Managing Learning and Turnover in Employee Staffing*



Similar documents
Bullwhip Effect Measure When Supply Chain Demand is Forecasting

REVISTA INVESTIGACION OPERACIONAL VOL. 31, No.2, , 2010

CHAPTER 22 ASSET BASED FINANCING: LEASE, HIRE PURCHASE AND PROJECT FINANCING

A Queuing Model of the N-design Multi-skill Call Center with Impatient Customers

Why we use compounding and discounting approaches

FORECASTING MODEL FOR AUTOMOBILE SALES IN THAILAND

The Term Structure of Interest Rates

Planning Demand and Supply in a Supply Chain. Forecasting and Aggregate Planning

Research Article Dynamic Pricing of a Web Service in an Advance Selling Environment

COLLECTIVE RISK MODEL IN NON-LIFE INSURANCE

Teaching Bond Valuation: A Differential Approach Demonstrating Duration and Convexity

Economics Honors Exam 2008 Solutions Question 5

Hanna Putkuri. Housing loan rate margins in Finland

A formulation for measuring the bullwhip effect with spreadsheets Una formulación para medir el efecto bullwhip con hojas de cálculo

TACTICAL PLANNING OF THE OIL SUPPLY CHAIN: OPTIMIZATION UNDER UNCERTAINTY

*The most important feature of MRP as compared with ordinary inventory control analysis is its time phasing feature.

Random Walk in 1-D. 3 possible paths x vs n. -5 For our random walk, we assume the probabilities p,q do not depend on time (n) - stationary

SECTION 1.5 : SUMMATION NOTATION + WORK WITH SEQUENCES

Data Protection and Privacy- Technologies in Focus. Rashmi Chandrashekar, Accenture

Financial Data Mining Using Genetic Algorithms Technique: Application to KOSPI 200

UNDERWRITING AND EXTRA RISKS IN LIFE INSURANCE Katarína Sakálová

Single-machine Scheduling with Periodic Maintenance and both Preemptive and. Non-preemptive jobs in Remanufacturing System 1

Derivative Securities: Lecture 7 Further applications of Black-Scholes and Arbitrage Pricing Theory. Sources: J. Hull Avellaneda and Laurence

4. Levered and Unlevered Cost of Capital. Tax Shield. Capital Structure

The Application of Multi Shifts and Break Windows in Employees Scheduling

Improving Survivability through Traffic Engineering in MPLS Networks

Monitoring of Network Traffic based on Queuing Theory

DBIQ USD Investment Grade Corporate Bond Interest Rate Hedged Index

Experience and Innovation

FEBRUARY 2015 STOXX CALCULATION GUIDE

Capital Budgeting: a Tax Shields Mirage?

Mechanical Vibrations Chapter 4

An Approach for Measurement of the Fair Value of Insurance Contracts by Sam Gutterman, David Rogers, Larry Rubin, David Scheinerman

Forecasting, Ordering and Stock- Holding for Erratic Demand

Learning objectives. Duc K. Nguyen - Corporate Finance 21/10/2014

INTEREST RATE FUTURES AND THEIR OPTIONS: SOME PRICING APPROACHES

A panel data approach for fashion sales forecasting

Introduction to Statistical Analysis of Time Series Richard A. Davis Department of Statistics

Amazon.com, Inc. started in 1995 as an Internet

DYNAMIC MODELS FOR VALUATION OF WRONGFUL DEATH PAYMENTS

Incremental calculation of weighted mean and variance

Double Entry System of Accounting

CHAPTER 3 THE TIME VALUE OF MONEY

MTH6121 Introduction to Mathematical Finance Lesson 5

AGC s SUPERVISORY TRAINING PROGRAM

Chapter 5. Aggregate Planning

ACCOUNTING TURNOVER RATIOS AND CASH CONVERSION CYCLE

An Optimal Control Approach to Inventory-Production Systems with Weibull Distributed Deterioration

Outline. Role of Aggregate Planning. Role of Aggregate Planning. Logistics and Supply Chain Management. Aggregate Planning

Bond Valuation I. What is a bond? Cash Flows of A Typical Bond. Bond Valuation. Coupon Rate and Current Yield. Cash Flows of A Typical Bond

