Professional Documents
Culture Documents
P. Agrell
C3: DK Dairy Giants: We’ll have asymmetric information. One page brief to write
based on the assessment of the scenario and the asymmetric information.
Afterwards, we’ll make some other teams. This is a two-session case.
C4: Make-Up-Task – voluntary task that replaces the lowest grade on any other task.
Deadline: the end of the semester (20th).
6.1 Introduction
Supply Chain members are primarily concerned with optimizing their own objectives.
Yet, an optimal supply chain performance requires the execution of a precise set of
action that is not always in the best interest of the members in the supply chain.
However, optimal performance is achievable if the firms coordinate by contracting on
a set of transfer payments such that each firm’s objective becomes aligned with the
supply chain objective.
We’re going to discuss supply chain models. In each model, the supply chain optimal
actions are feasible for each firm. However, firms lack the incentive to implement
those actions. To create that incentive the firms can adjust their terms of trade via a
contract that establishes a transfer payment scheme.
The Norwegian government built a military aircraft production during World War II.
After the war, they had too much production capacity. They first made a joint venture
and developed a 34-seats aircraft (secondary and high frequency lines). Afterwards,
they developed a 50-seats airplane for other markets, ones with lower frequency
lines. At that time, the aircrafts are MTO – Made to Order – and take 6 months to
produce.
Problems and solution: (1) market sluggish1 – buyers require a lead-time of 3 months
instead of 6. This is made possible because of bargaining power2 (if you can do it,
alright, if you don’t I’ll go somewhere else). To solve this problem, the policy had to be
changed from MTO to MTS and ATO – Assembled to Order. This implies the
1
Sluggish: lent
2
How can the bargaining power of the buyers go up? There are some competitors that can produce
on forecast, which means they can build quicker for the buyer.
1
suppliers have to produce on forecast, but they do not agree, they want to make it on
an order. SAAB has to forecast and order to the suppliers, which means they have to
carry the risks. Once you do that, what happens to the market? The aircraft producer
doesn’t care about long-term relations with the supplier. The buyers slow down their
orders wait to the very last minute so that the aircraft producer has to sell his aircrafts
to a smaller price. The buyers push it so far, that they don’t want to buy it anymore:
“How about you make one for me – with my logo on it – and then you keep it and I
lease it? If I don’t need it, if it’s not filled up enough, then you take it back, free of
charge.” This is the Rolls Royce of contracts. This means the seller takes the risks
from the entire supply chain, from the suppliers to the final consumer. Is this
sustainable? No, that’s why in 1998 the decision was taken to cease production. So,
what went wrong here? Whose fault is it? SAAB Management? Suppliers? Market?
State? The State had a role because competitors survived the crisis while SAAB
didn’t because they were subsided. On the other side, competitors had some other
strategies. Fokker, for example, split everything in one military part and one civilian
part. The military part went fine while the civilian part didn’t as well (See graph). This
means, that even when the civilian part sales goes down, Fokker still sold in the
military part. Why were the suppliers so strict with SAAB and not to another one?
SAAB had violated here one of the principles of coordination. They had everything
set up and planned and the suppliers knew about it. They accepted a 10% cut in the
first place because they had made their forecasting on forehand and knew that there
were enough margins to accept the 10% cut. We call this a hold-up.
There was no risk sharing mechanisms, there was no trust anymore in how it was
going to work, it was a chicken race, and somebody had to turn the steering wheel.
This story indicates that it is not only important to optimize production but the
company has also to be sure that the 3 parts (production, …, …) have to be
coordinated well.
“The best Supply Chains aren’t just fast and cost-effective. They are also agile and
adaptable, and they ensure that all their companies’ interests stay aligned.”
Hau L. Lee
The major Supply Chain Management focus in 1980-1990 was to be efficient, lean
meaning supply chains are able to grow worldwide. Everything was about maximizing
speed and minimizes supply costs. Yet, those “efficient” supply chains aren’t able to
deliver the goods for several reasons: (i) high-speed, low-cost supply chains are
unable to respond to unexpected changes in demand or supply; (ii) centralization of
manufacturing and distribution facilities (scale economies) implies deliveries with
container loads of products making it impossible to ship products when demand
rises. This results in stock-out stores and turning away consumers. When the
container finally arrives, markdowns are needed to sell the whole container of
products, way too much regarding what the supermarket really needed to satisfy the
fluctuating demand.
2
Besides, the failure of those “efficient” supply chains to deliver lasting positions is due
to (i) higher competition in markets, (ii) faster production innovation, (iii) shorter
product life cycles by technological innovation and (iv) a greater interdependency
between markets.
Top-performing supply chains possess three different qualities: (1) agility – they
react speedily to sudden changes in demand or supply; (2) adaptability – they adapt
over time as market structures and strategies evolve; (3) alignment – they align the
interests of all the firms in the supply network so that companies optimize the chain’s
performance when they maximize their interests.
Agility
- Design products so that they share common parts and processes initially and
differ substantially only by the end of the production process (mass
customization). It’s the best way to adapt to consumers preferences and
respond quickly to demand fluctuations.
Adaptability
It is the ability to adjust the supply chain’s design to meet structural changes in
markets, strategies, products and technologies. It is about adapting your product
portfolio to whatever you need. You adapt your supply chain not only in volume but
also in the way of doing it. For example, offer goods and services and not only goods.
The best supply chains identify structural shifts, sometimes before they occur, by
capturing the latest date, filtering out noise, and tracking key patterns. They then
relocate facilities, change sources of supplies, and, if possible, outsource
manufacturing.
Building an adaptable supply chain requires two key components: the ability to spot
trends and the capability to change supply networks.
3
Alignment
Great companies take care to align the interest of all the firms in their supply chain
with their own. That’s critical because every firm tries to maximize only its own
interests. In any company’s interests differ from those of the other organizations in
the supply chain, it actions will not maximize the chain’s performance.
One way companies align their partners’ interests with their own is by redefining the
terms of their relationships so that firms share risks, costs, and rewards equitably.
Although such arrangements require trust and commitments on the different parts of
the supply chain, it is a powerful way to align the interests of companies in supply
chains.
Today, the production innovation is faster and faster and product life is getting shorter
and shorter, you’re relying more and more on suppliers. This means, even if you’re a
lean producer you can go out of the market if you don’t adapt to the market structure.
Clever companies create alignment in supply chains in several ways: (i) alignment of
information, so that all the companies in a supply chain have equal access to
forecasts, sales data, and plans; (ii) alignment of identities – the manufacturer must
define the roles and responsibilities of each partner so that there is no scope for
conflict; (iii) alignment of incentives, so that when companies try to maximize returns,
they also maximize the supply chain’s performance. To ensure that happens,
companies must try to predict the possible behavior of supply chain partners in the
light of their current incentives. Then they must redesign incentives so partners act in
ways that are closer to what’s best for the entire supply chain.
Stockout Problems
Stockout problems are major problems for the front end of the supply chain. The front
end is were the product made in the chain is sold. The average stockout ratio is
12,7%. Is that a small number? It’s an enormous number. Sellers invest in people
knowing a product and when the buyer wants to buy the product, he can’t find it. In
toys department, the stockout is of 20%.
The average stockout ratio is 12,7% for an estimated annual opportunity cost of 367
M€.
Remark: The opportunity cost represents the loss, including the damage to the
brand.
4
Is it just a question of forecasting?
What will happen when we have a stockout on the right side (side of brands),
customers will try products on the left side (white brand products). The retailer has
the largest margin on the left side. It is a game where part of the opportunity cost is
the fact that we invest in R&D, develop new products etc. but if the product cannot be
found, people will buy retailers products. The distributor has to take care of
everything from the beginning till the end, because the retailer is not going to make
sure they have enough of their products because then they can sell theirs. You don’t
need to coordinate only your chain but the entire sector to make sure buyers do buy
your products. In the pharmaceutical sector, the drug producer needs to tell buyers
that the other product (generic drugs) are bad (generic distributors don’t take any
responsibility on bad products because they have no brand). It’s their responsibility.
Each parties are concentrated on their own interests: (1) Distributors purchase
quantities to take advantage on price promotion and (2) manufacturer’s divers are
paid commission based on the sales generated at each store they serviced.
There is something from within the structure. Imagine that Sony has a lot stock and is
selling it to agents and each agent is selling to retailers. Let’s say that I give agents
discount (the more you buy, the less you pay). What is the problem? If we don’t set it
up correctly, the chain is going to be untrusted. Agent 1 can say that because he likes
the retailer he can give him a good price, this means he pushes. The retailer is not
going to buy those products because he doesn’t trust the agent. Thus, Sony is not
going to sell to agent 1 anymore and one of the chains will be underexploited.
Contracts
Contract components
5
(4) Quantity flexibility: if you want quantity flexibility downstream, you need
flexibility along the whole chain.
(5) Buyback or returns policies
(6) Allocation rules: if you have rationing, because you don’t have enough, the
contract specifies how the product is going to be allocated to the different
buyers.
(7) Lead time
(8) Quality
Limitations to Coordination
Roles of SC Contracts
(1) Coordination
(2) Rent allocation: business risks – profit allocation
3
Monitoring: analyser, surveiller
6
The principal and the agent are linked by a contract but there are also signals,
information that are shared from one to another that’s isn’t part of the contract. On
the other hand, the agent is taking private action 𝑎 that we cannot observe. The
outcome 𝑥 of the production comes back to the principal. So what’s the contract? It’s
not a good idea to sign a contract on action 𝑎 but we can link everything to the
outcome 𝑥. That means I pay you afterwards. I don’t care how you did it or in how
much time but I do care about the outcome which is fully observable. Thus the agent
takes the risks: “If you deliver the right stuff, I will pay the right amount.”
