msom china
TRANSCRIPT
-
8/8/2019 Msom China
1/37
Outline
Introduction
The Model
Results
Multi-ChannelPricing
Computational
Analysis
Conclusionsand OngoingResearch
Optimal Pricing and Production Planning forSubscription-Based Products
W. T. Huh S. Kachani A. Sadighian
Department of Industrial Engineering and Operations Research,Columbia University, New York
MSOM Annual Meeting - Beijing, ChinaJune 18-19, 2007
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
-
8/8/2019 Msom China
2/37
Outline
Introduction
The Model
Results
Multi-ChannelPricing
Computational
Analysis
Conclusionsand OngoingResearch
Outline
1 Introduction
MotivationLiterature Review
2 The ModelDynamics of the model
DP Representation3 Results
One Stage RepresentationMonopolyDuopoly
4 Multi-Channel Pricing5 Computational Analysis
Symmetric FirmsAsymmetric Firms
6
Conclusions and Ongoing ResearchHuh, Kachani, Sadighian Optimal Pricing and Production Planning
-
8/8/2019 Msom China
3/37
Outline
IntroductionMotivation
LiteratureReview
The Model
Results
Multi-Channel
Pricing
ComputationalAnalysis
Conclusionsand OngoingResearch
MotivationSubscription-Based Products (Magazines)
Specific business model:
Subscriber BaseAdvertisement Sales vs. Product Sales
Example: Magazine Industry - Columbia Journalism
Review (CJR) one of the leading Journalism Magazines inUS
Sources of Revenue:
Newsstand single copy SalesSubscription Revenue
Advertisement Sales
Circulation revenue for 2006 > $10 billion
Advertisement revenue: 54% of total salesMagazine sales through subscription: 32% of total sales
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
-
8/8/2019 Msom China
4/37
Outline
IntroductionMotivation
LiteratureReview
The Model
Results
Multi-Channel
Pricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Literature Review - Quantity and Pricing Models
Environment: Both price and quantity decisions
Inventory Replenishment Problem
Main theme: Choice of correct quantity to
maximize(minimize) revenue(cost) in multiple periodsOne special case: Newsvendor
Models for join quantity-price decision making
Petruzzi and Dada (1999) Price-setting newsvendorFedergruen and Heching (1999) single retailer withstochastic demand, comprehensive model on monopolist(refer for further literature review)
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
-
8/8/2019 Msom China
5/37
Outline
IntroductionMotivation
LiteratureReview
The Model
Results
Multi-Channel
Pricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Literature Review - Competition Models
Competitive Newsvendor
Parlar(1988), later Lippman and McCardle (1997), Firmscompeting over the overflow of the unmet demand of theother firms
General Existence results
Kirman and Sobel (1974): No spill-over, backlog or lost,demand as a function of priceBernstein and Federgruen (2004): Demand as a functionof price and a second measure service level
Ahn and Olsen (2005): Magazine problem, built onsupermodular games, prove existence of a Markov PerfectEquilibrium
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
-
8/8/2019 Msom China
6/37
Outline
IntroductionMotivation
LiteratureReview
The Model
Results
Multi-Channel
Pricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Goals
Develop a model for joint pricing and production planningfor subscription-based products
Model should capture both stand-price andsubscription-price decisions
Model should yield either closed-form or computationallytractable solutions
Extend our model to the duopoly setting
Extend our model to address multi-channel pricing of
subscription-based products with consumers choiceTest our results numerically and apply our work to themagazine industry in New York area
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
-
8/8/2019 Msom China
7/37
Outline
Introduction
The Model
Dynamics of themodel
DPRepresentation
Results
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
1 Introduction
MotivationLiterature Review
2 The ModelDynamics of the modelDP Representation
3 ResultsOne Stage RepresentationMonopolyDuopoly
4 Multi-Channel Pricing5 Computational Analysis
Symmetric FirmsAsymmetric Firms
6 Conclusions and Ongoing Research
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
-
8/8/2019 Msom China
8/37
Outline
Introduction
The Model
Dynamics of themodel
DPRepresentation
Results
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Dynamics of the Model
Firm selling its product in multiple periods
Independent demand in each period
Decisions in each period
Newsstand Quantity (qt)
