Re k Gull u Bo gazi˘ci University Industrial Engineering ... · Pricing in Service Systems with...
Transcript of Re k Gull u Bo gazi˘ci University Industrial Engineering ... · Pricing in Service Systems with...
Pricing in Service Systems with Strategic Customers
Refik Gullu
Bogazici UniversityIndustrial Engineering Department
Istanbul, Turkey
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 1 / 86
Service Systems with Strategic Customers
Before the talk
My professor’s advice on queueing theory versus game theory
mathematical difficulty versus conceptual maturity
How to learn new stuff?
teaching, writing a code, thesis supervision
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 2 / 86
Service Systems with Strategic Customers
Before the talk
A review from a personal perspective
A great place to start reading:Rafael Hassin and Moshe Haviv, To queue or not the queue:equilibrium behavior in queueing systems, Springer, 2003.http://www.math.tau.ac.il/ hassin/book.html
A follow up survey book, “Rational Queueing” to appear soon
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 3 / 86
Service Systems with Strategic Customers
Outline
A simple example of an “unobservable” queue
Parameter uncertainty
Effect of delay information
Observable queues: residual service time
Multiple customer types: identical price
Multiple customer types: differentiation
An inventory model
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 4 / 86
Service Systems with Strategic Customers
A Framework Example
A student cafeteria in a university
University administration regulates the price
possibly in the form of a subsidy
Students arrive according to a Poisson process
Λ is the rate of potential students
A single server with exponential rate µ > Λ.
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 5 / 86
Service Systems with Strategic Customers
A Framework Example
Cafeteria is sort of far away from the main building
Students decide eating there or not before observing the congestion
There are other dining facilities on campus
Once a decision is made, it can not be changed
Students are identical
with respect to their valuation of the service, value of time, and theirbehaviour towards riskall are rational decision makers
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 6 / 86
Service Systems with Strategic Customers
A Framework Example
R : the value of service (as judged by students)
c : the unit cost for waiting
p : fee for dining at the cafeteria
The expected utility of a student from the service
R− p− cE[sojourn time]
the system is at steady state
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 7 / 86
Service Systems with Strategic Customers
A Framework Example: equilibrium behaviour
As students are “identical”, their equilibrium behaviour is expected tobe the same
Each student choose to enter the cafeteria with probability q
Let U(qtagged, qothers) be the utility of a tagged student when all theothers behave with qothers
Best response against qothers: U(q′, qothers) ≥ U(q, qothers) for all q.
Symmetric Nash equilibrium: best response against itself
U(qe, qe) ≥ U(q, qe) for every q
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 8 / 86
Service Systems with Strategic Customers
A Framework Example: equilibrium behaviour
0 ≤ qe(p) ≤ 1 is the equilibrium probability of joining the cafeteriawhen the fee is p
λe(p) = Λqe(p) < µ
For effective arrival rate λ < µ
w(λ) = 1/(µ− λ)
3 cases need to be examined
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 9 / 86
Service Systems with Strategic Customers
Case 1
Nobody else is joining and
p+ cw(0) > R =⇒ R < p+ c1
µ
µ <c
R− p
qe(p) = 0λe(p) = 0
w(λe(p)) = 1/µ
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 10 / 86
Service Systems with Strategic Customers
Case 2
If everybody else is joining and
p+ cw(Λ) ≤ R =⇒ R ≥ p+ c1
µ− Λ
µ ≥ Λ +c
R− p
qe(p) = 1λe(p) = Λ
w(λe(p)) = 1/(µ− Λ)
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 11 / 86
Service Systems with Strategic Customers
Case 3
p+ cw(0) ≤ R < p+ cw(Λ)
R = p+ cw(λe(p))
