quality - speed conundrumorg.business.utah.edu/opsconf/pages/veeraraghavan_slides.pdf · 2010. 1....

Quality - Speed ConundrumTradeoffs in Customer Intensive Services

Senthil K. Veeraraghavan

joint work with:

Krishnan Anand, David Eccles School of Business, University of UtahM. Fazıl Paç, Wharton School, University of Pennsylvania

University of UtahUtah Winter Operations Conference, January 2010

Veeraraghavan (Wharton) Quality - Speed Conundrum Utah January 2010 1 / 29

Quality Speed Tradeoff in Services

Generally service speed and service value are considered asindependent performance metrics

Focal point of service quality in queueing literature: service deliverylead times/waiting timesFast service/low waiting times often considered better.

Customer intensive services: Primary health care, hospitality andeducation.

Value provided increases with time spent with customer.Longer service times are preferred and provide more value...... but no one likes to wait.


Preview of some main results

For customer-intensive services, market coverage is often partial.(i.e. many customers remain unserved).In equilibrium, the servers “slow” down.Equilibrium demand may increase with market price.Service Competition may increase prices!


Psychology of Waiting vs. Service Duration

In many scenarios, customers prefer to be served well, but dislikewaiting for the same service (Schelling 1975).


Customer-Intensive Services

Some services are surely more customer-intensive!


Examples

Beauty-care services (hairdressers, spas ... )

Consulting services (business, legal, financial ...)

Informative call centers (nurse-lines)


Baumol’s Cost Disease (Baumol 1996) for Queues

Empirical evidence: Productivity in customer-intensive (labor-intensive)service industries is low, although it is high for services in general(Varian 2004, Triplett and Bosworth 2003, Yarnall et. al. 2003).

Cost Disease and Customer Intensity:"You can control drug costs and limit expensive new procedures, but when it comes tohospital care and doctor visits, the only way to improve productivity is to shrink the size ofthe staff and have doctors spend less time with patients. To lower prices you have to lowerquality." (Surowiecki 2003)

Cutting Corners⇐⇒ Increase in Productivity.


Service Value in Primary Care

"Regular" DoctorsRushing between patients. "I was seeing 30 people a day and always rushing. Patients were

dissatisfied.... I was dissatisfied." (Dr. Kaminetsky, NYU, Testimony to Joint Economic Committee U.S. Congress)

Not enough time for preventive care.

Service delays are likely, due to demand related congestion.Charge "relatively" lower prices.

"Concierge" ("Boutique") Doctors

(MDVIP, MD2, Current Health, Qliance Primary Care,...)More time with each patient (highly customized treatment).Offer immediate response (no service delay),Charge high prices.


Speed vs. Quality in Primary Care

Does Good care take longer? [Chen et al 2009, Health CareReform].Making Sense of Primary Health Care. [NY times, January 17,2010]


Related Literature

Rational Queue JoiningNaor (1969), Kalai et al (1992), Li (1992), Li and Lee (1994), Lederer and Li (1997), Gilbert and Weng (1998), Armony and Haviv(2000), Edelson and Hildebrand (1971), Wang and Olsen (2008), Allon and Federgruen (2007), Chen and Wan (2003).

Service Value/CongestionCachon and Harker (2002), Afeche (2006), Gans (2002), Hopp, Iravani and Yuen (2007), Ren and Wang (2009), Wang, Debo andScheller-Wolf (2008), Lu, van Mieghem and Savaskan (2009), Hasija, Pinker and Shumsky (2009), de Vericourt and Sun (2009).

Productivity/QualityChase and Tansik (1983), Oliva and Sterman (2001).

Majority of the literature considers service speed and service value asindependent performance metricsQuality-speed tradeoff in services: Interaction between service speedand service quality.Empirical evidence: Aiken et al (2002), Diwas and Terwiesch (2008),Tucker et al (2006).


Model of Customer-Intensive Service

Queueing models of advance access in primary care settings (Bowers and Mould 2002,Green and Savin 2008)


Value of Customer-Intensive Service


Modeling Value of a Customer-Intensive Service

mu0.5 1.0 1.5 2.0 2.5 3.0

0

10

20

30

40

a = 10

a = 2

a = 1

V

a = 0

Vb = 5, mb

= 2

(a) V = Vb + αg(µb)− αg(µ)

t1 2 3 4 5

0

1

2

3

4

5

6

7

a = 1

a = 2

a = 0

a = 10

Vb = 5, mb

= 2

(b) V = Vb + αf (tb)− αf (t)

Figure: Service Value for labor-intensive services varies with service time

Valuation as a function consistent with diagnostic services with sequential hypothesistesting (Wang et al 2008)


Service Provider’s Operating Region

Service has to be fast enough: µ > A1(α)No one will wait forever even if the service value is high.

Service cannot be "too fast": µ < A2(α)Not possible to provide valuable service at really high speeds.


Queueing Game

No Market Coverage: No customer joins the queue if V (µ)− p ≤ c/µ.

Full Coverage: All customers join the queue IF V (µ)− p ≥ c/(µ− Λ).

Partial Coverage: V (µ)− p < c/(µ−Λ) and V (µ)− p > c/µ, there existssome mixed equilibrium such that some customers join, and others balk.


Service Provider’s Revenue Maximization

Service provider maximizes revenues with respect to price andservice rate:

max{p≥0,µ∈F(α)} {R(µ,p) = pλe(µ,p)}

Equilibrium demand:

λe(µ, p) =

Λ if 0 ≤ p ≤ V (µ)−WC(µ,Λ)µ− c

V (µ)−p if V (µ)−WC(µ,Λ) < p ≤ V (µ)−WC(µ, 0)

0 if V (µ)−WC(µ, 0) < p.