Outcome 𝑥 is not only a function of action 𝑎 but also of the state of nature
𝑞→𝑥 = 𝑓(𝑎, 𝑞). The state of nature 𝑞 is influenced by contingencies. For example – a
patient at the dentist: the average patient takes 15 minutes to be fixed, but you can
have one that takes 6 days to fix, so 𝑞 is a risk the patient takes.
The dentist will not sign a contract based on the outcome because the outcome –
that you don’t suffer anymore – is only observable from one side, unilaterally. The
challenge is to find something that can make everyone optimize his or her own
function.
You cannot make a contract about something that is not observable on both sides but
on top of that it has to be verifiable – can I measure it, find information about it
(Example: the quality of food is not measurable) and enforceable – is the measure I
made good enough to put it in a contract and go to court if needed. For example, can
you prove that something is quality? Can you go to court and say that the quality is
bad? Thus, an action or outcome is contractible if it is observable, verifiable and
enforceable.
𝑥 = 𝑥(𝑎)
7
The agent will here choose the action 𝑎 that is optimal for him, given the incentive
scheme 𝑠(𝑥).
(1) Moral Hazard: When one firm can take actions that affect (negatively) the
entire chain, but those actions are not contractible. The agent is being
opportunistic post-contract. He follows his own interests and not the supply
chains interest.
Imagine a car manufacturer selling cars through independent car dealers in the USA.
What’s the advantage of doing it with independent car dealers? (In Europe,
you have concessions per regions and they make an agreement on who can
sell to which region) The big advantage is that it’s cheaper to go into the
market. Now imagine Renault sells cars to the independent car dealers
through agents. The car dealer (i) sells new cars (Renaults are imported in the
US easily), (ii) sells used cars and (iii) takes care of the after-sales service.
The profit of the car dealer is his profit minus his private costs. What is the
most expensive part of its activity? The after-sales service is a huge cost
(trained personnel etc.) but this cost is not observable, it’s messy, costly and
not in the interest of the car dealer. The car dealer will try to minimize the
after-sale service. The car dealer can say that the car, Renault brand, is of
very poor quality and that it has nothing to do with its garage. To the customer
coming back, 2 years later with a damaged car, they can say they stopped
selling Renaults and do no offer an after-sale service anymore.
Is there a solution to moral hazard? You have to connect it with a fixed investment to
have mono-brand dealers, strongly committed to that single brand which
makes it difficult to switch to another brand because it will need a high
reinvestment.
8
(2) Idiosyncratic investments: When one firm invests in an asset that has value
only with respect to its relationship with a specific supply chain partner. This is
the SAAB problem.
(3) Adverse selection: When the two firms have different levels of information
about a key contract parameter prior to entering into the contract. Before I
sign the contract I know something about myself that the others do not know.
Contraband and the dolphin-free tuna are adverse selection problems.
Example: The supplier has better information about the quality of its product than
does the buyer.
Solutions
(1) Contract-based: We know contracts are never complete. Yet, we will try to
design contracts that give the right incentive to optimize the supply chain. Indeed,
even incomplete contracts may work if they are sound and actors are repeating
interaction.
→ To solve moral hazard: define contractible outcomes that correlate with the right
action.
→ To solve adverse selection: define after-sales provisions to signal quality
(warranties, return-rights…)
(2) Information-based: It is based on the idea that more information could lead to
voluntary coordination. Thereby, the sunshine regulation is an important concept: you
create some kind on incentive for the agents to do what has to be done by giving
information. For example, publicly publish the outcome of benchmarking data on all
companies of a sector to embarrass the least performing entities and to put the best
performing entities in the highlight. The Sunshine regulation comes from the SUN
company. The CEO organizes a meeting with all suppliers and goes through
everything that happens during the year. Doing this, he made public everything
someone could have been doing wrong or well.
The information-based solution is about making investments in observability to
reduce the profitability of moral hazard or adverse selection.
SUNKIST is a tuna brand. They buy fish from a fisherman. The guy on the fishing
boat is the agent (a). The tuna has two different market values. When the agent is
going fishing he may kill some dolphins too. The consumer down the chain is willing
to pay two prices: one dolphin free and the other one not, with the dolphin free price
higher than the other one. How to make sure the dolphin free tuna is effectively
dolphin free? When I give tuna, I cannot see if dolphins were killed or not. It is based
* *
on trust. How to assure process compliance? 𝑎≠𝑎 (𝑎 = 𝑎𝑟𝑔𝑚𝑖𝑛 𝑈(𝑎)). (1) Can we
equip the boat with something to be sure no dolphin is hurt? A GPS? But there are kind of ways to
𝑐𝑎𝑡𝑐ℎ
circumvent the GPS problem. (2) We could use a ration 𝑠𝑜𝑙𝑑 𝑐𝑎𝑡𝑐ℎ but the agent also sells to local
(3) We could give an incentive to the village to denunciate the agents if they
retailers.
are not doing what they should. We have asymmetric information implying a
monitoring problem. The problem is the inobservability of the action. We can create
joint monitoring effect by creating group incentive. The fisher has all incentives not to
respect SUNKISTS will. Yet, there are other fishers. We’re not signing a contract with
one fisherman but with a whole structure. The fishermen jointly produce the fish. How
to translate the incentive? I force the fishermen to coordinate, which is more
expensive so I need to offset them. I will give them a higher price (premium price) as
offset. To be sure everything is going allright, I can have an inspector that comes
inspecting. He doesn’t care about who did what but if he finds something that violates
the contract then the whole group will be punished. They’ll find some other village.
You penalize the entire group. When you penalize the group, what happens to the
monitoring effect? It is higher. The higher the group penalty, the higher the monitoring
effect.
What about other fishermen joining the village? It is possible in the contract because
SUNKIST only pays per ton of fish. But what will the village think about it? The cost of
monitoring those other fishermen is higher because they are further away. So it is
taking a high risk for not more money because SUNKIST is paying per ton. SUNKIST
is never going to give economies of scale so that the fishermen don’t see too big.
Whatever the quantity of fish, the price per ton remains the same.
Monitoring just as a way to police the fishermen is a bit a repressive system. Is there
any positive point? You can internalize problems of one fisherman for the whole
village. Indeed, a problem of one of them is a problem for the whole village. It
increases the cohesion and it limits monitoring costs because you know each other
10
better. The more non-monitory benefits they get from being part of the club, the more
all fishermen will do everything to stay in it. They do want to keep their social state
and benefits.
What is the reputation value? Can we generalize the tuna story? It is the same for
child labor, horse meat etc. The reputation value is sometimes more effective than
inside monitoring. Indeed, in Bangladesh for instance, people are more mobile;
monitoring is less effective because the village is less well defined whereas in the
fishermen village each fisherman stays in the village his whole life such that
monitoring is effective. The reputation value is also linked to the premium price.
Imagine you have a supplier with a particular plant that has a particular capacity. You
cannot certify the products come from a sweatshop somewhere else and then are
sent to the plant.
To create reputation, you need to force the agents to make costly investments for
future transactions.
(4) Structure-based: You set up the production processes such that the whole
structure depends on the quality of each production step. You internalize outcomes to
overcome incentive problems. Besides, vertical integration creates legal and
economic means to align.
Let’s take Mauritius for example. They live of textile industry, are paid well and kids
aren’t working. Imagine you have to choose between supplier A with 𝑐𝐴 and supplier
B with 𝑐𝐵 with 𝑐𝐴 lower than 𝑐𝐵. You go visit the plant of supplier A to be sure
everything is all right because the price is so low. Everything seems to be OK but
actually the plant is continuing to work at night to make Nike’s shoes and during the
day they make their own shoes, which they sell on the black market. Can you create
group incentives here? No. You have to well design the supply chain. The structure
needs to kill the intermediation possibility between the different processes. You have
to make them make something that has no value on the market. For example, you
can not give them all the holograms so that it can be seen that it is contraband but
sometimes it cannot be seen. In Nike’s case, you make them do only the left shoes
and another one, in Asia for example, only the right shoes. The value for selling the
right shoes is very low and Mauritians are not going to make an arrangement to put
right and left shoes together because that would be too expensive. Give two other
examples of supply chains with huge problems of contraband? Luxury products,
pharmaceutics, tobacco etc.
Why not select 𝑐𝐵? If you select B, agent A can make the same shoes for the same
price and yet not pay the workers what they should get.
Structural solutions solve trivially the problem in theory but incentive provision in
large bureaucracies is costly and management is less effective.
4
Tendering: Soumission d’offres.
11
- Rotation of roles;
- Product design (module interfaces).
Summary
Select one or several categories of solution types and design a solution to address
the specific incongruence.
Conclusions
In the real world, we need to identify the decisions that most critically require
coordination, and give priority to aligning the incentives for these critical decisions in
the design of supply chain structures, design of contracts, the use of trust and the
collection and analysis of information.
The effectiveness of any specific solution depends on its potential power, but
also on the current level of opportunism of the partners involved in a
transaction or relation.
Term Paper
Model extension: you take a model and you change assumptions. You need to
understand the model and work on it analytically and also explain.
Literature Survey: You select a topic, for example, “the role of container standards for
coordination”. You may not fill the whole container with your product if it has specific
dimensions. Thus, our life, products are conditioned by the containers dimensions.
12
Eventually, you will find some branches with no research during the literature survey.
Check examples of a literature survey. You need to survey the literature.
Base Model
We study coordination in a supply chain with one supplier and on retailer. There is
one selling season with stochastic demand and a single opportunity for the retailer to
order inventory from the supplier before the selling season begins.
In this problem, the retailer faces the newsvendor’s problem: he must choose and
order quantity before the start of a single selling season that has stochastic demand.
We always start with two players, the supplier (upstream) and the retailer
(downstream) but those are generic names – the supplier could be a distributor, a
retailer etc.