qe(p) = λe(p)/Λw(λe(p)) = 1/(µ− λe(p))
λe(p) = µ− c
R− p
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 12 / 86
Service Systems with Strategic Customers
Administration’s Problem: social optimization
The administration cares about the overall performance
Solves the following problem
max0≤λ≤Λ
{λ(R− c 1
µ− λ)}
strictly concave
maximum occurs at
λ∗ = µ−√cµ
R
λ∗ ≥ 0 (by R ≥ c/µ)
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 13 / 86
Service Systems with Strategic Customers
Administration’s Problem: social optimization
By considering the constraint λ ≤ Λ
λ∗ = min{Λ, µ−√cµ
R}
if Λ ≥ µ−√
cµR
optimal objective function value
(√Rµ−
√c)2
w(λ∗) =√
Rcµ
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 14 / 86
Service Systems with Strategic Customers
Administration’s Problem: social optimization
if Λ ≤ µ−√
cµR
optimal objective function value
Λ(R− c
µ− Λ)
w(λ∗) = 1µ−Λ
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 15 / 86
Service Systems with Strategic Customers
By assuming R ≥ c/µ
λe(0) = µ− c
R≥ µ−
√cµ
R= λ∗
Individual optimization leads to longer queues than imposed by socialoptimization
Admission fee can regulate this
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 16 / 86
Service Systems with Strategic Customers
Revenue maximization
Let pm be the admission fee charged for dining
pm = R− cw(λ)max
0≤λ≤Λpmλ
Same as the social optimization objective
The socially optimal arrival rate can be induced by the fee
pm = p∗ = R− cw(λ∗) = R−
√cR
µ
λe(p∗) = λ∗, profit = (
√Rµ−
√c)2
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 17 / 86
Service Systems with Strategic Customers
Rafael Hassin and Moshe Haviv, To queue or not the queue: equilibriumbehavior in queueing systems, Springer, 2003. (Chapter 3)
Bell, C. E., Stidham, Jr., 1983, Individual versus Social Optimization inthe Allocation of Customers to Alternate Servers, Management Science,29, 831-839.
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 18 / 86
Service Systems with Strategic Customers
Parameter Uncertainty
The preceding model considered an “unobservable” system
The queue length or the waiting times upon arrival are unobservable
Need to be careful
Many things are known and/or intelligently computable: service rate,expected waiting time, service value, etc.
These parameters are known with certainty
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 19 / 86
Service Systems with Strategic Customers
Uncertainty in the Service Rate
Suppose µ, the service rate, can take two values
µ =
{µ1 with probability αµ2 with probability 1− α
µ1 > µ2
Do students know the realised service rate?
No, they are “uninformed”Yes, they are “informed”, and the server charges either a different feeor the same fee
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 20 / 86
Service Systems with Strategic Customers
Service Rate Uncertainty
For the “uninformed” case
v = (R− p)/cv = α
µ1−λ + 1−αµ2−λ in equilibrium
v is a solution of a nonlinear equation
Πun = λ(R− cv)
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 21 / 86
Service Systems with Strategic Customers
Service Rate Uncertainty
“informed” case with two prices
λi = µi − cR−pi , pi = R−
√cRµi
Πin2 =
α(√Rµ1 −
√c)2 + (1− α)(
√Rµ2 −
√c)2 if R ≥ c
µ2
α(√Rµ1 −
√c)2 if c
µ1< R < c
µ2
0 if R ≤ cµ1
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 22 / 86
Service Systems with Strategic Customers
Service Rate Uncertainty
“informed” case with a single price
λi = µi − cR−p (rate of arrivals given the service rate)
If p is small enough to attract customers for both values of µ
average arrival rate for the single price p
λ = α(µ1 −c
R− p) + (1− α)(µ2 −
c
R− p)
= µ− c
R− p
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 23 / 86
Service Systems with Strategic Customers
Service Rate Uncertainty
Maximizing pλ = pµ− pcR−p
(R− p)2 = Rc/µ
=⇒ p = R−
√Rc
µ
Resulting profit(√Rµ−
√c)2
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 24 / 86
Service Systems with Strategic Customers
Service Rate Uncertainty
But a higher price may be chosen: so that customers opt out whenµ = µ2.