Subscription Price (st)Newsstand Price (pt)
Subscription Mechanism
Conversion of newsstand sales to subscribers t()new subscribers = tE[min(qt, Dt)]
Attrition rate t()leaving subscribers = (1 t)nt
Subscription dynamicsnt(qt, nt1) = (1 t)nt1 + tE[min(qt, Dt)]
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
-
8/8/2019 Msom China
9/37
Outline
Introduction
The Model
Dynamics of themodel
DPRepresentation
Results
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
DP Representation
Revenue and Cost Structure
Variable Cost: c(nt + qt)Advertisement and Subscription Revenue: (st + m)nt (madvertisement revenue)
Newsstand Revenue: ptE[min(qt, Dt)]
Profit in each period
E[ptmin(qt, Dt) c(qt + nt)] +t
k=1(sk + m)ntk
Analysis done for finite and infinite horizon cases
Finite horizon case has a linear salvage value v + nvChallenge: The profit generated in period t depends onprevious decisions
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
-
8/8/2019 Msom China
10/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
1 Introduction
MotivationLiterature Review
2 The ModelDynamics of the modelDP Representation
3 ResultsOne Stage RepresentationMonopolyDuopoly
4 Multi-Channel Pricing5 Computational Analysis
Symmetric FirmsAsymmetric Firms
6 Conclusions and Ongoing Research
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
-
8/8/2019 Msom China
11/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
One Stage Representation
Key Observation: Revenue in each time period is aseparable function of previous decision
Proposition: We can transform the problem to a series ofseparable independent problems
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
M l
-
8/8/2019 Msom China
12/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
MonopolyQuantity Decision
Decision variables:1 Production Quantity qt
The problem can be reduced to a newsvendor problem in
each period with no carry over of inventoryRevenue: pt +
t(st+mc)1(1t)
Cost: c
Structure in the optimal quantity values:
Decreasing subscription prices leads to decreasing optimal
production quantityIncreasing planning horizon increases the optimalproduction quantity
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
M l
-
8/8/2019 Msom China
13/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
MonopolyQuantity and Subscription Price Decision (1)
Decision variables:1 Production Quantity qt
2 Subscription Price st
Attrition rate t() and conversion rate t() are notconstant anymoret = t(st) and t = t(st)
Two interpretations for t
Macro: Percentage of people subscribingMicro: Probability of one person joining
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
M l
-
8/8/2019 Msom China
14/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
MonopolyQuantity and Subscription Price Decision (2)
Under certain regularity condition for t and t, theoptimal subscription price can be determinedindependently of optimal production quantity
If advertisement revenue is strictly positive, it is possiblethat the optimal subscription price is 0 but in absence ofadvertisement revenue optimal subscription price ispositive
After determining optimal st
finding optimal productionquantity reduces to the previous case
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
M l Q tit
-
8/8/2019 Msom China
15/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Monopoly - Quantity,Subscription Price and Stand Price Decision (1)
Decision variables:1 Production Quantity qt
2 Subscription Price st
3 Stand Price pt
In this new setting conversion rate t() is assumed to beboth a function of stand price pt and subscription price st
The same one-stage representation can be achieved
Optimal Subscription Price:Assumption: Dependency of subscription price on stand
price and subscription price is of the multiplicative formt(st, pt) = t1 (st)t2 (pt)
Proposition: the choice of optimal subscription price isthe same as monopoly case without stand price decision
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
M l Q tit
-
8/8/2019 Msom China
16/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Monopoly - Quantity,Subscription Price and Stand Price Decision (2)
Transformation for one stage representation
Concept of Virtual Price pt :
pt pt + t(st,pt)(st+mc)1(1t(st))
Type of random demand on newsstand:Additive: t(pt) = a bpt +
Multiplicative: (p)t = aptb
Assuming certain functional forms for (), demand as a
function of Virtual price can be written in additive ormultiplicative form too. Thus, the problem is reduced to aprice setting newsvendor problem.
