DataMOCCA Data MOdels for Call Center Analysis Project Partners: Technion: Paul Feigin, Avi Mandelbaum, Sergey Zeltyn (IBM Research) SEElab : Valery Trofimov,

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


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


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


11

Daily Report of April 20, 2004 – Heavily Loaded Day


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


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


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


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


Rela

tive f

req

uen

cie

s %

Empirical Total Lognormal Lognormal Lognormal Lognormal Lognormal

Fitting Mixture of 5 Lognormal Components



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


Rel

ativ

e fr

equ

enci

es %



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


Rel

ativ

e fr

equ

enci

es %


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


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


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


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


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


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


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


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).

Documents

DataMOCCA Data MOdels for Call Center Analysis Project Partners: Technion: Paul Feigin, Avi Mandelbaum, Sergey Zeltyn (IBM Research) SEElab : Valery Trofimov,