The risk creates the need for coordination. The risk is introduced from the stochastic
demand 𝑞 > 0 that follows a distribution 𝐹 that I know.
13
The retailer sells a product; he has market access, and sells the product for price 𝑝,
fixed and known. If the retailer decides to stock on those products, then for every
stocked product he will have to pay 𝑐𝑅, it’s the retailer marginal production cost.
If there is anything left when the retailer sells things (leftover 𝐼(𝑞)), then he will
salvage this. The salvage value is 𝑣. What is the salvage value? It is what you got if
you have something left over. You sell it through another channel at another price.
What’s the difference between 𝑝 and 𝑣? 𝑣 is much lower otherwise the risk
disappears. 𝑝 has to be large and 𝑣 is small.
Why does the retailer has to find a way to sell leftovers and why not sell them back to
the supplier? The salvage value 𝑣 is always lower at the supplier, so it should stay at
the retailer to optimize the supply chain.
The supplier gets an order quantity 𝑞 and he has a marginal production cost 𝑐𝑆 higher
than 0. Note that 𝑐𝑆 + 𝑐𝑅 < 𝑝. The supplier is the principal in our game. It’s the
supplier that makes the product and he will propose a contract.
The terms of the contract are represented by 𝑇. We are seeking for the set of
contracts that coordinate the supply chain and arbitrarily allocate its profit. We are
going to design this 𝑇 to achieve coordination. 𝑇 is were the game is played. You
respond to the terms of payment.
Both the supplier and the retailer try to maximize their profit. The supplier has an
independent profit function π𝑆(𝑞) and the retailer has one that is depend on 𝑞
contingence to 𝑇, π𝑅(𝑇(.)).
Notations
Assumptions
There is one retailer and one supplier. Both firms are risk-neutral, so each firm
maximizes expected profit. They have common information about all parameters. The
retailer is the only one having access to the market. There is one selling season with
stochastic demand following a known distribution and a single opportunity for the
retailer to order inventory from the supplier before the selling season begins.
14
There is a deterministic fixed sales price. The salvage at the retailer is more
profitable than at the supplier. The supplier’s and the retailer’s direct cost is lower
than the sales price (𝑐𝑆 + 𝑐𝑅 < 𝑝).
The retailer faces the newsvendor problem: he must choose an order quantity before
the start of a single selling season that has stochastic demand.
→𝐷 is the demand during the selling season.
→𝐹 is the distribution function of demand and 𝑓 its density function.
What’s the goodwill cost? It’s an additional cost of not having the demanded article.
When you don’t sell something because you don’t have it anymore, you lose a sale.
You lose the price but you may also lose the customer, future sales and
opportunities. You lose more than just the margin on that article. You may lose
complementary sales to the one you just lose.
Imagine you sell bread, meat etc. with low added value and beverages and liquors
with higher added value. If you don’t have bread or meat, the customer may go
somewhere else to buy bread and in the same time beverages and you’ll lose more
than just a lost sale of bread.
Example: If you don’t have the electric toothbrush, you won’t sell the toothpaste.
In what situation could we have a goodwill cost for the retailer? Imagine you have a
Danone product and a Carrefour n° 1 product. I have a stockout of Danone products.
The customer shows up to buy the Danone product, which initially attracted him to
the shop. The customer might then buy Carrefour n° 1 product’s. The substitution
makes it even profitable for the retailer. So the goodwill cost for the retailer is not
really a problem. Yet, it is for the supplier. The goodwill cost of the supplier may be a
problem. Indeed, part of the lost sale of Danone, may deviate to other stores where
Danone is sold. The higher the substitutions rate between the products, the higher
the 𝑔𝑆 and the lower the 𝑔𝐶.
15
The retailer earn 𝑣 < 𝑐 per unit unsold at the end of the season. Assuming the
supplier’s net salvage value is no greater than 𝑣, so its optimal for the supply chain to
salvage left over inventory at the retailer.
16
If the true demand is
(1) 𝑑≤𝑞, we’ll sell 𝑑 at full price and we’ll salvage (𝑞 − 𝑑). The income for this
period is 𝑝 * 𝑑 + 𝑣 * (𝑞 − 𝑑).
When 𝑞 is very large, (𝑝 * 𝑑) converges towards what? What is the expected sale? I
will have no constraint so I will simply sell the expected value µ. So (𝑝 * 𝑑)
approaches (𝑝 * μ), that’s the maximal expected revenue I can have.
𝑣 is always smaller than 𝑐 so if we have too much inventory 𝑞, the profit approaches 0
and might even go in the negatives.
(2) 𝑑 > 𝑞, we’ll sell 𝑞 because that’s the quantity we have and there will be no
salvage. The income for this period is 𝑝 * 𝑞 − 𝑔 (𝑑 − 𝑞). There is some
goodwill cost due to lost sales.
Let 𝑇 be the expected transfer payment from the retailer to the supplier.
Both players are (1) risk neutral and (2) profit-maximizers. (1) What if somebody is
not risk neutral? They’ll be more averse to the risk of losses. We, as private person,
are risk averse because we have a limited budget and a finite horizon. Firms can be
risk averse too if those conditions are fulfilled. (2) The firm could have another
function than profit-maximizer such as achieving growth.
17
𝑔𝑆𝐿(𝑞) means that if there is a stockout at the retailer, there is also a cost for the
supplier.
^ ^
We want to find 𝑞 that optimizes the objective function: 𝑞 = 𝑎𝑟𝑔𝑚𝑎𝑥 π𝑅(𝑇). To solve
this problem we need to work as in a Stackelberg game. We have a Stackelberg
model where the leader is the supplier and the follower is the retailer. The leader will
give the contract that will optimize his profit. The Stackelberg game is a game of full
information and anticipation.
^
First step: The supplier anticipates the retailer’s reaction function 𝑞(𝑤).
~ ^
Second step: The supplier optimizes 𝑤 = 𝑎𝑟𝑔𝑚𝑎𝑥 π𝑆(𝑞(𝑤)).
We want to help the firms to optimize their profit. The first thing to do is to look at the
coordination criteria. Take an arbitrary 𝑇(𝑞, 𝑤).
The efficiency of a contract is the ratio of joint profit under autonomous decision
making to maximum joint profit. A coordinating contract is 100% efficient
0 0
→𝑞 is the coordinating production decision if and only if 𝑞 = 𝑎𝑟𝑔𝑚𝑎𝑥 Π(𝑞).
^ 0 ^ ^
→ 𝑇(𝑞, 𝑤) is a coordinating contract if and only if 𝑞 = 𝑎𝑟𝑔𝑚𝑎𝑥 π𝑅 𝑇, 𝑤 ( ) AND
^ ^
(
𝑤 = 𝑎𝑟𝑔𝑚𝑎𝑥 π𝑆 𝑞(𝑤) . )
^ ^ 0 ^ ^
( )
If both do the right thing (𝑤 = 𝑎𝑟𝑔𝑚𝑎𝑥 π𝑆 𝑞(𝑤) and 𝑞 = 𝑎𝑟𝑔𝑚𝑎𝑥 π𝑅 𝑇, 𝑤 ) then ( )
we’re maximizing the whole profit.
0
Let’s take a closer look to 𝑞 .
Π(𝑞) = π𝑆(𝑞) + π𝑅(𝑞)
= 𝑝𝑆(𝑞) + 𝑣𝐼(𝑞) − 𝑔𝑅𝐿(𝑞) − 𝑐𝑅𝑞 − 𝑇(𝑞) + 𝑇(𝑞) − 𝑐𝑆𝑞 − 𝑔𝑆𝐿(𝑞)
18
( )
= 𝑝𝑆(𝑞) + 𝑣(𝑞 − 𝑆(𝑞)) − 𝑔𝑆 + 𝑔𝑅 (μ − 𝑆(𝑞)) − 𝑐𝑅 + 𝑐𝑆 𝑞 ( )
= (𝑝 − 𝑣 + 𝑔)𝑆(𝑞) − (𝑐 − 𝑣)𝑞 − 𝑔μ
N.B.: 𝐿(𝑞) = μ − 𝑆(𝑞) and 𝐼(𝑞) = 𝑞 − 𝑆(𝑞). We can express everything regarding
𝑆(𝑞).
Remarks: (1) Is it normal − 𝑇(𝑞) + 𝑇(𝑞) are cancelling each other off in joint
optimization? Yes, transfer costs do not make sense in joint optimization. It is
( )
independent of the organization. (2) 𝑐𝑅 + 𝑐𝑆 = 𝑐 is the fixed production and
( )
retail cost per article. (3) 𝑔𝑆 + 𝑔𝑅 = 𝑔. (4) 𝑔μ is constant and represents the
negative utility of not selling at all.
' '
Π (𝑞) = (𝑝 − 𝑣 + 𝑔)𝑆 (𝑞) − (𝑐 − 𝑣) = 0
𝑞
𝑆(𝑞) = 𝑞 − ∫ 𝐹(𝑥)𝑑𝑥.
0
' 𝑐−𝑣
𝑆 (𝑞) = 1 − 𝐹(𝑞) = 𝐹(𝑞°) = 𝑝−𝑣+𝑔
𝑐−𝑣 𝑝−𝑣+𝑔−(𝑐−𝑣)
𝐹(𝑞°) = 1 − 𝑝−𝑣+𝑔
= 𝑝−𝑣+𝑔
Afterwards, we need to verify we’re not minimizing the profit. This validation goes
through the Second Order Condition.
'' ''
Π (𝑞) = (𝑝 − 𝑣 + 𝑔)𝑆 (𝑞) =− (𝑝 − 𝑣 + 𝑔)𝑓(𝑞°) < 0
19
Analysis Logic: Testing a contract T
0
What if I say no to the first question? It means 𝑞 will never fall of the game. You’ll
start something that cannot fully optimize the supply chain.