p = R−√
Rcµ1
Resulting profit: α(√Rµ1 −
√c)2
Two profit terms are equal for
R
c= η =
(1−√α√
µ−√µ1α
)2
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 25 / 86
Service Systems with Strategic Customers
Service Rate Uncertainty
Πin1 =
(√Rµ−
√c)2 if R
c ≥ ηα(√Rµ1 −
√c)2 if 1
µ1≤ R
c ≤ η0 if R
c ≤1µ1
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 26 / 86
Service Systems with Strategic Customers
Parameters Uncertainty
The service provider benefits from revealing the service rate, and frompricing accordingly
Πin2 ≥ Πin
1 ≥ Πun
As the variability in service rate increases, Πin2 increases
The server provider may lose (1− α) fraction of the customers butextracts higher revenue from the remaining
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 27 / 86
Service Systems with Strategic Customers
Parameters Uncertainty
Hassin, R., 2007, Information and Uncertainty in a Queuing System,Probability in the Engineering and Informational Sciences, 21, 361-380.
waiting cost uncertainty
service valuation uncertainty
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 28 / 86
Service Systems with Strategic Customers
Delay Information
So far: “unobservable” systems
What if the customers are revealed information on the possible delaybefore they decide to join or not
”observable” systems
Three levels of information
1. No information (same as before)
2. Partial information: how many customers are in front ofme?
3. Full information: what is my exact waiting time?
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 29 / 86
Service Systems with Strategic Customers
Delay Information
M/M/1 type service system
W : waiting time (in the queue)
θ: customer type parameter, a random variable, θ ∈ [0, 1] with cdf H,pdf h
c(t): cost of waiting t time units
U = R− θE[c(W )]
Previously: θ ≡ 1, c(t) = ct
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 30 / 86
Service Systems with Strategic Customers
Delay Information
Scale R and c(t) so that R = 1
Assume that c(0) > 1
Customers with θ > 1/c(0) balk
Scale Λ (ignore them) to λ, and assume (new) c(0) = 1
Customers with θ ≈ 1 are also attracted to join when W = 0.
U(no waiting) = 1− θ
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 31 / 86
Service Systems with Strategic Customers
Delay Information: who stays in the system?
Information I, a random variable
Want: U |I = 1− θEW [c(W )|I] ≥ 0
Given information I = i, an arriving customer stays if
θ ≤ θi =1
EW [c(W )|I = i]
Pr{stays|I = i} = H(θi)
Fraction of customers who stay: EI [H(θI)]
ThroughputλEI [H(θI)]
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 32 / 86
Service Systems with Strategic Customers
Average utility
Define
J(θ) =1
θ
∫ θ
0H(x)dx
Average utility
u = E[U+] = Eθ,I [(1− θEW [c(W )|I])+]
= EI
[∫ θI
0(1− xEW [c(W )|I])h(x)dx
]= EI
[H(θI)− (1/θI)
∫ θI
0xh(x)dx
]= EI [J(θI)]
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 33 / 86
Service Systems with Strategic Customers
Case 1: No information
The equilibrium arrival rate:
λNI = λH
(1
E[c(WNI)]
)ρNI = λNI/µ
c(s) =∫∞
0 e−stc(t)dt LST of c(t).
