Post on 16-Jan-2016
DataMOCCAData MOdels for Call Center Analysis
Project Partners:
Technion: Paul Feigin, Avi Mandelbaum, Sergey Zeltyn (IBM Research)
SEElab: Valery Trofimov, Ella Nadjharov, Igor Gavako
Students, RA’s
Wharton: Larry Brown, Noah Gans, Haipeng Shen (N. Carolina), Linda Zhao
Students, Wharton Financial Institutions Center
Companies: US Bank, IL Telecom (both Aspect-Based), IL Bank (Genesys), …
2
Project DataMOCCA
Goal: Designing and Implementing a
(universal) data-base/data-repository and interface
for storing, retrieving, analyzing and displaying
Call-by-Call-based data/information
Enables the Study of:
- Customers
- Service Providers / Agents
- Managers / System
Wait Time, Abandonment, Retrials
Loads, Queue Lengths, Trends
Service Duration, Activity Profile
3
DataMOCCA History: The Data Challenge
- Queueing Research lead to Service Operations (early 90’s)
- Services started with Call Centers which, in turn, created data-needs
- Queueing Theory had to expand to Queueing Science: Fascinating
- WFM was Erlang-C based, but customers abandon! (Im)Patience?
- (Im)Patience censored hence Call-by-Call data required: 4-5 years saga
- Finally Data: a small call center in a small IL bank (15 agents, 4 services)
- Technion Stat Lab, guided by Queueing Science: Descriptive Analysis
- Building blocks (Arrivals, Services, (Im)Patiece): even more Fascinating
4
DataMOCCA History: Research & Teaching
- Large well-run call centers beyond conventional Queueing Theory:
- Both Quality and Efficiency Driven (vs. tradeoff)
- Multi-Disciplinary view: OR/OM, HRM (Psychology), Marketing, MIS
- Research: Asymptotic analysis of the Palm/Erlang-A model, in the
Halfin-Whitt regime = QED Regime
- Teaching: Service Management + Industrial Engineering =
Service Engineering / Science
- Wharton: Mini-course (OPIM, Morris Cohen) + Seminar (Statistics, Larry
Brown + Call Center Forum (Noah Gans) = cooperation with a large(r)
banking call center (1000 agents), …
5
DataMOCCA: Present
System Components:
1. Clean databases: operational histories of
individual calls, agents and sometimes customers (ID’s).
2. SEEStat (friendly powerful interface): online access to (mostly)
operationa and (some) administrative data; flexible customization,
statistics
Currently Three Databases:
- US Bank (220/40M calls, 1000 agents, 2.5 years;7-48GB; 3GB DVD).
- Israeli Telecom (800 agents, 3.5 years; 55GB; ongoing).
- Israeli Bank (500 agents, 1.5 years; ongoing).
Primary arrivals to queue Retail August 2003
050
100150200250300350400450
07:00 09:00 11:00 13:00 15:00 17:00 19:00 21:00 23:00
Time (Resolution 5 min.)
numb
er o
f ca
ses
04.08.2003 06.08.2003
Customer service time Retail Total August 2003
0.0
0.5
1.0
1.5
2.0
2.5
0 15 30 45 60 75 90 105 120
Time (Resolution 5 sec.)
Rela
tive
fre
quen
cies
, %
04.08.2003 06.08.2003
Customers in queue(average) Retail August 2003
0
20
40
60
80
100
120
140
07:00 09:00 11:00 13:00 15:00 17:00 19:00 21:00 23:00
Time (Resolution 5 min.)
numb
er o
f ca
ses
04.08.2003 06.08.2003
6
Agents(CSRs)
Back-Office
Experts
VIP(Training)
Arrivals(Business Frontier
of the 21th Century)
Redial(Retrial)
Busy(Rare)
Goodor
Bad
Positive: Repeat BusinessNegative: New Complaint
Lost Calls
Abandonment
Agents
ServiceCompletion
ForecastingStatistics,
HumanResourceManagement
)HRM(
New Services Design (R&D)
Operations,Marketing,
MIS
Organization Design:Parallel (Flat)Sequential (Hierarchical)Sociology, Psychology,
Operations Research
HRM
Service Process Design
Quality
Efficiency Skill Based Routing (SBR) DesignMarketing, HRM,Operations Research,
MISCustomers Interface Design
Computer-Telephony Integration - CTIMIS/CS
Marketing
(Turnover up to200% per Year)
(Sweat Shops of the
21th Century)
Operations/BusinessProcessArchive
DatabaseDesignData Mining:
MIS, Statistics, Operations Research, Marketing
InternetChatEmailFax
Lost Calls
Service Completion(75% in Banks)
(Waiting Time Return Time)
Logistics
Customers Segmentation - CRM
Psychology, Operations Research,
Marketing
Expect 3 minWilling 8 minPerceive 15 min
PsychologicalProcessArchive
Psychology,Statistics
TrainingJob Enrichment
Marketing,Operations Research
Human Factors Engineering
VRU/IVR
Queue(Invisible)
VIP Queue
(If Required 15 min,then Waited 8 min)
(If Required 6 min ,then Waited 8 min)
Information Design
Function
Scientific Discipline
Multi-Disciplinary
Index
Tele-StressPsychology
Operations Research,
Economics, HRM
Game Theory,Economics
Incentives
Service Engineering: Multi-Disciplinary ViewCall Center Design
7
DataMOCCA: Future
1. Operational (ACD) data with Business (CRM) data, plus Contents
2. Contact Centers: IVR, Internet, Chats, Emails, …
3. Daily update (as opposed to montly DVD’s)
4. Web-access (Research, Applications)
5. Nurture Research, for example
- Skills-Based Routing: Control, staffing, design, online; HRM
- The Human Factor: Service-anatomy, agents learning,
incentives,
6. Hospitals (already started, with IBM, Haifa hospital): Operational, Human-
Factors, Medical & Financial data; RFID’s for flow-tracking
8
DataMOCCA Interface: SEEStat
- Daily / monthly / yearly reports & flow-charts for a complete
operational view.