Remark: After having determined the optimal quantity we need to go back and check
on the supplier, as he is the one initiating. Can we get the retailer to get the optimal
quantity? If yes, then it’s a coordinating contract otherwise it’s not. Afterwards, the
allocation has to be coordinating. If that’s the case, we have a good contract.
Quick intuition
𝑔 = 0;
𝑝−𝑐 𝑚𝑎𝑟𝑔𝑖𝑛 𝑓𝑜𝑟 𝑠𝑎𝑙𝑒𝑠 𝑡ℎ𝑎𝑡 𝑔𝑜 𝑡ℎ𝑟𝑜𝑢𝑔ℎ 𝑎𝑡 𝑝𝑟𝑖𝑚𝑎𝑟𝑦 𝑚𝑎𝑟𝑘𝑒𝑡
𝐹(𝑞°) = 𝑝−𝑣
= 𝑙𝑜𝑠𝑡 𝑚𝑎𝑟𝑔𝑖𝑛 𝑤ℎ𝑒𝑛 𝑦𝑜𝑢 ℎ𝑎𝑣𝑒 𝑜𝑟𝑑𝑒𝑟𝑒𝑑 𝑡𝑜𝑜 𝑚𝑢𝑐ℎ
𝑝−𝑐
𝑃(𝐷≤𝑞°) = 𝑝−𝑣
𝑝−𝑐
When 𝑝 is very large, 𝑝−𝑣
tends to 1, so I’ll always make sure I have enough
𝑝−𝑐
inventory. When 𝑐 is very large, 𝑝−𝑣
tends to 0 meaning I don’t want any stock; my
stocking rate will go down and down.
𝑝−𝑐
𝑣 is always smaller or equal to 𝑐. If 𝑣 = 𝑐, 𝑝−𝑣
goes to 1. I can salvage the cost,
there is no risk anymore and I’ll make sure to have enough inventory.
It’s the easiest contract. You just take a fixed cost and you sell every product at that
price. The supplier charges the retailer a fixed wholesale price per unit ordered. You
take one unit you pay 𝑤, you take two units you pay 2𝑤 etc.
20
It is shown that the wholesale price contract generally does not coordinate the supply
chain.
With a wholesale price contract the supplier charges the retailer 𝑤 per unit
purchased: 𝑇𝑤(𝑞, 𝑤) = 𝑤𝑞.
Out of this, we can confirm that 𝑤≤𝑐𝑠 meaning the wholesale price contract
coordinates the channel only if the supplier earns a non-positive profit (he sells at
marginal costs). So the supplier clearly prefers a higher wholesale price. As a result,
the wholesale price contract is generally not considered a coordinating contract. This
points out the problem of double marginalization: coordination failure because there
are two margins and neither firm considers the entire supply chain’s margin when
making a decision. Yet, this type of contract is commonly observed in practice. Why?
For instance, it is simple to administer. A supplier may prefer the wholesale price
contract over a coordinating contract is the additional administrative burden
associated with the coordinating contract exceeds the supplier’s potential profit
increase.
^ ' 𝑤+𝑐𝑅−𝑣
( 0) = 1 − 𝑝−𝑣+𝑔
(1) The retailer optimal order quantity satisfies 𝐹 𝑞(𝑤) = 1 − 𝑆 𝑞 ( ) 𝑅
=
𝑝−𝑤−𝑐𝑅+𝑔𝑅 ^ 𝑝−𝑤−𝑐
𝑝−𝑣+𝑔𝑅
, let’s assume for simplicity that 𝑔 = 0, then 𝐹(𝑞(𝑤)) = 𝑝−𝑣
. 𝑅
𝑝−𝑐
This should be equal to 𝑝−𝑣
as both should give us 𝑞°.
𝑝−𝑤−𝑐𝑅
𝑝−𝑣
=
𝑝−𝑐
𝑝−𝑣 (
= 𝑞°↔𝑐 = 𝑤 + 𝑐𝑅 →𝑤 = 𝑐𝑆 )
The contract only coordinates when 𝑤 = 𝑐𝑆.
For this to coordinate the chain, what should the price be?
If 𝑝 = 0, the retailer will stockout.
If 𝑤 > 𝑐𝑠, what will the retailer do? He will understock.
Is it coordinating? No, this can only coordinate if 𝑤 = 𝑐𝑠. But then how do we allocate
the profit? The retailer gets 100% of the profit and the supplier gets 0. So that
situation is certainly not going to happen in the reality.
^ ' 𝑤+𝑐𝑅−𝑣
(
𝐹 𝑞(𝑤) = 1 − 𝑆 𝑞 ) ( 0) = 1 − 𝑝−𝑣+𝑔𝑅
21
*
This gives a unique wholesale price that induces the retailer to order 𝑞𝑟 units, 𝑤(𝑞),
(
↔𝑤(𝑞) = 𝑝 − 𝑣 + 𝑔𝑅 𝐹(𝑞) − (𝑐𝑅 − 𝑣) )
( )
Retailer’s profit function: π𝑅(𝑞) = 𝑝 − 𝑣 + 𝑔𝑅 𝑆(𝑞) − 𝑐𝑅 − 𝑣 𝑞 − 𝑔𝑅μ − 𝑤𝑞 ( )
0 0
Can this become 𝑞 ? This reaction function has to give the 𝑞 we computed here
^
above. We solve the FOC for the retailer’s reaction function regarding 𝑞(𝑤).
' '(𝑞)
(
π𝑅(𝑞) = 𝑝 − 𝑣 + 𝑔𝑅 𝑆 ) (
− 𝑐𝑟 − 𝑣 + 𝑤 = 0 )
^ 𝑐𝑅−𝑣+𝑤 ^ 0
()
→𝑆' 𝑞 = 𝑝−𝑣+𝑔𝑅
→𝑞=𝑞
𝑐 −𝑣+𝑤 0 '
This ratio ( 𝑝−𝑣+𝑔
𝑅
𝑅
) has to give 𝑞 and 𝑆 𝑞 ( 0) = 𝑐−𝑣
𝑝−𝑣+𝑔
.
Let’s assume,
𝑐𝑅−𝑣+𝑤
λ= 𝑝−𝑣+𝑔𝑅
𝑐−𝑣
λ= 𝑝−𝑣+𝑔
These two ratios are not the same but they have to same proportion. We need to
have the same relationship between all elements.
𝑝 − 𝑣 + 𝑔𝑅 = λ(𝑝 − 𝑣 + 𝑔)
𝑐𝑅 − 𝑣 + 𝑤 = λ(𝑐 − 𝑣)
How to interpret λ? We’ll bring in lambda into the big profit equation.
π𝑅(𝑞 * λ) = λ(𝑝 − 𝑣 + 𝑔)𝑆(𝑞) − λ(𝑐 − 𝑣)𝑞 − 𝑔𝑅µ
The joint profit has to be equal to zero for the first order condition.
Π(𝑞) = (𝑝 − 𝑣 + 𝑔)𝑆(𝑞) − (𝑐 − 𝑣)𝑞 − 𝑔μ = 0
What is lambda? If you know the solution of the big profit function you know the
solution of the retailer’s profit given a certain balancing that is given by a constant.
22
We can find out that lambda is the retailer’s share of the supply chain’s surplus.
This means such a contract can coordinate and lambda represents the profit
allocation of it. But we need to check in detail what it means for the exact solution.
When
𝑤+𝑐𝑅−𝑣 𝑝−𝑣+𝑔𝑅
𝑝−𝑣+𝑔𝑅
=
𝑐−𝑣
𝑝−𝑣+𝑔
→𝑤 = 𝑝−𝑣+𝑔 ( )
(𝑐 − 𝑣) − 𝑐𝑅 − 𝑣 = λ(𝑐 − 𝑣) − (𝑐𝑅 − 𝑣)
Set λ = 1. Who has the power? The retailer, he takes everything because of the
following,
( )
𝑤 = (𝑐 − 𝑣) − 𝑐𝑅 − 𝑣 = 𝑐 − 𝑐𝑅 = 𝑐𝑆
The retailer buys the products at the marginal production cost and obtains all profit.
We have shown that in fact if we put 𝑤 to be λ(𝑐 − 𝑣) − (𝑐𝑅 − 𝑣), then in fact the
wholesale contract can coordinate for a specific 𝑤 = 𝑐𝑆 and λ = 1. But can we
allocate any profit among the players? Here, this contract gives no profit to the
supplier. Let’s try with a positive profit for the supplier.
How about 0 < λ < 1 (the supplier has a positive profit). In this case,
What will happen in this case? Let’s say λ = 0 to see the bound. In that case,
( )
𝑤 =− 𝑐𝑅 − 𝑣 = 𝑣 − 𝑐𝑅. We know that (𝑐 − 𝑣) is positive, as 𝑐 is always bigger
than 𝑣. Thus, (𝑐𝑅 + 𝑐𝑆) > 𝑣. Which means (𝑣 − 𝑐𝑅) is negative, thus a negative
wholesale price, which is impossible.
Remark: The compliance regime doesn’t matter in this contract: for a fixed wholesale
price no less than 𝑐𝑆 the supplier’s profit is non-decreasing in 𝑞, so the supplier surely
produces and delivers whatever quantity the retailer orders.
We could say the retailer’s profit represents compensation for bearing risk: with the
wholesale price contract there is no variation in the supplier’s profit, but the retailer’s
profit varies with the realization of demand.
23
With a buyback contract the supplier charges the retailer 𝑤 per unit purchased, but
pays the retailer 𝑏 per unit remaining at the end of the season.
𝑇(𝑞, 𝑤, 𝑏) = 𝑤𝑞 − 𝑏𝐼(𝑞)
where 𝑏 is the buyback (return) payment.
Note that a retailer should not profit from left over inventory, so assume 𝑏≤𝑤.