The Interest Rate Risk of Mortgage Loan Portfolio of Banks

Circularity and the Undervaluation of Privatised Companies

Predicting Indian Stock Market Using Artificial Neural Network Model. Abstract

Combining Adaptive Filtering and IF Flows to Detect DDoS Attacks within a Router

DBIQ Regulated Utilities Index

Analysis of Tailored Base-Surge Policies in Dual Sourcing Inventory Systems

Enhancing Oracle Business Intelligence with cubus EV How users of Oracle BI on Essbase cubes can benefit from cubus outperform EV Analytics (cubus EV)

Analogue and Digital Signal Processing. First Term Third Year CS Engineering By Dr Mukhtiar Ali Unar

Stock Trading with Recurrent Reinforcement Learning (RRL) CS229 Application Project Gabriel Molina, SUID

THE FOREIGN EXCHANGE EXPOSURE OF CHINESE BANKS

Modelling Time Series of Counts

Basic Elements of Arithmetic Sequences and Series

1. C. The formula for the confidence interval for a population mean is: x t, which was

PROCEEDINGS OF THE YEREVAN STATE UNIVERSITY AN ALTERNATIVE MODEL FOR BONUS-MALUS SYSTEM

TEMPORAL PATTERN IDENTIFICATION OF TIME SERIES DATA USING PATTERN WAVELETS AND GENETIC ALGORITHMS

3. Cost of equity. Cost of Debt. WACC.

Solving Logarithms and Exponential Equations

How to set up your GMC Online account

OMG! Excessive Texting Tied to Risky Teen Behaviors

Research on Inventory Sharing and Pricing Strategy of Multichannel Retailer with Channel Preference in Internet Environment

Fuzzy Task Assignment Model of Web Services Supplier

Heuristic Approach to Inventory Control with Advance Capacity Information

Ranking Optimization with Constraints

APPLICATIONS OF GEOMETRIC

.04. This means $1000 is multiplied by 1.02 five times, once for each of the remaining sixmonth

I. Why is there a time value to money (TVM)?

Hilbert Transform Relations

France caters to innovative companies and offers the best research tax credit in Europe

Reaction Rates. Example. Chemical Kinetics. Chemical Kinetics Chapter 12. Example Concentration Data. Page 1

Abstract. 1. Introduction. 1.1 Notation. 1.2 Parameters

Confidence Intervals. CI for a population mean (σ is known and n > 30 or the variable is normally distributed in the.

Behavior Analysis of a Biscuit Making Plant using Markov Regenerative Modeling

Time Value of Money. First some technical stuff. HP10B II users

To c o m p e t e in t o d a y s r e t a i l e n v i r o n m e n t, y o u n e e d a s i n g l e,

Time Consisency in Porfolio Managemen

11/6/2013. Chapter 14: Dynamic AD-AS. Introduction. Introduction. Keeping track of time. The model s elements

Installment Joint Life Insurance Actuarial Models with the Stochastic Interest Rate

Kyoung-jae Kim * and Ingoo Han. Abstract

Transcription:

Maagig Learig ad Turover i Employee Saffig* Yog-Pi Zhou Uiversiy of Washigo Busiess School Coauhor: Noah Gas, Wharo School, UPe * Suppored by Wharo Fiacial Isiuios Ceer ad he Sloa Foudaio

Call Ceer Operaios cusomer populaio call rouig ad CSR schedulig algorihms CTI resources CSR populaio Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

