Professional Documents
Culture Documents
Term Project - Call Center Operations: SOM 686, Fall 2006 Darren Mitchell Hayden Gilbert Serge Suprun
Term Project - Call Center Operations: SOM 686, Fall 2006 Darren Mitchell Hayden Gilbert Serge Suprun
SOM 686, Fall 2006 Darren Mitchell Hayden Gilbert Serge Suprun
Agenda
2
Overview Call Center Statistics (October), Definitions & Forecast Process View of the Call Center Call Center Competencies, Architecture, Three Key Measures & Littles Law Call Center Flow, Delays, Queues & Process Attributes Flow Rate-related Measures of Call Center Capacity Flow Time-related Measures of Customer Delay Inventory-related Measures of Customer Queues Performance Improvements & Managing Capacity
Overview
3
Time Warner Cable Los Angeles Bought Adelphia Exchanged properties with Comcast Now provides cable, internet & telephone services to over 2 million customers. Time Warner Advanced support is local Teamed up with outsourced partner to provide basic support for all three products. Basic support includes: billing, basic repair, changes to the accounts & general questions. Outsourced partner provides basic support from 3 call center locations: USA, Canada & Argentina.
Overview
4
Call Center in Argentina supports English & Spanish-speaking customers. There are 36 agents employed in the center 12 are Spanish only 24 are bilingual On average, the call center handles 40,000 calls per month. However, over last 2 months handled calls were 36.5% over forecast. Service level suffered over past 2 months too In October, only 32.4% of calls were answered within appropriate time limit.
DATE
FCAST
NCO
ABA
SVL%
ABA%
10/31/2006
1,263
41,168
2,237
54,494
77.1%
32.4%
238
11,118
46.5%
37.7%
10.6%
20.4%
156
241
73
74
447
462
NCO number of calls offered ABA number of calls abandoned; ABA% - percentage of calls abandoned SVL service level ASA average speed of answer Hand average handle time
Forecast
6
Past 2 months, Argentina used nave forecast: (Based on the last year historical data) Forecast didnt take into the account recent product changes & integration projects As a result October call volume was 32.4% over forecast. We are suggesting to use Moving average forecast.
Ex: MA103 = (A10+ A9+ A8)/3, using last months data is (54,494+56,221+50,221)/3=53,636 Therefore, F11=53,636. November forecast should be for 53,636 calls. This would allow Argentina to staff appropriately.
Inputs/Outputs customers calling for service/customers completed the call Flow Units customers Network of Activities and Buffers answering customers calls Resources customer service agents (CSA), phone automated system (PAS), etc. Information Structure account management system, reference materials, etc.
4 Dimensions for measuring the competence of the call center: Process Cost Process Flow Time Process Flexibility Process Quality. Argentina call center focuses on the low cost. Call center provides high-quality Spanish support. Argentina call center process architecture is defined by the types of resources (CSA, PAS, etc). Call center falls somewhere along the spectrum between two extremes (flow shop and job shop). Flow shop fits better, as call center uses specialized resources that perform limited tasks & produce large volumes with high precision and speed.
Flow Time = Time Customer spends: In automated phone system Waiting in the queue for an agent Talking to an agent. Ex: October Average PAS time is 46, Queue average waiting time is 32 & handle time is 462 seconds. T=540 seconds
Flow Rate = Number of customers that flow through a specific point in the call center process per unit of time. Inventory = Total number of customers present within call center boundaries. Ex: Argentina call center is 24/7 & call patterns are very different. Therefore, to simplify calculations we will use I=27 customers
Littles Law
10
Throughput = Average number of customers that flow through the call center per unit of time. Littles Law - Average inventory (=) Throughput (x) Average flow time. I=RxT
Number:
Time: In:
Ii
Ti
+
+
Ip
Tp
=I
=T
Inflow Rate Ri = Average rate of customer arrivals per unit time. In the Argentinean call center, Ri = 5 customers/minute Processing Time Tp = Average time required by agent to process the customer. Tp = 462 (agent) + 46 (PAS) = 508 seconds c = Number of agents in the resource pool c = 36 agents in Argentina Process capacity (Rp) = Total processing rate at which customers are processed by agents in the resource pool.
Rp = c/Tp or Rp = 36/8.47 = 4.25 customers/minute Buffer capacity (K) = Maximum number of customers that can wait in queue. K = 120 (there are 10 lines that can hold 12 each at any given time)
Throughput rate (R) = Average rate at which customers flow through the call center process R = min (Ri, Rp) In our case, Rp is smaller, so R = Rp = 4.25 Capacity utilization () = Average fraction of the resource pool capacity that is occupied in processing customers = R/Rp In our case, R = Rp and = 1
Our resource pool is constantly busy processing customers.
Safety capacity (Rs) = Excess processing capacity available to handle the customers inflows. Rs = Rp Ri In our case, Rs = 4.25 5 = -0.75
All the available capacity is busy processing arrivals.
Average waiting time (Ti) = Time that a customer spends in queue. Ti = 32 seconds. Average theoretical time = Average processing time of a customer. Tp = 462 seconds + 46 seconds (automated system) = 508 seconds Average flow time in the process (T) = Average time that a customer spends waiting in queue & being served T = Ti + Tp or T 508 + 32 = 540 seconds or 9 minutes Flow time efficiency = Proportion of time that a customer spends being served rather than waiting in queue Tp / T = .94
Average queue length = Average number of customers waiting for service Ii = R x Ti Ii = 4.25 x 0.53 = 2.25 customers waiting for service
Average number of customers in service = Average in-process inventory Ip = R x Tp Ip = 4.25 x 8.47 = 35.99 customers in service
Average total number of customers in the process I = I i + Ip I = 2.25 + 35.99 = 38.24 customers in the process
4.25 customers/min
Number:
Time: In:
2.25
508
+
+
35.99
32
= 38.24
= 540 seconds
Queue +
Service = Process
Performance Improvements
17
The following levers improved process performance: Decrease variability in customer interarrival & processing times. Decrease capacity utilization (or increase safety capacity) either by Decreasing the arrival rate or increasing the unit processing rate Increasing the number of servers Synchronize the available capacity with demand.
Managing Capacity
18
Capacity utilization ( = Ri/Rp ) can be reduced by increasing average processing rate (Rp) In order to increase processing rate (Rp = c/Tp) we recommended decreasing average processing time (Tp) To achieve a decrease in processing time: Identified that billing & escalated calls took longer to handle in this call center vs. similar centers. Thus recommended & implemented 2 separate hour training segments
Billing prorates explanation How to handle escalated calls
DATE
FCAST
NCO
% O/U FCAST
ABA
SVL%
ABA%
ASA
Hold
Hand
Processing time decreased to 410 seconds Tp = 410+46 = 456 seconds or 7.6 minutes Process capacity Rp = c/Tp or Rp = 36/456 = 4.74 customers/minute Result: Capacity Utilization = Ri/Rp or 5/4.74 = 1.06
20