- Graphs and tables, in customized resolutions (month, days, hours,
minutes, seconds) for a variety of (pre-designed) operational
measures )arrival rates, abandonment counts, service- and wait-
time distribution, utilization profiles,…(.
- Graphs and tables for new user-defined measures.
- Direct access to the raw (cleaned) data: export, import.
9
Daily Report of April 13, 2004 – Regular Day
EntriesExitsAbnormal Termination
3760
3716
44
VRU
28968
2895
3
15
Announce
84645
4254
2
42103
Message
10394
1039
4
Direct
48240
603
2954247
3772
3
9130
1387
OfferedVolume
Handled
AbandonShort
AbandonCancel Disconnect
Continued
DataMOCCA – SEEStat
10
Daily Report of April 27, 2004 - Holiday779
22650
46849
4941
EntriesExitsAbnormal Termination
765
14
VRU22
64 7
3
Announce
2686 0
19989
Message
4941
Direct
22157
435
1072215
1858
1
2867
709
OfferedVolume
Handled
AbandonShort
AbandonCancel
Disconnect
Continued
DataMOCCA – SEEStat
11
Daily Report of April 20, 2004 – Heavily Loaded Day
EntriesExitsAbnormal Termination
560955
67
42
VRU
26614
2658 6
28
Announce
157820
9382 7
63993
Message
15964
1596
4
Direct
62866
1024
432
1579
1
4974
8
11138
1980
OfferedVolume
Handled
AbandonShort
AbandonCancel
Disconnect
Continued
Interesting Scenarios: Peak Days
12
Pre-designed Operational Measures: %Abandonment, TSF
Time Series
DataMOCCA - SEEStat
ILTelecom, Week days
0
5
10
15
20
25
30
35
40
45
Jan-04 Mar-04 May-04 Jul-04 Sep-04 Nov-04 Jan-05 Mar-05 May-05
Month
Rat
e, %
Abandons proportion Engineering Abandons proportion Technical
Fraction waiting > 30 Engineering Fraction waiting > 30 Technical
13
Pre-designed Operational Measures: %Abandonment
Time Series
Interesting Scenarios: Platinum Customers (VIP’s ?)
Abandons proportionUSBank, Mondays
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
Mar-01 Jul-01 Nov-01 Mar-02 Jul-02 Nov-02 Mar-03 Jul-03
Month
Rate
, %
Retail Premier Business Platinum
14
Daily Reports
DataMOCCA - SEEStat
Pre-designed Operational Measures: Arrival Rates, 2/2005
Agent statusILTelecom
0
50
100
150
200
250
300
350
400
450
0:00 2:00 4:00 6:00 8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 0:00
Time (Resolution 5 min.)
Nu
mb
er
of
cases
01.02.2005 02.02.2005 03.02.2005 04.02.2005 05.02.2005 06.02.2005 07.02.2005
08.02.2005 09.02.2005 10.02.2005 11.02.2005 12.02.2005 13.02.2005 14.02.2005
15.02.2005 16.02.2005 17.02.2005 18.02.2005 19.02.2005 20.02.2005 21.02.2005
22.02.2005 23.02.2005 24.02.2005 25.02.2005 26.02.2005 27.02.2005 28.02.2005
15
Histograms
Pre-designed Operational Measures: Service Times
DataMOCCA - SEEStat
Customer service time, RetailUSBank, April 2001, Week days
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
00:00 01:10 02:20 03:30 04:40 05:50 07:00 08:10 09:20 10:30 11:40 12:50 14:00 15:10 16:20 17:30 18:40
Time (Resolution 1 sec.)
Rela
tive f
req
uen
cie
s %
Customer service timeUSBank, April 2001, Week days Retail Total
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000
Time (Resolution 1 sec.)