CSR Plas Are Made a 3 Levels Mome-by-mome o-he-fly adjusmes of call rouig, overime, luch, breaks Week-by-week assigme of CSRs o schedules Mohly, quarerly arge saffig levels se ad hirig plas made Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Our Hierarchical Approach Focus o high level saffig problem formulae as discree-ime dyamic program Embed lower level schedulig ad call rouig problems schedulig sofware Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Relaed Lieraure Aggregae Plaig HMMS Model (1960) Orrbeck, Schuee, ad Thompso (1968) Griold ad Saford (1974) Eber (1976) Khoshevis ad Wolfe (1986) Gerchak, Parlar, ad Segupa (1990) Huma Resources Barholomew, Forbes, ad McClea (1991) Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Relaed Lieraure Iveory Theory Arrow, Karli ad Scarf (1958) Veio (1965) Iglehar ad Jacque (1969) Karli (1960) ad Zipki (1989) Gerchak ad Wag (1994) Call Ceer Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Saffig Problem Descripio Sochasic o-saioary service requireme Ucerai service capaciy: people's service capaciies chage Radom learig Radom urover Log hirig/raiig leadime Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Service Requireme Varies by Time of day Day of week / moh Moh of year Promoio Special eves umber of calls per day 2500 2000 1500 1000 500 0 1 15 29 43 57 71 85 day umber Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Learig Source: Gusafso, 1982 Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Turover Call ceers have high urover raes Turover decreases wih job eure Empirical evidece Naioal Logiudial surveys (Ligh ad Ureda 1990) AT&T service represeaives (Gusafso 1982) Theoreical explaaios loger eure implies beer ``fi" (Jovaovic 1979) repeaed searches for a beer job (Parsos 1973) Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Oulie Model Srucural Properies of The Opimal Policy Model Deails Numerical Aalysis Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Model: Dyamics Idices i = 1, 2,, m : employee ype = 0, 1,, T : discree ime periods Sae variables i, : ype i employee, ime period (before hirig) x : ype 1 employee hired, ime period (decisio variable) y : ype1 employee, ime period (afer hirig) y = 1, + x Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Model: Dyamics (cod) Trasiio variables q i, ( i, ) : ype-i employees who ur over, ou of radom variable i, l i, ( i, ) : ype-i employees who lear, ou of radom variable i, q i, ( i, ) = i, qi, l i, ( i, ) = i, (1 qi, ) li, Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Model: Coss α : discou facor h : ui hirig cos W i : per sage wage cos, ype i employee O ( y,,,...,, ; D : operaig cos 2 m icludig overime ad ousourcig coss joily covex i ) ( y,,..., ) 2, m, Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004 Dyamic Programmig Model { } + + + + + + = + + + = ),..., ( ),...,, ( ) ( mi ),..., ( 1, 1 1, 1 },...,,,..., {, 2, 1, 2, 1, 1 0, 1, 1, 1,, 1, m l l q q m m i i i x m V E x O W x W hx V m m α

Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004 Cosrais Subjec o he sysem dyamics: ) ( ) ( ) ( ) ( ) ( 1, 1,,,,,, 1, 1, 1, 1 1, 1, i i i i i i i i l l q y l y q y x y + + + = = + = 1 m. < i

Sysem Dyamics Illusraed ype i i, demad ad coss realize q ( ) l ( ) + l ( i, i, i, i, i, i 1, i 1, ) + + 1 + x learig ad hirig urover hirig ime period ype 1 y y q ( y ) l ( y 1, 1, ) Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Opimaliy of Base-Level Policy Base-Level Policy: here exis saedepede base-levels y ( 2, *,..., m, ) such ha he opimal umber o hire is x = * * y ( 2,,..., m, ) 1, if 1, < y ( 2,,..., m, 0 oherwise * ) Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Proof Idea Covexiy of oe-period cos fucio propagaes V (,..., ) where 1, m, ( h+ W1 ) y + O( y,..., m, ) = mi h + y { } 1, + α E V+ 1( 1, + 1,..., m, + 1) 1, + 1 i, + 1 m, + 1 = (1... = (1... = (1 q q q 1, i, m, )(1 )(1 ) m, l 1, l i, ) y ) + l i, m 1, + l i 1, m 1, i 1, m W 1, i i, i= 2 Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Proof Idea Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Model Deails (I): Operaig Cos Fucio O (y m, ; D ) Sub-ierval: 1 2 3........ s..........., S D = D1 D 2 D3 Ds DS w = work schedule 1 I(w,s)=0 I(w,s)=1 Type-1 (42) Type-2 (26). work schedule w Type-3 (15) Type-4 (9). work schedule WS Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Operaig Cos Fucio O() s = 1,, S : sub-iervals i ay hirig period w = 1,, WS : work schedules (I(w,s) = 1 if w requires workig i sub-ierval s, 0 oherwise). x i,w = he umber of ype-i employees assiged o work schedule w z s : he amou of work o be ousourced i sub-ierval s Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Operaig Cos Fucio O() Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