Pr{WNI > t} = ρNIe−µ(1−ρNI)t
E[c(WNI)] = (1− ρNI) + ρNIµ(1− ρNI)c(µ(1− ρNI))
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 34 / 86
Service Systems with Strategic Customers
Case 1: No information
An example: Uniform customers with Linear Cost
c(t) = 1 + t
c(s) = 1s + 1
s2
λNI =λ
1 + ρNI/(µ(1− ρNI))
=⇒ (1− µ)(ρNI)2 + (µ+ λ)ρNI − λ = 0
=⇒ ρNI =−(µ+ λ) +
√(λ+ µ)2 + 4λ(1− µ)
2(1− µ)
πNIn = (1− ρNI)(ρNI)n
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 35 / 86
Service Systems with Strategic Customers
Case 2: Partial information
The service provider tells the arriving customer N(t)
The customer computes cn = E[c(W )|N(t) = n]
Stays if θ ≤ θn = 1/cn
Birth-death process with state dependent arrival rate λn = λH(θn)
Steady-state probabilities
Θn =
n−1∏m=0
H(θm), Θ =
∞∑n=1
Θn(λ/µ)n
πPI0 = 1/(1 + Θ)
πPIn = Θn(λ/µ)nπPI0
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 36 / 86
Service Systems with Strategic Customers
Case 2: Partial information
An example: Uniform customers with Linear Cost
c(t) = 1 + t
θm = 11+m
µ, Θn =
∏n−1m=0
11+m
µ
Θn(λ/µ)n = Γ(µ)Γ(µ+n)λn
πPI0 =1
1 + γ(µ, λ)λ1−µeλ, πPIn = πPI0
Γ(µ)
Γ(µ+ n)λn
Γ(x) =
∫ ∞0
tµ−1e−tdt and γ(µ, λ) =
∫ λ
0tµ−1e−tdt
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 37 / 86
Service Systems with Strategic Customers
Case 3: Full information
The service provider tells the arriving customer workload V (t) = v
Critical point: θv = 1/c(v)
Effective arrival rate λ(v) = λH(θv)
f(v): the pdf of the stationary workload V
under linear cost and uniform customers
πFI0 =1
1 + λeµµ−(λ+1)γ(λ+ 1, µ)
fFI(v) = λπFI0 (1 + v)λe−µv, v > 0
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 38 / 86
Service Systems with Strategic Customers
Benefit of Information
The service provider would like to have large throughput λEI [H(θI)]
Customers would like to have a large average utility EI [J(θI)]
EI
[1
θI
∫ θI
0H(x)dx
]What is the impact of information on these measures?
Clearly the answer depends on H(x).
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 39 / 86
Service Systems with Strategic Customers
Benefit of Information
There are cases where the service provider and the customers arealigned
H(x) = xα, α > 0
J(θ) = 1θ
∫ θ0 x
αdx = 1α+1θ
α = 1α+1H(θ)
Average utility ∼ Throughput
More information is better for all parties
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 40 / 86
Service Systems with Strategic Customers
Benefit of Information
Consider constant θ
No information: throughput is λ for sufficiently small λ
Partial information: there is a threshold n∗ beyond which customersdo not join
throughput < λ
The service provider may hide information
Essentially whether information beneficial to one party or the otherdepends on the shape of H(θ)
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 41 / 86
Service Systems with Strategic Customers
Benefit of Information
If H(1/x) is convex in x ≥ 1, then more information benefits theservice provider by increasing throughput
If J(H−1(y)) is convex on [0, 1], then more information benefits thecustomers by increasing the average utility.
H(x) =γe−γx
1− e−γx ∈ [0, 1]
=⇒ γ ≤ 2 (h(x) does not decrease too rapidly)
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 42 / 86
Service Systems with Strategic Customers
Benefit of Information
If H(1/x) is convex, then πPI0 ≤ πNI0
πNI0 = 1− λ
µH
(1
E[cNNI ]
)πPI0 = 1− λ
µE
[H
(1
cNPI
)]πPI0 > πNI0 =⇒ NPI ≤st NNI =⇒ E[cNPI ] ≤ E[cNNI ]
=⇒ H
(1
E[cNNI ]
)≤ E
[H
(1
cNPI
)]by convexity of H(1/x) and Jensen’s inequality
a contradiction
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 43 / 86
Service Systems with Strategic Customers
Guo, P., Zipkin, P., 2007, Analysis and Comparison of Queues withDifferent Levels of Delay Information, Management Science, 53, 962-970.
Guo, P., Zipkin, P., 2009, The Effects of the Availability of Waiting-timeInformation on a Balking Queue, EJOR, 198, 199-209.