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Duopoly
-
8/8/2019 Msom China
17/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Duopoly -General Setting
Two firms competing over unsatisfied demand
New demand function:
Dti = ti + i(
tj q
tj )
+
Each firm has a loyal customer base tiSecond part of demand function: the portion of unsatisfieddemand that switch to a new product
The same micro and macro level interpretation exists for i
In this setting the demand for firm i is a function of(pti , qt
j) and indirectly pt
j
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Duopoly
-
8/8/2019 Msom China
18/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Duopoly -Quantity Decision (1)
Proposition: The one stage representation of the problemstill holds with modification of demand function
Expected value of sales will be adjusted to
Emin(qt1,t
1+ 1(
t
2 qt
2)+)
We will assume uniform demand to derive closed formsolution
Using the one stage representation the problem ofcompeting firms can be reduced to the problem of two
competing newsvendorsWe use results from Lippman and McCardle (97) tocharacterize the equilibrium
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Duopoly
-
8/8/2019 Msom China
19/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Duopoly -Quantity Decision (2)
Notation i = ci/pi
The equilibrium point for the two competing newsvendoris the pair (q1 , q
2) iff (q
1, q
2) satisfy
P(Di > qi) = i
Notice that i is the ratio of cost to the virtual price
Using the above result we derive the equilibrium point
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Duopoly Quantity Decision
-
8/8/2019 Msom China
20/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Duopoly - Quantity DecisionTypes of the Quantity Game Equilibria (1)
Type of resulting equilibrium depends on and parameters
If the cost structure of the two firms is not very different
the emerging equilibrium is symmetric (They use thesame type of best response functions)
Types of equilibria (type of best response functions):
Type I (symmetric): The competing firms produce withinthe support of their loyal customers demand (i)Type II (asymmetric): One of the competing firm producesmore than the upper bound of its loyal customers demand
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Duopoly -
-
8/8/2019 Msom China
21/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Duopoly -Types of the Quantity Game Equilibria (2)
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 11
2 Type I(Symmetric)
Type II
Type II
(i constant)
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Duopoly -
-
8/8/2019 Msom China
22/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Duopoly Types of the Quantity Game Equilibria (3)
Interpretation of value as degree of separation ofmarkets:
Recall Dti = ti + i(
tj q
tj )
+
Proposition: For any combination of (c1, p1) and (c2, p2)there exists a = 1 = 2 such that resulting equilibria issymmetric for all
If firms operate separate markets the symmetric type of
equilibrium will emerge, meaning they will use the sametype of best response
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Duopoly - Convergence of
-
8/8/2019 Msom China
23/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Duopoly Convergence ofBest Response Functions in Quantity Game
Proposition: If the firms start from a non-equilibriumpoint a tatonnement scheme using their best responsefunctions will eventually converge to the equilibrium point
Moreover the convergence rate when the best responsefunctions are symmetric is linear
Implication of the above result:
If the firms start with a limited view of what they shouldproduce (one might dismiss the possibility of producing
above its loyal customer base demand) the best responsefunctions will eventually lead them to the equilibriumpoint!
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Duopoly -
-
8/8/2019 Msom China
24/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Duopoly Quantity and Subscription Price Quantity Game (1)
As in monopolist case we consider the case werecompetitors decide on their subscription price too
The equation governing optimal choice of subscription
price is:1(s1)
s1+m1c1
1 (s1)1(s1)1(11(s1))
+ 1(s1)
= 0
The optimal subscription price can be determined beforeentering the competition
Optimal subscription price independent of quantity decisionOptimal subscription price independent of other firmsquantity and subscription price decision
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Duopoly -
-
8/8/2019 Msom China
25/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Duopoly Quantity and Subscription Price Quantity Game (2)
After setting the subscription rates the problem reduces tothe previous case
Proposition: Two firms competing with quantity andsubscription price first set their subscription prices like amonopolist and then choose their production level as theprevious case. This decision will determine the uniqueequilibrium point.
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Duopoly - Effect of
-
8/8/2019 Msom China
26/37
Outline
Introduction
The Model
Results
One StageRepresentation
Monopoly
Duopoly
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
uopo y ect oSubscription Price on the Equilibrium Point
Observation: Recall
Type of emerging equilibria depends on = c/p
pp+ (s,p)(s+mc)1(1(s))
Although choice of s can be done without entering thecompetition, however it will have an effect on the type ofemerging equilibrium.
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Multi-Channel Pricing
-
8/8/2019 Msom China
27/37
Outline
Introduction
The Model
Results
Multi-ChannelPricing
Computational
Analysis
Conclusionsand OngoingResearch
gGeneral Setting (1)
One publishing firm offering the magazine throughdifferent channels (traditional paper-based, electronic,combination)
Consumers enter the system and based on a consumer
choice model (for example multinomial logit choose theirpreferred product)
In each period the consumers re-evaluate their subscriptiontype
There are different types of customers (i.e. different utilityfunction co-efficient with regard to different subscriptiontypes)
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Multi-Channel Pricing
-
8/8/2019 Msom China
28/37
Outline
Introduction
The Model
Results
Multi-ChannelPricing
Computational
Analysis
Conclusionsand OngoingResearch
gGeneral Setting (2)
Decision
Print
Online
Print
+
Online
ot
ct
srt
(Pst
)
ort(Po
t)
crt(Pc
t)
Pst
Pot
Pct
st
Demand
Exit
Nnt
Nrt
Nt
e
t
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Multi-Channel Pricing
-
8/8/2019 Msom China
29/37
Outline
Introduction
The Model
Results
Multi-ChannelPricing
Computational
Analysis
Conclusionsand OngoingResearch
gChallenges
Number of subscribers for each product depends on theprice decisions in all previous periods.