This is not to be confused with a physical contract: the remaining units are not
physically returned to the supplier unless the supplier’s net salvage value is greater
than the retailer’s net salvage value. Let’s assume the retailer’s net salvage value is
the highest. When the retailer overstocks newspapers, he bundles them and sells
them at salvage value and gets a deduction, a credit or “markdown money”, for those
units on the invoice for the new newspapers. We assume the salvage value is zero.
It’s a financial agreement.
What sign does this contract has? Is this a positive contract – positive means a
contract from the retailer to the supplier? The quantity 𝑞 is positive. What’s the upper
bound for the number of leftovers? 𝑞. Thus this contract is positive.
Is it coordinating?
( ) (
π𝑅(𝑞, 𝑤, 𝑏) = 𝑝 − 𝑣 + 𝑔𝑅 − 𝑏 𝑆(𝑞) − 𝑤 − 𝑏 + 𝑐𝑅 − 𝑣 𝑞 − 𝑔𝑅µ )
We can compare this to the joint profit we had in the wholesale contract.
Again we can use the lambda-trick. We can use the same approach as above for a
set of buyback parameters {𝑤, 𝑏} and such that λ≥0.
𝑝 − 𝑣 + 𝑔𝑅 − 𝑏 = λ(𝑝 − 𝑣 + 𝑔)
𝑤 − 𝑏 + 𝑐𝑅 − 𝑣 = λ(𝑐 − 𝑣)
* 0
(i) It follows immediately that 𝑞𝑅 = 𝑞 is optimal for the retailer. 𝑇(𝑤, 𝑏) can
coordinate.
(ii) The buyback contract coordinates with voluntary compliance as long as λ≤1.
Moreover, voluntary compliance increases the robustness of the supply chain.
0
Indeed, suppose the retailer is not rational and orders 𝑞 > 𝑞 . Since the supplier is
allowed to deliver less than the retailer’s order quantity, the supplier corrects the
0
retailer’s mistake by delivering only 𝑞 units. This doesn’t work if the retailer orders
0
𝑞 < 𝑞 because he can refuse to accept more than he orders.
Compliance is about selling less or the exact amount that was ordered. If the last
item you deliver is decreasing your profit you could choose not to deliver it.
The good news is that the contract is (i) coordinating, that there is (ii) voluntary
compliance and that it is (iii) flexible in profit allocation.
(iii) The retailer’s profit is increasing in λ and the supplier’s profit is decreasing in λ,
so the λ parameter acts to allocate the supply chain’s profit between the two firms.
Actually, the parameter λ can be loosely interpreted as the retailer’s share of the
supply chain’s profit; it is precisely the retailer’s share when 𝑔𝑅 = 𝑔𝑆 = 0.
We need to solve the contract for (𝑤, 𝑏). We present one contract parameter in terms
of the other
𝑏+𝑔
𝑤(𝑏) = 𝑏 + 𝑐𝑆 − (𝑐 − 𝑣)⎡⎢ 𝑝−𝑣+𝑔
𝑆 ⎤
⎥
⎣ ⎦
(a) If the supplier has all the power, what do we get then, what buyback contract will
be proposed? In that case, the supplier gets all the profit. What’s happening to the
lambda if he has all the power? It goes to 0. That means,
𝑝 − 𝑣 + 𝑔𝑅 − 𝑏 = 0
𝑤 − 𝑏 + 𝑐𝑅 − 𝑣 = 0
25
→ 𝑏 = 𝑝 − 𝑣 + 𝑔𝑅
→𝑤 = 𝑣 − 𝑐𝑅 + 𝑏 = 𝑝 − 𝑐𝑅 + 𝑔𝑅
[𝑤 = 𝑝 − 𝑐𝑅 + 𝑔𝑅 𝑏 = 𝑝 − 𝑣 + 𝑔𝑅
𝑝−𝑣+𝑔
𝑤 = 𝑝 − 𝑣 + 𝑔𝑅 + 𝑐𝑆 − (𝑐 − 𝑣)⎡ 𝑝−𝑣+𝑔 ⎤
⎣ ⎦
(b) What if the retailer has all the bargaining power? What does he propose? Then
lambda equals 1.
𝑝 − 𝑣 + 𝑔𝑅 − 𝑏 = 𝑝 − 𝑣 + 𝑔
𝑤 − 𝑏 + 𝑐𝑅 − 𝑣 = 𝑐 − 𝑣
→𝑏 = 𝑔𝑅 − 𝑔 =− 𝑔𝑆
→𝑤 = 𝑏 + 𝑐 − 𝑐𝑅 = 𝑏 + 𝑐𝑆 = 𝑐𝑆 − 𝑔𝑆
What will happen to the supplier when he has a positive goodwill? For every piece
the retailer buys, he will pay less than the costs. “I know it costs you 1€ to produce
but you need me to sell them so you’re going to sell them to me at 0,5€.” The retailer
goes back to the supplier that subsidized the sales and tells him the supplier needs
to reimburse him for what he didn’t sell. The supplier makes a profit of 0.
(c) What does the profit share of the retailer depend on?
What would guarantee me a profit share in this situation? If I have a low or a high
goodwill cost, what happen in that case?
The buyer reserves 𝑞 units of capacity for a fee 𝑠 and pays another fee 𝑢 for each
unit of capacity used. It’s analogous to a buyback contract with 𝑤 = 𝑠 + 𝑢 and 𝑏 = 𝑢
: the buyer pays 𝑠 + 𝑢 for each unit of capacity that is reserved and receives 𝑢 for
each unit of capacity not utilized.
The buyer’s demand depends on the contract parameters and the uncertain spot
price for additional capacity: if the spot price is less than 𝑢 then the buyer satisfies his
demand via the spot market exclusively, but with higher spot prices the buyer uses
and optimal mixture of the reserved capacity and the spot market.
What happens when the spot price is lower than 𝑢? What can I do as a retailer? The
analogy with the buyback contract falls. That’s the price risk. Do not understand.
26
With a revenue sharing contract the supplier charges 𝑤𝑅 per unit purchased plus the
retailer gives the supplier a percentage of his revenue.
We will share the revenues (not necessarily the costs) that will come for both the
sales and the salvage. Let ϕ be the fraction of supply chain revenue the retailer
keeps, so (1 − ϕ) is the fraction the supplier earns.
In bleu we have the sales revenue and in orange the salvage revenue.
0 0
Is it coordinating? We need to check if 𝑞 = arg 𝑎𝑟𝑔 𝑚𝑎𝑥π𝑅(𝑞) , if 𝑞 is maximizing the
retailers profit.
( )
π𝑅(𝑤, ϕ) = (ϕ(𝑝 − 𝑣) + 𝑔𝑅)𝑆(𝑞) − 𝑤 + 𝑐𝑅 − ϕ𝑣 𝑞 − 𝑔𝑅µ
^ ~~ 0 ^ ^
( )
Is 𝑞 𝑤, ϕ = 𝑞 𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝑤, ϕ?
Special case (illustration): Let’s say 𝑔 = 0 and 𝑣 = 0 and 𝑐𝑅 = 0. Then, it that case,
π𝑅(𝑤, ϕ) = ϕ𝑝𝑆(𝑞) − 𝑤𝑞
In this case what happens? It is good news. The retailer gets a scaled version of the
joint profit. We found immediately a contract that shares profit. This (ϕ∏ (𝑞)) is only
0
optimized when you put 𝑞 to be 𝑞 . Revenue Sharing contracts coordinates for (at
^ ^
least one) setting of (𝑤, ϕ). In our example, we have fixed 𝑤 to be ϕ𝑐 but not ϕ.
27
D. Coordinating contracts
28
REVENUE SHARING 𝑂𝐾 𝑂𝐾
CONTRACT
There is an equivalence
between the buyback contract
and the revenue sharing
contract. Thus, as the buyback
contract is coordinating, the
revenue sharing contract is
too.
29
QUANTITY FLEXIBILITY 𝑂𝐾 𝑂𝐾
CONTRACT
SALES REBATE 𝑂𝐾 𝑂𝐾
CONTRACT
QUANTITY DISCOUNT 𝑂𝐾 𝑂𝐾
CONTRACT
30
E. The Quantity Flexibility Contracts
With a quantity flexibility contract, the supplier charges 𝑤𝑞 per unit purchased but
then compensates the retailer for his losses on unsold units.
Buyback logic was that the supplier protected the retailer partially for everything he
has unsold. It’s a partial reimbursement for all unsold units. If this 𝑏 becomes too big,
then the retailer responds by ordering way too much because he has less risk. The
retailer gets a partial protection on its entire order.
The quantity flexibility contract is about unloading the risk of the retailer by getting
back what he didn’t sold but the supplier will do only a full reimbursement of only a
part of the order δ. The quantity flexibility contract is a contract with two parameters,
𝑤 and δ. The retailer is fully protected on a portion of its order.
The retailer receives a credit from the supplier at the end of the season equal to
( )
𝑤𝑞 + 𝑐𝑅 − 𝑣 𝑚𝑖𝑛(𝐼, δ𝑞), where 𝐼 is the amount of leftover inventory, 𝑞 is the number
of units purchased and δ∈[0, 1] is a contract parameter.
If the supplier did not compensate the retailer for the 𝑐𝑅 cost per unit then the retailer
would receive only partial compensation on a limited number of units, which is called
a backup agreement.
( ) ( )
π𝑅(𝑞|𝑤, δ) = 𝑝 − 𝑣 + 𝑔𝑅 𝑆(𝑞) − 𝑐𝑅 − 𝑣 𝑞 − 𝑇(𝑤, δ, 𝑞) − µ𝑔𝑅
𝑞
( ) ( )
= 𝑝 − 𝑣 + 𝑔𝑅 𝑆(𝑞) − 𝑤 + 𝑐𝑅 − 𝑣 𝑞 + (𝑤 − 𝑣 + 𝑐𝑅) ∫ 𝐹(𝑦)𝑑𝑦 − µ𝑔𝑅
(𝐼−δ)𝑞
The second part is there to improve the wholesale contract, which is actually the first
part of the equation here above.