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 44 / 86
Service Systems with Strategic Customers
Observable Queues: Residual Service Time
Consider an observable M/G/1 queue
The arriving customer observes the queue length before joining
If the service time is exponential
Customer joins if the number in the system is less than⌊R
cE[Service Time]
⌋With non-exponential service times
Residual service time matters
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 45 / 86
Service Systems with Strategic Customers
Observable Queues: Residual Service Time
A customer who observes n customers upon arrival joins withprobability qn
(q1, q2, . . .)
The behaviour of others has an effect on the assessment of residualservice time
E[RSTn] = E[residual ST when the arriving customer finds n in the system]
E[RSTn] = fn(q1, q2, . . . , qn) (a recursive expression)
E[RST1] =E[ST ]
1− G(Λq1)− 1
Λq1
Suppose deterministic service time and q1 is highinformation about the current service state
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 46 / 86
Service Systems with Strategic Customers
Observable Queues: Residual Service Time
Once q1, q2, . . . , qn−1 are known
if nE[ST ] + fn(q1, . . . , qn−1, 1) ≤ R/c =⇒ qn = 1if nE[ST ] + fn(q1, . . . , qn−1, 0) ≥ R/c =⇒ qn = 0otherwise nE[ST ] + fn(q1, . . . , qn−1, q) = R/c =⇒ qn = q
Intuition: q1 ≥ q2 ≥ q3 ≥ · · ·Which turns out to be wrong!
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 47 / 86
Service Systems with Strategic Customers
Observable Queues: Residual Service Time
ST = 1 with probability ε (small), ST = 0 with probability 1− εE[RST1] = 1
1−e−Λq1− 1
Λq1
Solve ε+ E[RST1] = R/c to find Λq1.
Λ ≤ Λ1 =⇒ q1 = q2 = 1Λ1 < Λ ≤ Λ2 =⇒ 0 < q1 < 1, q2 = 1Λ > Λ2 =⇒ 0 < q1 < q2 < 1
The fact that there are two customers (one in service) means that thewe are probably nearing the end of the current service time, and theservice time of the next customer is very likely to be zero anyway
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 48 / 86
Service Systems with Strategic Customers
Observable Queues: Residual Service Time
If the service time distribution is of type “decreasing mean residuallife” (DMRL)
that is, E[ST − t|ST > t] is monotone decreasing in t
There is ne as the smallest integer satisfying
nE[S] + fn(1, 1, 1, . . . , 1) ≥ R/c
qe ∈ [0, 1) satisfying
neE[S] + fn(1, 1, 1, . . . , qe) = R/c
qn =
1 n < neqe n = ne0 n > ne
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 49 / 86
Service Systems with Strategic Customers
Observable Queues: Residual Service Time
Intuition: waiting time difference between states n and n+ 1
(n+ 1)E[ST ] +E[RSTn+1]−nE[ST ]−E[RSTn] > E[ST ]−E[RSTn] ≥ 0
Waiting times are increasing in nThere exist at most one n with mixed strategy[RSTn] is increasing in q for (q1, . . . , qn−1, q)Hence qe is unique.
“Avoiding the crowd” versus “following the crowd”
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 50 / 86
Service Systems with Strategic Customers
Naor, P., 1969, The Regulation of Queue Size by Levying Tolls,Econometrica, 37, 15-24.
Kerner, Y., 2011, Equilibrium Joining Probabilities for an M/G/1 queue,Games and Economic Behavior, 71, 521-526.
Haviv, M., Kerner, Y., 2007, On Balking from an Empty Queue, QueueingSystems, 55, 239-249.
Kerner, Y., 2008, The Conditional Distribution of the Residual ServiceTime in the Mn/G/1 Queue, Stochastic Models, 24, 364-375.
Manou, A., Economou, A., Karaesmen, F., 2014, Strategic Customers in aTransportation Station: When Is It Optimal to Wait?, OperationsResearch, 62, 910-925.