Rela
tive f
req
uen
cie
s %
Empirical Three-Parameter Lognormal )location=-4.882768 scale=5.263772 shape=0.8415916(
Fitting Distribution
16
Fitting Mixtures of Distributions for Customer service timeUSBank, April 2001, Week days Retail Total
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000
Time (Resolution 1 sec.)
Rela
tive f
req
uen
cie
s %
Empirical Total Lognormal Lognormal Lognormal Lognormal Lognormal
Fitting Mixture of 5 Lognormal Components
DataMOCCA – SEEStat
Fitting Mixtures of Distributions for Customer service timeUSBank, April 2001, Week days Retail Total
0.0000
0.0005
0.0010
0.0015
0.0020
0.0025
0.0030
0.0035
0.0040
0.0045
0.0050
0.0055
0.0060
0.0065
0.0070
100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 2100 2200 2300 2400 2500 2600 2700
Time (Resolution 1 sec.)
Rel
ativ
e fr
equ
enci
es %
Empirical Total Lognormal Lognormal Lognormal Lognormal Lognormal
Fitting Mixtures of Distributions for Customer service timeUSBank, April 2001, Week days Retail Total
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
0.80
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
Time (Resolution 1 sec.)
Rel
ativ
e fr
equ
enci
es %
Empirical Total Lognormal Lognormal Lognormal Lognormal Lognormal
1
2
34
5
17
Fitting Mixture of 5 Gamma Components
Interesting Scenarios: VRU/IVR
Fitting Mixtures of Distributions for VRU only timeUSBank, April 2001, Week days
0.0
0.1
0.1
0.2
0.2
0.3
0.3
0.4
0.4
0.5
0.5
0.6
0.6
0.7
0 50 100 150 200 250 300 350 400 450 500 550 600
Time (Resolution 1 sec.)
Rela
tive f
req
uen
cie
s %
Empirical Total Gamma Gamma Gamma Gamma Gamma
18
US Bank: Service Time Histograms for Telesales, 2001-3
Interesting Scenarios: Service Science
Agent service time, TelesalesUSBank, Week days
0.00
0.25
0.50
0.75
1.00
1.25
1.50
1.75
2.00
2.25
2.50
00:00 02:00 04:00 06:00 08:00 10:00 12:00 14:00 16:00 18:00
Time (Resolution 5 sec.)
Rela
tive f
req
uen
cie
s %
May-01 May-02 May-03
19
Israeli Bank: Histogram of Waiting Time
Interesting Scenarios: Psychological vs. System
Wait time(handled)ILBank, January 2006, Week days
0.000
0.050
0.100
0.150
0.200
0.250
0.300
0.350
0.400
0.450
0.500
0.550
0.600
0.650
00:12 01:12 02:12 03:12 04:12 05:12 06:12 07:12 08:12
Time (Resolution 1 sec.)
Rel
ativ
e fr
equ
enci
es %
20
Interesting Scenarios: Skills-Based-Routing
Average wait time(waiting)ILTelecom, October 2004, Week days
0
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
0:00 2:00 4:00 6:00 8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
Time (Resolution 30 min.)
Mea
ns
Private Private Platinum
But VIP customers wait less Performance level: Regular versus VIPMore Regular get service immediately
Served without waiting October 2004, Week days
0.0
20.0
40.0
60.0
80.0
100.0
120.0
7:00 9:00 11:00 13:00 15:00 17:00 19:00 21:00 23:00
Time (Resolution 30 min.)
Rate
, %
Private Private Platinum
21
Israeli Call Center: Technical Service
Interesting Scenarios: Scenario Analysis
Unhandled Calls on May 24th, 2005 Agents Status on May 24th, 2005
TechnicalILTelecom, 24.05.2005
0
25
50
75
100
125
150
175
200
225
0:00 2:00 4:00 6:00 8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
Time (Resolution 30 min.)
Num
ber
of c
ases
Arrivals to queue Unhandled
Agent status, TechnicalILTelecom, 24.05.2005
0.0
2.5
5.0
7.5
10.0
12.5
15.0
17.5
20.0
22.5
25.0
27.5
30.0
32.5
35.0
0:00 2:00 4:00 6:00 8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00 0:00
Time (Resolution 30 min.)
Num
ber
of c
ases
Total Available Business Line
22
23
Data-Based Research: Must (?) & Fun
- Contrast with “EmpOM”: Industry / Company / Survey Data (Social Sciences)
- Converge to: Measure, Model, Validate, Experiment, Refine )Physics, Biology,…(
- Prerequisites: OR/OM, (Marketing) – for Design; Computer Science,
Information Systems, Statistics – for Implementation.
- Outcomes: Relevance, Credibility, Interest; Pilot (eg. Healthcare). Moreover,
Teaching: Class, Homework (Experimental Data Analysis); Cases.
Research: Test (Queueing) Theory / Laws, Stimulate New Models / Theory.
Practice: OM Tools (Scenario Analysis), Mktg (Trends, Benchmarking).