31
0 0
should give 𝑞 (which we do know), if it is coordinating. The FOC should have 𝑞 as
stationary point. Afterwards, we’ll need to verify its profit maximizing and not
minimizing with the second order condition.
0
𝑞(𝑤, δ) = 𝑞
0
δπ𝑅(𝑞 )
δ𝑞
= 0 is a necessary condition
This holds when 𝑣 − 𝑐𝑅≤𝑤(δ)≤𝑝 + 𝑔𝑅 − 𝑐𝑅. This range is satisfied with δ ∈[0, 1]
because,
Observations:
● 𝑤 is increasing in δ. The supplier will signal to the retailer that if he wants a
higher δ, he’ll set a higher wholesale price, as the supplier will take more risks.
● Any allocation of profit is possible through {𝑤, δ}. All allocations are possible.
You can share the profit, as you want. (See here below)
● δ can be negotiated with the retailer.
● In fact, quantity flexibility contracts require no more information than buyback
contracts. Actually, it even requires less information. If δ is small, it requires
less control.
32
Is this kind of contract implementable? Is it tricky to handle? No, as it is equivalent to
the buyback, it is fine as well.
The supplier wants the retailer to make better orders with the time so the cost of
stockout has to be part of the profit function of the retailer. Good orders, is better for
the supplier as to make better forecasts.
0
For supply chain coordination the supplier must also wish to deliver 𝑞 to the retailer.
The supplier’s profit function is,
𝑞
(
π𝑆(𝑞, 𝑞(δ), δ) = 𝑔𝑠𝑆(𝑞) + 𝑤(δ) − 𝑐𝑆 𝑞 − 𝑤(δ) + 𝑐𝑅 − 𝑣 ) ( )(1−δ∫ )𝑞 𝐹(𝑦)𝑑𝑦 − µ𝑔𝑆
and,
δ π𝑆(𝑞,𝑤(δ),δ)
δ𝑞 ( ) ( )
= 𝑔𝑆(1 − 𝐹(𝑞)) + 𝑞(δ) − 𝑐𝑆 − 𝑤(δ) + 𝑐𝑅 − 𝑣 (𝐹(𝑞) − (1 − δ)𝐹((1 − δ)𝑞)
( )
= 𝑔𝑆(1 − 𝐹(𝑞)) − 𝑐 + 𝑣 + 𝑤(δ) + 𝑐𝑅 − 𝑣 (1 − 𝐹(𝑞)) + (1 − δ)𝐹((1 − δ)𝑞)
0
The supplier’s FOC at 𝑞 is satisfied:
0
(
δ π𝑆 𝑞 ,𝑤(δ),δ )
δ𝑞 (
= 𝑔𝑆 1 − 𝐹 𝑞 ( 0)) − 𝑐 + 𝑣 + (𝑝 − 𝑣 + 𝑔𝑅)(1 − 𝐹(𝑞0)) = 0
0
However the sign of the second-order condition at 𝑞 is ambiguous,
2
δ π𝑆(𝑞,𝑤(δ),δ) 2
δ𝑞
2 (
=− 𝑤(δ) 𝑓(𝑞) − (1 − δ) 𝑓((1 − δ)𝑞) − 𝑔𝑆𝑓(𝑞) )
The voluntary compliance does not hold for these specific cases,
Hence, supply chain under voluntary compliance is not assured with a quantity
flexibility contract even if the wholesale price is 𝑤 (δ). Channel coordination is
achieved with forced compliance since then the supplier’s action is not relevant.
→ When δ = 0, the retailer earns at least the supply chain’s optimal profit,
33
(
π𝑅(𝑞, 𝑤(0), 0) = 𝑝 − 𝑣 + 𝑔𝑅 𝑆(𝑞) −
0
)
0
( 𝑝−𝑣+𝑔𝑅
𝑝−𝑣+𝑔
0 0
)(𝑐 − 𝑣)𝑞 − µ𝑔
0
0
𝑅
Given that the profit functions are continuous in δ, it follows that all possible
0
allocations of ∏(𝑞 ) are possible?
F. The Sales Rebate Contracts
Here (Cashon’s literature), the retailer buys 𝑞 units at 𝑤𝑞. The supplier gives a
discount if the retailer buys 𝑞≥𝑡 units, discount of –𝑟 per unit sold for each unit that he
sells over a threshold 𝑡. This contract depends on three parameters: 𝑤, 𝑟 and 𝑡. The
supplier actually says that if the retailer sells a lot, he will diminish his price.
The logic here is to make the retailer suffer even more from understocking: he loses
profit from not selling and he could have been paying less for those items. The
marginal revenue is no longer constant. You change (𝑝 − 𝑣) in (𝑝 + 𝑐𝑢). You
increase the reward for marginal sales above 𝑡 and doing that you increase the
understocking cost 𝑐𝑢.
𝑞
𝑇(𝑤, 𝑟, 𝑡, 𝑞) = 𝑤𝑞 𝑞 < 𝑡 (𝑤 − 𝑟)𝑞 + 𝑟(𝑡 + ∫ 𝐹(𝑦)𝑑𝑦) 𝑞≥𝑡
𝑡
When 𝑞≥𝑡 the retailer pays 𝑤 − 𝑟 for every unit purchased, an additional 𝑟 per unit
for the first 𝑡 units purchased and an additional 𝑟 per unit for the units not sold above
the 𝑡 threshold. The discount is only given on the units beyond 𝑡.
δπ𝑅 '
δ𝑞 ( )
= 𝑝 − 𝑣 + 𝑔 𝑅 𝑆 (𝑞 ) − 𝑐 𝑅 − 𝑣 −( ) δ𝑇
δ𝑞
(𝑞) = 0
If 𝑡 is set too high, then we get back to a wholesale contract and it will not be
coordinating.
34
δπ𝑅 0
→𝑞 < 𝑡: δ𝑞 (
(𝑞 ) = 𝑝 − 𝑣 + 𝑔𝑅 𝑆 𝑞 ) '( 0) − (𝑐𝑅 − 𝑣 − 𝑤) = 0. This is the wholesale
price contract, which is only coordinating for the special case where the wholesale
price is equal to the marginal cost.
δπ𝑅 0
→𝑞≥𝑡: δ𝑞 (
(𝑞 ) = 𝑝 − 𝑣 + 𝑔𝑅 𝑆 𝑞 ) '( 0) − (𝑐𝑅 − 𝑣) ) 0
− (𝑤 − 𝑟 𝐹(𝑞 ) = 0. When we
compute this, we can compute a wholesale price and it will depend on the 𝑟 we use
here. The equation can be solved for 𝑤(𝑟). If we stick to that wholesale price, what
do we have?
(
𝑤(𝑟) = 𝑝 − 𝑣 + 𝑔𝑅 + 𝑟 𝐹 𝑞) ( 0) − 𝑐𝑅 + 𝑣
This profit function is concave in 𝑞 for 𝑞 > 𝑡, which means the second-order condition
0
is fulfilled and 𝑞 is a local maximum.
0
If 𝑞(𝑤, 𝑟, 𝑡) = 𝑞 then the sales rebate contract is coordinating.
Any profit can be allocated. You know the overall profit. You can allocate, through a
constraint, all profit through the two players.
Let’s consider the supplier’s production decision. The supplier’s profit function given
a coordinating sales rebate contract is,
π𝑆(𝑞, 𝑤(𝑟), 𝑟, 𝑡) =− 𝑔𝑆(µ − 𝑆(𝑞)) − 𝑐𝑆𝑞 + 𝑇𝑆(𝑞, 𝑤, (𝑟), 𝑟, 𝑡)
For 𝑞 > 𝑡,
δπ𝑆(𝑞,𝑤(𝑟),𝑟,𝑡)
δ𝑞
= 𝑔𝑆𝐹(𝑞) − 𝑐𝑆 + 𝑤(𝑟) −
𝑟
𝑟𝐹(𝑞) ( )(
= 𝑟 − 𝑔𝑆 𝐹(𝑞) − 𝐹 𝑞 ( 0))
0
The above is positive for 𝑞 < 𝑞 only if 𝑟 < 𝑔𝑆. But if 𝑟 < 𝑔𝑆, then 𝑤(𝑟)≤𝑐𝑆; the
supplier cannot earn a positive profit with 𝑟 < 𝑔𝑆. As a result it must be that 𝑟 > 𝑔𝑆,
which implies the supplier loses money on each unit delivered to the retailer above 𝑡.
So the sales rebate contract does not coordinate the supply chain with voluntary
compliance.
35
Those are stronger assumptions to the newsvendor model.
Initial assumption: the price is exogenous, single and fixed. The retailer impacts sales
only through his stocking decisions. This may not be realistic.
Here, we assume that the retailer sets a price 𝑝 such that 𝐹(𝑥|𝑝) is reacting as
δ𝐹(𝑥|𝑝)
follows: δ𝑝
> 0, which means the demand decreases stochastically in price.
Recall that F is the probability that demand is less than 𝑥.
We want to know if a contract that coordinates the retailer’s order quantity also
coordinates the retailer’s pricing. Buyback, quantity flexibility and sales-rebate
contracts do not coordinate in this setting. Those contracts run into trouble because
the incentive they provide to coordinate the retailer’s quantity action distorts the
retailer’s pricing decision. Revenue sharing coordinates if there are no goodwill
penalties. With goodwill penalties there exists a single coordinating revenue sharing
contract that provides only a single allocation of the supply chain’s profit. The quantity
discount does better: it coordinates and allocates profit avec is 𝑔𝑅≥0, but 𝑔𝑆 = 0 is
required. The price discount contract coordinates and arbitrarily allocates profit.