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 51 / 86
Service Systems with Strategic Customers
Multiple Customer Classes-identical price
Things get complicated with multiple customer classes
Consider two classes of customers with M/G/1 type service facility
R1, R2, c1, c2, Λ1, Λ2
Suppose that the customers are either
indistinguishable to the service provideror price discrimination is not possible
Service times are identically distributed
Customers are treated in a FCFS manner
They can not observe the queue length
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 52 / 86
Service Systems with Strategic Customers
Multiple Customer Classes-identical price
λ = λ1 + λ2, the equilibrium arrival rate
zi = Ri − (ci/µ), i = 1, 2
ui(λ, p) = Ri − p− ci(wQ(λ) + (1/µ)) = zi − p− ciwQ(λ)
z1 ≥ z2, and p ≤ zi (otherwise does not join)
ui(λ, p) is strictly decreasing and concave in λ
since wQ(λ) is convex increasing
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 53 / 86
Service Systems with Strategic Customers
Multiple Customer Classes-identical price
For any price p and 0 ≤ λ < µ,
c1 ≤ c2 =⇒ u1(λ, p) ≥ u2(λ, p)If c1 > c2, there is a critical value λ so that
λ > λ =⇒ u1(λ, p) < u2(λ, p)
As system gets more congested, the one with higher sensitivity todelay hurts more
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 54 / 86
Service Systems with Strategic Customers
Multiple Customer Classes-identical price
By solving ui(λ, p) = 0 for λ and p
λi(p) =2µ2(zi − p)
2µ(zi − p) + ci(1 + cv2)
pi(λ) = zi −λci(1 + cv2)
2µ(µ− λ)
maximum arrival rate for i for a given price p
maximum price i is willing to pay for total arrival rate λ
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 55 / 86
Service Systems with Strategic Customers
Multiple Customer Classes-identical price
How do customers behave for a given price p?
Depends on c1 and c2
If c1 ≤ c2, then Λ1 and λ2(p) are compared
λ1(p) ≥ λ2(p)
Λ1 ≥ λ2(p) =⇒ (q1e , q
2e) = (min{1, λ1(p)/Λ1}, 0)
Λ1 < λ2(p) =⇒ (q1e , q
2e) = (1,min{1, (λ2(p)− Λ1)/Λ2})
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 56 / 86
Service Systems with Strategic Customers
Multiple Customer Classes-identical price
Suppose Λ1 ≥ λ1(p) ( =⇒ ≥ λ2(p))
First: Class-1 enters with rate λ1(p)
u2(λ1(p), p) ≤ u1(λ1(p), p) = 0
=⇒ q2e = 0
Next: Class-2 does not enter the system
u1(λ, p) > 0 ⇐⇒ λ < λ1(p)
u1(λ, p) is decreasing in λ
=⇒ q1e = λ1(p)/Λ1
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 57 / 86
Service Systems with Strategic Customers
Multiple Customer Classes-identical price
How the server provider sets the price?
maxp λ(p)p
If c1 ≤ c2, then depends on the market size of Type-1
Λ1“large” =⇒ pe = max{p∗1, p1(Λ1)}
and (q1e , q
2e) = (min{1, λ1(p∗1)/Λ1}, 0)
p∗1 = arg max pλ1(p)
p∗1 = z1 −√c1(1 + cv2)(c1(1 + cv2) + 2µz1)− c1(1 + cv2)
2µ
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 58 / 86
Service Systems with Strategic Customers
Zhou, W., Chao, X., Gong, X., 2014, Optimal Uniform Pricing Strategy ofa Service Firm when Facing Two Classes of Customers, POM, 23, 676-688.
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 59 / 86
Service Systems with Strategic Customers
Multiple Customer Types-differentiation
M customer types
Ri, ci, ΛiSome questions:
what will be the “control policy”what will be the price/delay menu?what is the information structure: who knows what?