The dynamic programming formulation is highly non-lineareven when simple linear functions are chosen for utilityfunctions.
Because of limited pool of customers not only the pricebut production quantity decision will also impact the
future states directly.
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
-
8/8/2019 Msom China
30/37
Outline
IntroductionThe Model
Results
Multi-ChannelPricing
Computational
AnalysisSymmetricFirms
AsymmetricFirms
Conclusionsand OngoingResearch
1 Introduction
MotivationLiterature Review
2 The ModelDynamics of the modelDP Representation
3 ResultsOne Stage RepresentationMonopolyDuopoly
4 Multi-Channel Pricing5 Computational Analysis
Symmetric FirmsAsymmetric Firms
6 Conclusions and Ongoing Research
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Effect of cost on value of considering the
-
8/8/2019 Msom China
31/37
Outline
IntroductionThe Model
Results
Multi-ChannelPricing
Computational
AnalysisSymmetricFirms
AsymmetricFirms
Conclusionsand OngoingResearch
competition - Case of similar firms
0%
10%
20%
30%
40%
0 1 2 3 4 5Cost
%o
fProfitImprovementby
ConsideringtheCompetition i= 1 i= 0.8
i = 0.6
Increase in cost of production increases the value ofconsidering the competition
m as bs ap bp a b p
13 0.1 0.005 0.3 0.001 0.05 0.001 0.9 6
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Effect of stand price on value of considering the
-
8/8/2019 Msom China
32/37
Outline
Introduction
The Model
Results
Multi-ChannelPricing
Computational
AnalysisSymmetricFirms
AsymmetricFirms
Conclusionsand OngoingResearch
competition - Case of similar firms
Increase in stand price decreases the value of consideringthe competition
m as bs ap bp a b c
13 0.1 0.005 0.3 0.001 0.05 0.001 0.9 3
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Who benefits more? Asymmetric firms
-
8/8/2019 Msom China
33/37
Outline
Introduction
The Model
Results
Multi-ChannelPricing
Computational
AnalysisSymmetricFirms
AsymmetricFirms
Conclusionsand OngoingResearch
Asymmetry in cost
Firms differ in their costs
Considering thecompetition has a highervalue for the firms whichare operating with highercost
m as bs ap bp13 0.1 0.005 0.3 0.001a b c1 p
0.05 0.001 0.9 0.8 3 6
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Who benefits more? Asymmetric firms
-
8/8/2019 Msom China
34/37
Outline
Introduction
The Model
Results
Multi-ChannelPricing
Computational
AnalysisSymmetricFirms
AsymmetricFirms
Conclusionsand OngoingResearch
Asymmetry in customer attraction power
Firms differ in their values, interpret ascustomer attraction power
Considering thecompetition has a highervalue for the firms otherfirms customers are lessloyal
m as bs ap bp13 0.1 0.005 0.3 0.001a b 1 c p
0.05 0.001 0.7 0.8 3 6
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
-
8/8/2019 Msom China
35/37
Outline
Introduction
The Model
Results
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
1 Introduction
MotivationLiterature Review
2 The ModelDynamics of the modelDP Representation
3 ResultsOne Stage RepresentationMonopolyDuopoly
4 Multi-Channel Pricing5 Computational Analysis
Symmetric FirmsAsymmetric Firms
6 Conclusions and Ongoing Research
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
Conclusions and Ongoing Research
-
8/8/2019 Msom China
36/37
Outline
Introduction
The Model
Results
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Conclusions and Ongoing Research
Presented a model for joint pricing and productionplanning for subscription-based products
Model captured both stand-price and subscription-pricedecisionsModel yielded closed-form and computationally tractable
solutions that brought insight into the problemExtended our model to the duopoly setting
Presented encouraging numerical results
Currently implementing our work at Columbia Journalism
Review (CJR)Will enhance the model to incorporate differentdistribution channels (e.g. online) and effect ofpromotional activities on subscription base
Huh, Kachani, Sadighian Optimal Pricing and Production Planning
-
8/8/2019 Msom China
37/37
Outline
Introduction
The Model
Results
Multi-ChannelPricing
ComputationalAnalysis
Conclusionsand OngoingResearch
Optimal Pricing and Production Planning forSubscription-Based Products
W. T. Huh S. Kachani A. Sadighian
Department of Industrial Engineering and Operations Research,Columbia University, New York
MSOM Annual Meeting - Beijing, ChinaJune 18-19, 2007
Huh, Kachani, Sadighian Optimal Pricing and Production Planning