(Insert (1)).
The retailer sets both the retail price and the quantity. He has two liberties. We
assume that the retailer sets his price at the same time as his stocking decisions and
the price is fixed throughout the season.
What happens to the coordination? We always begin with the first order condition of
the integrated channel because that’s what leads to the best profit.
(a) Integrated benchmark: Assume that the supplier and the retailer are in an
integrated entity maximizing their profit.
𝑞
𝑆(𝑞, 𝑝) = 𝑞 − ∫ 𝐹(𝑝)𝑑𝑦
0
𝑆(𝑞) is expected sales given the stocking quantity 𝑞 and the price 𝑝. Depending on
what 𝑝 is, you’ll get a different demand.
0 0
Now we want to compute the FOC for a stationary point {𝑞 , 𝑝 }, which is a finite (but
not necessarily) optimal quantity-price pair:
0 0
0 0
δ∏(𝑞 ,𝑝 )
δ𝑝 (
= 𝑆 𝑞 ,𝑝 ) − (𝑝0(𝑞) − 𝑣 + 𝑔) δ𝑝δ𝑆 = 0 (2)
0 0 0 0
( ) 𝑝0 − 𝑣 + 𝑔 − (𝑐 − 𝑣) = 0
δ∏(𝑞 ,𝑝 )
δ𝑞
=
δ𝑆 𝑞 ,𝑝
δ𝑝 ( )
36
0
A contract fails to coordinate if it is unable to satisfy the FOC at 𝑝 (𝑞), or it is able to
0
satisfy the FOC at 𝑝 (𝑞) only with parameters that fail to coordinate the quantity
decision.
The 𝑞 is reacting to the price just like it would do if the price was exogenous. Does it
change anything when we use any of the contracts that we had before?
(b) Let’s compare this with the buyback contract. For a fixed 𝑝, buyback is a
coordinating contract. Now, we’ll test the buyback with an endogenous 𝑝.
But 𝑏≥0, so we feel there is some kind of problem. We could imagine a buyback
contract with a negative buyback price. Yet even if 𝑏 < 0 we got 𝑤 − 𝑔𝑆 = 𝑐𝑆 − 𝑔𝑆, ( )
which is not acceptable to the supplier. Therefore, a buyback contract does not
coordinate the newsvendor with price-dependent demand.
The retailer sets the price in a buyback contract. He will only set the optimal price that
the chain would do if he gets a contract that is very extreme. Indeed, when he sets
𝑏 =− 𝑔𝑆, it means that for example if the retailer got newspaper leftovers, he will
have to reimburse the goodwill price to the supplier, give the supplier a little bonus.
When 𝑏 becomes negative, what interest does the retailer has to show this
information? Why would you tell him you have leftovers if you’re profit can only get
worse if you tell him? This contract cannot work. It is a very strange contract because
the expected profit for the supplier is zero. That’s why the supplier has to be paid for
leftovers because otherwise his profit could even be negative. This contract is not
coordinating.
You got similar disappointment for the other contracts (See Cashon 6.3).
Some intuition: Why is it like this? The buyback contract fails to coordinate in this
setting because the parameters of the coordinating contracts depend on the price.
The coordinating parameters are:
𝑏 = (1 − λ)(𝑝 − 𝑣 + 𝑔) − 𝑔𝑆
(
𝑤 = λ𝑐𝑆 + (1 − λ) 𝑝 + 𝑔 − 𝑐𝑅 − 𝑔𝑆 )
For a fixed λ, the coordinating buyback rate and wholesale price are linear in 𝑝.
Hence, the buyback contract coordinates the newsvendor with price dependent
demand if 𝑏 and 𝑤 are made contingent on the retail price chosen, or if 𝑏 and 𝑤 are
37
choses after the retailer commits to a price (but before the retailer chooses 𝑞). The
contract that the supplier will present to the retailer does not only depend on 𝐹 but
also on 𝑝, thus on 𝐹(. |𝑝). How to make this work? When you look at the FOC for the
integrated case, you see it’s working. It means you need the retailer to commit to a 𝑝
prior to the 𝑞 to make it work (Bernstein a Federgruen, 2000). This type of contract is
called a price-discount sharing contract. If you don’t proceed like that, the retailer will
only set the optimal 𝑝 if he gets the whole profit otherwise he will screw the price
down. Notice that the retailer gets a lower wholesale price if the retailer reduces his
price, i.e. the supplier shares in the cost of a price discount with the retailer.
When is the initial assumption of 𝑝 being exogenous true? It is true for no branded
commodities like tomatoes. This is not the typical assumption for commodities. The
price is very often a decision variable.
How do you implement this in reality? There are two ways of implementing this type
contract:
(1) What does Apple? They never make sales. Additional distributors have to
respect the price Apple has set. They control the price by requiring
commitment from the retailer. This is possible because they also have an
integrated channel so they can afford to say that if the distributor doesn’t
commit they will not sell through them.
(2) If you take some other branded goods, how do they do? As a supplier you’re
interested in having the optimal profit and share it. The supplier proposes a
recommended retail price that fits other things he does. It’s only if the retailer
sells at the recommended price that the supplier will help him with promotional
things and so on. The reason why the supplier can do that is because they
can anticipate the full profit they can get so that the retailer can set the price
that maximizes its profit.
So far, we have assumed the firms possess the same information when making their
decisions (full information). In that case, the optimal supply chain performance
requires more than the coordination of actions. It also requires the sharing of
information so that each firm in the supply chain is able to determine the precise set
38
of actions. Sometimes sharing information is difficult. Unfortunately, there is also the
possibility of opportunistic behavior with information sharing. For example, a
manufacturer may tell her supplier that demand will be quite high to get the supplier
to build a substantial amount of capacity.
Which contracts, if any, achieve supply chain coordination and how are rents
distributed with those contracts. In this model coordination requires (1) the supplier
takes the correct action and (2) an accurate demand forecast is shared.
There could be some kind of common belief on the cost because we don’t know
them.
The retailer knows more about the real demand than its supplier. How can the
supplier make sure that he gets the good forecast?
The first best solution is when you share information. The literature tries to find
contracts where there is a forecasting sharing. Weng found out that quantity
discounts do not achieve coordination.
Assume that one of the players, e.g. the retailer, has private information on a
parameter θ relevant for the profit/cost optimization. If θ is necessary for the supplier
^
to optimize ∏(.), then the procedure must reveal θ. The retailer will disclose θ which
^
may be a lie or the truth if θ = θ. A mechanism (contract) can be information
^
revealing if it leads to θ = θ (inducing truth telling).
39
Consider
π𝑆(𝑡)
𝑠. 𝑡. π𝑅(𝑞(𝑡, θ))≥0
This constraint is the participation constraint or individual rationality IR.
~ ~
( ( ))
π𝑅(𝑞(𝑡, θ)) > π𝑅 𝑞 𝑡, θ ∀θ≠θ
This constraint must be true for the contract to be revealing. This constraint is called
the incentive compatibility constraint ICC.
Let’s now go to a little example – Capacity Procurement (Cachon & Larivière, 2001)
(Insert 3). A manufacturer 𝑀 develops a new product with uncertain demand. There
is a single potential supplier 𝑆 for a critical component. The demand type information
is non-observable, non-verifiable and exposed verifiability doesn’t work either. The
supplier needs to invest in capacity 𝑘≥0 prior to production. We assume there is no
salvage value and no goodwill cost.
The integrated firm observes θ, there is no signal needed. The firm will never have a
𝑘 that is strictly larger than 𝑞 if it’s an integrated firm. I know exactly what I will order
and thus what capacity to invest in. The supply chain makes two decision: how much
capacity to construct and how much to produce. The latter is simple, produce
𝑚𝑖𝑛{𝑥, 𝑘, 𝑞}.
∏(𝑘, 𝑞)
0 0
An as said above 𝑘 > 𝑞 cannot be optimal, 𝑘 ⊂ 𝑞 .
𝑐𝑢
→ 𝐹θ 𝑘 ( 0) = 𝑟−𝑐𝑝
40
0
This means the supply chain coordination is achieved if the supplier builds 𝑘θ units of
capacity and defers all production until after receiving the manufacturer’s final order.
∏= 𝑤𝑂𝑞𝑖 + 𝑤𝑒𝑆θ(𝑞𝑖)
𝑎𝑛𝑑 ( )
π𝑚(𝑞) = 𝑟 − 𝑤𝑒 𝑆(θ) − 𝑤0𝑞
This is quite the same as a wholesale price contract (𝑤𝑒 + 𝑤0 is the wholesale price)
with a termination penalty charged for each unit the manufacturer cancels in stage 2
(where 𝑤0 is the termination penalty).
( )
𝑘 is equal to 𝑞 if the objective function is optimal. In that case, 𝑟 − 𝑤𝑒 = λ(𝑟 − 𝑐𝑝)
and 𝑤0 = λ𝑐𝑢. These conditions allow us to put the optimal range for λ ∈[0, 1].
π𝑚(θ) = λ∏(𝑘|θ)
The profit sharing parameter λ also says that the manufacturer has to pay part of the
capacity costs through the wholesale price.
δπ𝑆
δ𝑘 ( )
= (1 − λ) 𝑟 − 𝑐𝑝 (1 − 𝐹(𝑘|θ)) − 𝑐𝑢 = 0
=− λ𝑐𝑢 < 0
0
Is 𝑘 equal to 𝑞 ? Here, we have an underinvestment in capacity. The supplier builds
to low capacity and the contract is not coordinating. If the manufacturer wanted to
invest more, what can he do? He can signal that he has a strong demand all the
time, the supplier we’ll rerun its optimization and have a higher capacity. The supplier
knows about that and says that if the manufacturer says the demand is low, he
believes him because he will never have an incentive to lie about a low demand
because if you have a high-realized demand you will lose money. If you say the
demand is strong you might be lying and the supplier will ignore the signal. If the
signal is ignored, only 𝑞 can be used.
