrival patterns: better synchronization Reservation mgt, pre-sell, Disney’s FastPass – Decrease product variety: reduces variability of service time – Increase maximum queue capacity – Consignment program, fax, e-mail etc. Service Operations Slide 24 Demand mgt How do these insights related to our earlier “Levers for Reducing Flow Time?” “is to decrease the work content of (only ?) critical activities”, and/or move it to non critical activities. Reduce waiting time: – reduce variability arrivals & service requests synchronize flows within the process – increase safety capacity lower utilization Pooling – Match resource availability with flows in and out of process Service Operations Slide 25 Learning objectives: General Service Process Management Queues build up due to variability. Reducing variability improves performance. If service cannot be provided from stock, safety capacity must be provided to cover for variability. – Tradeoff is between cost of waiting, lost sales, and cost of capacity. Improving Performance – Reduce variability – Increase safety capacity Pooling servers/capacity – Increase synchronization between demand (arrivals) and service Service Operations Manage demand Synchronize supply: resource availability Slide 26