More Geeral O() Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Model Deails (II): Hirig/Traiig Lead Time Hirig/raiig lead ime λ > 0 modeled by addig employee ypes m > λ i = 1,, λ idicae hirig/raiig sages Hirig lead ime µ i = 0 W i may be 0 Traiig lead ime µ i may be 0 W i > 0 Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Numerical Aalysis Modelig of differe CSR classes Modelig of he sochasic elemes Compare Opimal policy: sae-depede recogize differe employees have differe capaciies ake io accou he sochasic elemes Heurisic: headcou policy check oly he oal umber of employees assume everyoe has he same (average) capaciy ake io accou he sochasic elemes Heurisic: LP-based recogize differe employees have differe capaciies use average learig ad urover raes Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Example Problem: Slow ad Fas CSRs CSR urover slow CSRs: 48% yearly (15% per quarer) fas CSRs: 34% yearly (10% per quarer) CSR learig: slow ad fas CSRs CSR becomes fas afer oe quarer of experiece % speedup wih learig: 20%, 40%, 60%, 80% Call volume Demad: 250,000 per quarer CSR capaciy: 10,000 per quarer o average Overime as % of regular ime: 10%, 20%, 30% Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Coss Used for Examples Wage is capaciy eural, for a avg. capaciy of 10,000 per quarer: Base wage: $4,500, beefis: $1,000 (22.2%) Overime: 1.5 imes base wage rae Ousource cos: 1.8, 18, 36, 91, 182 imes base wage rae Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Opimal v.s. Headcou: Ample Overime 700 0.6% 0.6% 0.4% 0.3% % cos icrease Average Aual Cos ($ 000's) 650 600 550 500 450 OT+OS Hirig Wage 400 HC 20% OPT 20% HC 40% OPT 40% HC 60% OPT 60% HC 80% OPT 80% policy, % learig Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Opimal vs Headcou: Limied Overime 700 0.7% 2.6% 6.6% 9.9% % cos icrease Average Aual Cos ($ 000's) 650 600 550 500 450 OT+OS Hirig Wage 400 HC 20% OPT 20% HC 40% OPT 40% HC 60% OPT 60% HC 80% OPT 80% policy, % learig Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Cos Variabiliy Average Cos +/- 1 Sadard Deviaio 500 250 0 HC 20% OPT 20% HC 40% OPT 40% HC 60% OPT 60% HC 80% OPT 80% $ (000's) Ample Overime 1500 1000 500 0 HC 20% OPT 20% HC 40% OPT 40% HC 60% OPT 60% HC 80% OPT 80% Limied Overime Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

probabiliy Opimal vs. Headcou 0.2 0.1 Disribuio of Regular Time Capaciy 30% overime 0.3 capaciy below which 0.3 ousourcig required service requireme head- 0.2 cou policy opimal policy 0.1 10% overime service requireme capaciy below which ousourcig required headcou policy opimal policy 0 150 200 250 300 350 400 0 capaciy (1,000 calls / quarer) 150 200 250 300 350 400 Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Our Fidigs Whe spare capaciy is cheap simpler hirig schemes do well less eed o rack employees carefully heurisics ha igore learig Whe spare capaciy is igh simpler hirig schemes ca do poorly value o rackig employees carefully more complex approaches: DP, IPA Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Opimal vs. LP Heurisic W/O raiig LT: LP does very well Whe here is a oe-period leadime: LP capaciy ofe falls shor 80% 60% 40% 20% 0% os=$1 $10 $20 $50 $100 10% 20% 30% learig Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Our Fidigs LP heurisic performs well, whe raiig is fas -- LP ca wai o observe he umber of employees available ad hire accordigly raiig akes ime, bu overime is ample, or OS (or service failure) is cheap -- OT ad OS ca cover capaciy shorfall whe urover is higher ha average The opimal base level srucure simplifies calculaio LP performs poorly, whe overime is limied ad OS (or service failure) is expesive If MDP is oo complicaed o use, he a leas add some buffer capaciy i cojucio wih LP Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004

Summary A ew framework o sudy employee saffig problems DP approach, allows radomess i he sysem Naural icorporaio of he schedulig ad rouig problem Opimal policy ad srucural properies Numerical resuls show whe i is worhwhile o more carefully moior sysem capaciy Call Cere Workshop, Cere de Recherches Mahemaiques, Uiversie de Moreal, 07/25/2004