Service Provider’s Price Decision

For a fixed service rate µ in the operating region F(α), serviceprovider maximizes revenues with respect to price.

Equilibrium outcome depends on the threshold λ̂(µ) = µ−√

cµV (µ) .

Optimal price:

p∗(µ) =

{V (µ) − WC(µ, Λ) if 0 ≤ Λ ≤ λ̂(µ)

V (µ) −√

cV (µ)/µ if λ̂(µ) < Λ.(1)

Resulting equilibrium arrival rate:

λe(µ, p∗(µ)) =

{Λ if 0 ≤ Λ ≤ λ̂(µ)

λ̂(µ) if λ̂(µ) < Λ.(2)

Corresponding equilibrium revenue:

R(µ, p∗(µ)) =

{(V (µ) − WC(µ, Λ))Λ if 0 ≤ Λ ≤ λ̂(µ)

µV (µ) − 2√

cµV (µ) + c if λ̂(µ) < Λ.(3)


Customer-intensity and Market Coverage

As services become customer-intensive, monopoly is less likely to coverthe market.⇒ Higher customer-intensity leads to partial coverage.


Partial Market Coverage: Equilibrium Price,Throughput and Revenue

Region 1 (Low µ): Increase price and admit more customers

Region 2 (Intermediate µ): Decrease price and admit more customers

Region 3 (High µ): Decrease price and admit less customers


Partial Market Coverage: Price-Throughput Tradeoffs -An Example

Increase service rate from µ = 4 to µ = 6.Examining equilibrium price-demand curve ...


Partial Market Coverage (Λ > λ∗α)

Optimal service rate: µ∗ = Vb+αµb2α .

As the service becomes more labor intensive (α ↑) the serviceprovider spends more time on each customer (µ∗ ↓).

Optimal price: p∗(µ∗) = Vb+αµb−2√

cα2

Optimal price falls (increases) as labor intensity increases forα < (>)c/µ2

b.

Equilibrium Demand: λe(µ∗,p∗(µ∗)) = Vb+αµb−2√

cα2α .

Equilibrium demand falls (increases) as labor intensity increases forα < (>)V 2

b /c.Higher waiting costs (c ↑) lead to lower prices and lower equilibriumdemand.


Full Market Coverage: (Λ < λ∗α)

Service provider serves all potential customers:λe(µ∗,p∗(µ∗)) = Λ

Optimal service rate: µ∗ = Λ +√

c/α.As the service becomes more labor intensive (α ↑) the serviceprovider spends more time on each customer (µ∗ ↓).

Optimal price: p∗(µ∗) = Vb + αµb − αΛ− 2√αc.

If Λ is low, then optimal price falls (increases) as labor intensityincreases for α < (>) c

(µb−Λ)2 .

Higher waiting costs lead to faster service at lower price.


Service Provider with Multiple Competing Servers

Multiple (n) servers owned by a single service provider providing service of intensity α.Service provider sets price to maximize total revenues.Servers set their own service speed to maximize their revenues.Surgeons who belong to the same health network (or hospital).Tax consultants at a single tax firm.


Service Provider with Multiple Servers: Partial MarketCoverage

Large Market (Λ ≥ nλ∗α)Servers act as monopolists.Competition has no effect on the optimal operating setting.Optimal service rate of server i is µ∗i = Vb+αµb

2α .

Optimal price is p∗ = Vb+αµb−2√αc

2


Service Provider with Multiple Servers: Full MarketCoverage

When Λ < nλ∗α, servers share the market demand in equilibrium, servingat rate µe

i = Λn +

√c/α.

Server spends more time on each customer than in the singleserver case.Equilibrium expected waiting cost of a customer is identical to theequilibrium waiting cost of the single server case.The optimal admission price is increasing in the number of servers.


Concluding Remarks

A Queueing Model based explanation for Baumol’s cost disease.Normative Model: Service quality vs. time in customer-intensiveservices.

The quality speed trade-off analyzed from both customers’ andservice providers’ perspectives.

Slow service is necessary to meet the service-level (quality)requirements.

Necessitating higher capacity investments⇒ Baumol’s Cost Disease.As customer-intensity of the service increases, server must spendmore time with each customer.

Faster service preferred over more valuable service whencustomer-intensity is low.


Concluding Remarks

Price and demand for the service move in tandem as thecustomer-intensity changes.In contrast with the downward-sloping demand curve because:

Price⇔ service quality⇔ service rate.Revenues are determined by⇒ the quality-speed trade-off.

Revenues are decreasing in the customer-intensity if thecustomer-intensity is low.Revenues are increasing in the customer-intensity if thecustomer-intensity is high.Competition in customer-intensive services may:

Increase the prices,Enhance the service value,Without increasing the congestion.


Acknowledgements

Thank You

Fishman-Davidson Center for Service Research.

Philipp Afeche, Baris Ata, Gérard Cachon, Francis de Vericourt, JackHershey, Ananth Iyer, Raj Rajagopalan, Rob Shumsky, Alan Scheller-Wolf,and anonymous reviewers.

Participants at MSOM Service SIG 2009, Carnegie Mellon University,PennState, Northwestern University, University of Maryland, University ofPennsylvania, University of Rochester, INFORMS JFIG CompetitionCommittee.


Research Questions...

Strategic Service and Expert Diagnosis (Pac and Veeraraghavan 2010).

Heterogeneity in Customer-intensity α.

Credence good services and information-asymmetry

Multi-stage Diagnostic Services.


quality - speed conundrumorg.business.utah.edu/opsconf/pages/veeraraghavan_slides.pdf · 2010. 1....

Documents