Incentive compatibility
pi + ciwi ≤ pj + ciwj j 6= i
individual rationalityRi ≥ pi + ciwi
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 60 / 86
Service Systems with Strategic Customers
Multiple Customer Types-Revenue Maximization Model
maxp,W,u
∑Mi=1 piλi
λi = Λi Pr{Ri ≥ pi + ciwi} ∀ ipi + ciwi ≤ pj + ciwj j 6= i
M∑i=1
λi < µ
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 61 / 86
Service Systems with Strategic Customers
Multiple Customer Types-General Results
If the service provider observes the types of the customers
Set priorities with a work conserving discipline (cµ rule)
If the service provider does not observe the types
Set priorities with strategic delaysWork conserving discipline may not be optimalDelay cost minimization is not the dominant criterionStrategic delay (for low priority items) deters high priority customerspurchasing a low priority menuThis accomplishes incentive compatibility
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 62 / 86
Service Systems with Strategic Customers
Multiple Customer Types-An example (Allon, 2010)
M/M/1 queue with µ = 1
λ1 = 0.2, R1 = 100, c1 = 20
λ2 = 0.3, R2 = 30, c2 = 4
Under cµ rule, w1 = 1/(µ− λ1) = 1.25,w2 = 1/(µ(1− ρ1)(1− ρ1 − ρ2)) = 2.5
p1 = R1 − c1w1 = 100− 20(1.25) = 75p2 = R2 − c2w2 = 30− 4(2.5) = 20Revenue: 0.2(75) + 0.3(20) = 21
Overall delay cost=2.5(4) + 1.25(20) = 35
Not IC: 75 + 20(1.25) > 20 + 20(2.5) = 70
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 63 / 86
Service Systems with Strategic Customers
Multiple Customer Types-An example
Highest IC price under cµ:
p1 = p2 + c1w2 − c1w1 = 20 + 20(2.5)− 20(1.25) = 45
with revenue=45(0.2) + 20(0.3) = 15
If one can have w2 ← 2.5 + 1 = 3.5
Price for Type-2 becomes: p2 = 30− (3.5)4 = 16
IC Type 1 price
p1 = 16 + 20(3.5)− 20(1.25) = 61
Revenue=61(0.2) + 16(0.3) = 17.
Overall delay cost=3.5(4) + 1.25(20) = 39
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 64 / 86
Service Systems with Strategic Customers
A manufacturing and service system
A common part is used for service activity of heterogenous customers
AudiA4 and VW Passat use the same engine, transmission and someother featuresDesign for after-sales-service
Customers have different sensitivity for waiting and service valuation
Service provider keeps a common spare parts inventory
Operates with a base stock policy (base stock level y)
Parts are replenished through a finite capacity system
M/M/1 but can be generalized
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 65 / 86
Service Systems with Strategic Customers
A manufacturing and service system
Whenever there is on-hand stock, customer demand is satisfiedirrespective of the type
If on-hand stock is zero, customers have to wait
Non-preemptive priorities
A customer is tagged with an incoming part (irrespective of the type)
y = net inventory + outstanding parts (N)
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 66 / 86
Service Systems with Strategic Customers
An Illustration
Manufacturer
N=0
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 67 / 86
Service Systems with Strategic Customers
An Illustration
Manufacturer
N=1
W=0
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 68 / 86
Service Systems with Strategic Customers
An Illustration
Manufacturer
N=1
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 69 / 86
Service Systems with Strategic Customers
An Illustration
Manufacturer
N=2
W=0
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 70 / 86
Service Systems with Strategic Customers
An Illustration
Manufacturer
N=2
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 71 / 86
Service Systems with Strategic Customers
An Illustration
Manufacturer
N=3
W=0
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 72 / 86
Service Systems with Strategic Customers
An Illustration
Manufacturer
C1
N=4
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 73 / 86
Service Systems with Strategic Customers
An Illustration
Manufacturer
N=5
C1C2
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 74 / 86
Service Systems with Strategic Customers
An Illustration
Manufacturer
N=5
C1C2
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 75 / 86
Service Systems with Strategic Customers
An Illustration
Manufacturer
N=6
C1C2C3
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 76 / 86
Service Systems with Strategic Customers
An Illustration
Manufacturer
N=5
C1C2
W>0
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 77 / 86
Service Systems with Strategic Customers
The service provider determines
base stock level yprice menu p = (p1, . . . , pM )the priority scheme
Customers react by arriving with λ = (λ1, . . . , λM )
Ri = pi + ciE[waiting time]
i = 1, 2, . . . ,M
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 78 / 86
Service Systems with Strategic Customers
The distribution of the outstanding parts
M/M/1: Pr{N = k} = ρk(1− ρ)
ρ =∑Mi=1 λi/µ
By PASTA property
E[waiting with y ≥ 0] = Pr{N ≥ y}E[waiting in a standard queue]
= ρyE[waiting in a standard queue]
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 79 / 86
Service Systems with Strategic Customers
The service provider’s problem
maxy,λ,u
{M∑i=1
λiRi − ρyM∑i=1
ciλiwi − hE[(y −N)+]
}
u = (u(1), u(2), . . . , u(M)) the priority order
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 80 / 86
Service Systems with Strategic Customers
Main Results
Restricted to work-conserving disciplines:
cµ rule is optimal: c1 ≥ c2 ≥ · · · cMGiven λ, optimal base stock level:
y∗(λ) = min
{y ≥ 0 : ρy+1 ≤ h
h+ vH(λ)
}
v = (1− ρ)/ρ, H(λ) =
M∑i=1
ciλiE[waiting(λ)]
Prices given by the first order conditions are incentive compatible
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 81 / 86
Service Systems with Strategic Customers
A reduction:
If Ri ≥ Rj and ci ≤ cj , then Type-i dominates Type-j
λ∗j = 0 (p∗j = Rj)
ci = c, Rk = max{Ri} or Ri = R, ck = min{ci}
maxy≥0,ρ∈[0,1)
{Rµρ− cρy ρ
1− ρ− h(y − ρ
1− ρ(1− ρy))
}y(ρ) = min{y ≥ 0 : ρy+1 ≤ h/(c+ h)}
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 82 / 86
Service Systems with Strategic Customers
A continuous approximation: Pr{N ≥ y} ≈ e−vy
Variable Value ∂/∂h ∂/∂c
λ µ−√µK/R ↓ ↓
y(√RµK −K
)/h ↓ ↑, ↓
E[W ]
√R/
(µK) ↓ ↓
p R− c√R/
(µK) ↑ ↓
K h log(1 + c/h)
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 83 / 86
Service Systems with Strategic Customers
K = h log(1 + c/h)→ c as h→∞Compare p above with
p = R−
√cR
µ= R− c
√R
µc
Optimal profit: (√Rµ−
√K)2
K < c =⇒ (profit with y > 0) ≥ (profit with y = 0)
Attracts more demand (with smaller price) and achieves a higher totalprofit.
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 84 / 86
Service Systems with Strategic Customers
Mendelson, H., Wang, S., 1990, Optimal Incentive-Compatible PriorityPricing for the M/M/1 Queue, Operations Research, 38, 870-883.
Afeche, P., 2013, Incentive-Compatible Revenue Management in QueueingSystems: Optimal Strategic Delay, MSOM, 15, 423-443.
Allon, G., 2010, Pricing and Scheduling Decisions, Wiley Encyclopedia ofOperations Research and Management Science edited by James J.Cochran, Wiley.
Maglaras, C., Yao, J., Zeevi, A., 2013, Optimal Price and DelayDifferentiation in Queueing Systems, Working Paper.
Guler, G., Bilgic, T., Gullu, R., 2014, Joint Inventory and Pricing Decisionswhen Customers are Delay Sensitive, to appear in IJPE.
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 85 / 86
Service Systems with Strategic Customers
Some less explored topics
Alternative cost and reward structures
“willingness to wait”: E[(W −WtoW )+]
Correlation of R and c
Competition and cooperation among service providers
Distribution free bounds
maxλ
minfR
λE[(R− cw(λ))+]
Estimation errors in parameters
Bounded rationality (Huang, Allon and Bassamboo, 2014)
Refik Gullu (Bogazici University) YEQT VIII Eindhoven 86 / 86