41
We have a new constraint in our models that make sure we have given the right
information and that the game does not lead to a suboptimal.
Our model is the Stackelberg game and the quantity is called the reaction function.
The supplier is the leader and the follower is the retailer. The leader draws first and
the follower follows. The leader knows what the follower will do and he maximizes his
profit with respect to the wholesale price. The follower will maximize his profit with
respect to the quantity.
The best response function is a strategy that maximizes the payoff of the player.
Given the 𝑛-player game, player 𝑖’s best response (function) to the strategies 𝑥−𝑖 of
*
the other players is the strategy 𝑥𝑖 that maximizes player’s 𝑖’s payoff π𝑖(𝑥𝑖, 𝑥−𝑖):
*
𝑥𝑖 (𝑥−𝑖 = arg 𝑎𝑟𝑔
A Nash equilibrium is when the best response is the best response to all the other
* * ~ ~
strategies. How do we know if we have a Nash equilibrium? If π𝑖(𝑥𝑖 , 𝑥−𝑖)≥π𝑖(𝑥𝑖, 𝑥−𝑖)
~ * * *
for ∀ 𝑥𝑖 ≠𝑥𝑖 , ∀𝑖 then (𝑥𝑖 , 𝑥−𝑖) is a Nash equilibrium. If I have no incentive to change,
then it is a Nash equilibrium. It’s about unilateral change; it assumes you only look at
*
your move. You cannot change 𝑥−𝑖. An outcome is a Nash equilibrium of the game
* * * * * * *
{𝑥1, 𝑥𝑖 , …, 𝑥𝑖−1, 𝑥𝑖+1, …, 𝑥𝑛} when 𝑥𝑖 is a best response to 𝑥−𝑖 for all 𝑖 = 1, 2…𝑛.
* * * * * *
Note that 𝑥−𝑖 = {𝑥1, 𝑥𝑖 , …, 𝑥𝑖−1, 𝑥𝑖+1, …, 𝑥𝑛}.
42
Let’s take an example. We have two players and in the first picture player 1 starts by
choosing left or right. Once he’s done that player 2 has the chance to choose left or
right. What are the strategies for player 1? {Left, Right}. What are the strategies for
player 2? {left|Left, left|Right, right|Left, right|Rights}. A strategy here is something
that has been calculated before. There is no need for playing the game, we have it
already. The left game can be seen as a sequential game.
In the right game, player 2 will have to make a decision without knowing what player
1 chooses. It’s a simultaneous game. The approach for this is to use tables where
you put the payoffs and possibilities. Using the definition we need to find the
equilibrium. If {𝑅, 𝑙} is an equilibrium no one should have an incentive to move. This is
not true. Is {𝑅, 𝑟} an equilibrium in this game? Yes, it is. If {𝑅, 𝑟} is played, they will
stick to it. {𝐿, 𝑙} is also an equilibrium. We have a multitude of equilibriums but they
are not equally good. {𝑅, 𝑟} is Pareto-optimal, the other is not.
I play once a sequential game. I wanted to play right because that’s the best. If that’s
the equilibrium I’ll take it but I tremble and I take the less good equilibrium. In the next
game, simultaneous game, what will player 2 choose? He might go to the left.
Assume that if he unilaterally moves to the right, he thinks he will be worse off as he
thinks player 1 will choose left and then he’ll get − 1. He thinks that the player 1 will
play left because he gets higher payoffs (It’s not true but it’s a belief linked to the first
decision). If you get a poor equilibrium in the beginning it may be very harmful
because of the stickiness of such decision.
43
Let’s take a look to the prisoner’s dilemma.
COOPERATE DEFECT
COOPERATE (− 1, − 1) (− 10, 0)
DEFECT (0, − 10) (− 8, − 8)
What is the link between this game and supply chain? Imagine two franchisees. Each
of them has the choice of investing high effort (good service to client etc.). If one
franchisee free rides on the effort of the other one, he’ll be better off. If none of us
invest, they end up by killing the brand because unilaterally I know the other one is
not going to invest. This is an explanation to why we have moral hazard. What can
we do? When you run a dynamic game you can use strategies to penalize the
players. If you repeat this game, and you use strategies to penalize and you
communicate on it before the game begins. This makes both players to coordinate
and if one of them cheats the next game they will make less profit. “I’ll start
cooperating and if you defect I’ll defect for all next games.” The idea is to tie together
future payoffs and to commit to your choice.
Subgame Strategies: Imagine two players. The first one is incumbent (he’s on the
market) and the entrant is the second one. This is a stage game where player 2 can
either enter the market either stay out. If he stays out the first player has either a
monopoly (10,0) either no profit (0,0). If the entrant comes in, the incumbent can
either decide to fight (-8, -8) either decide to truce (3, 3) – rather than having the
monopoly profit they split the duopoly profit in two.
ENTER STAYOUT
FIGHT (− 8, − (10, 0)
TRUCE (3, 3) (0, 0)
In this game, player 1 knows what player 2 has done. Before the game, player 1
warns player 2 that he is going to fight if he enters the market. Let’s say that player 2
doesn’t believe that and he enters. What will player 1 do? Truce because once player
2 has entered he cannot revoke and player 1 will be better of by trucing. Once he has
entered, there is a subgame. Player 2 ignores player’s 1 warning as he cannot
44
commit to it (cheap talk). How do you implement this in practice? This is happening
for industries with low differentiation. Example: Belgian company buys a plant that
has to run all the time; they cannot lower their volumes nor export. If someone wants
to come in they transform the cheap talk in pre-commitment. If they come in, the only
outcome they can do is (− 8, − 8) because (3, 3) doesn’t exist anymore. The first
one who enters the market is going to build a lot of capacity so he can pre-commit.
Cooperative games assume that players can make coalitions, binding commitments.
In non-cooperative games, there is no way you can split your profit among different
player. There are no coalitions possible. We assume complete information: all players
observe all draws.
Why would you collaborate? To obtain some skills, resources more quickly, to learn
from you partner, to share costs and risks, etc.
In cooperative games you focus only on the outcome, which you split in the end. You
take the players that decide on a common strategy. The value of the coalition is the
one that is split between the players. In non-cooperative comes you focus on the
actions and not on the outcome.
Assumptions:
● Players 𝑁 can form any coalitions 𝑆 that are beneficial to them. There is no
prior assignment of power.
● The value of the coalition 𝑆, 𝑣(𝑆) does not depend on non-coalition member’s
actions.
A core is a solution where the players are better off than before, if it exists. The idea
is that if there is a core and if it’s unique than usually it’s a predictive one. The
collusions, the coalitions etc. are the ones that are going to happen. The core may
not exist and may not be unique.
Problems with independence in SCM: It should be independent, i.e. the value of the
coalition may not depend on the players. The value of the coalition is the surplus that
is made. In imperfect competition, the players that are not in the coalition have an
impact on the revenues. Yet, they do not have any impact on the variables and fixed
costs. 𝑣(𝑆) is not independent of residual coalitions, standard formulations must
assume competitive markets.
𝑣(𝑆) = 𝑆𝐶 𝑠𝑢𝑟𝑝𝑙𝑢𝑠 = ∏(𝑆) = 𝑟𝑒𝑣𝑒𝑛𝑢𝑒𝑠 − 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒 𝑐𝑜𝑠𝑡𝑠 − 𝑓𝑖𝑥𝑒𝑑 𝑐𝑜𝑠𝑡𝑠
= 𝑝(𝑆, 𝑁 − 𝑆)𝑞(𝑆) − 𝑐(𝑆)𝑞(𝑆) − 𝐾(𝑆)
45
One particular approach to calculate the value of the coalition is the Shapley value.
The Shapley value is unique. The logic behind it is that you take the marginal gains
from random coalition configuration: You get a marginal contribution of each player,
you take the expected value of each player and you weight them together. Then you
get the Shapley value, which is the value of you joining a random coalition. It is
relatively low predictive.
Example – Cobranding Intel Inside: You have a game with a supplier (Intel) and two
retailers (IBM and ISUS). ISUS is an unknown Chinese PC brand. Imagine that a
supplier can supply only one retailer with a product at normalized marginal cost (0).
● Time 0: The IBM can invest in co-branding.
● Time 1: The IBM negotiates with IBM and ISUS.
Intel needs to sell its product to one of the retailers. If they sell it as a processor IBM
can sell it’s PCs at 9 and ISUS for 3 (Values retail with no co-branding
{ }
𝑅1, 𝑅2 = {9, 3}. If they play this game without co-branding, they can get anything
from 3 to 9. 9 if they get all the profit from IBM and 3 if they get all the profit from
ISUS. This means that Intel would sell to IBM and IBM would buy at 3. Now suppose
you place the cobranding (Intel Inside). If they do cobranding IBM would always sell
it’s PCs to 9 because they are already good and ISUS to 7 because of the Intel Inside
{ }
(Value retail with co-branding 𝑅1, 𝑅2 = {9, 7}. With co-branding, Intel can get
anything between 7 and 9. The equilibrium would be that IBM buys it at 7.
Cobranding No Cobranding
Supplier’s Profit {3, 9} *
{(7 − 1), 9 }
Contract {𝑅1, 𝑅2} = {1, 0} { }
𝑅1, 𝑅2 = {1, 0}
Value {𝑅1, 𝑅2} = {9, 0} {𝑅1, 𝑅2} = {9, 0}
* There is also a co-branding cost. If Intel co-brands, they incur a cost of -1 and get
{7 − 1, 9} and get {3, 9} if they do not co-brand. Still cobranding is more profitable so
that’s what they are going to choose.
It shows that it is important to have recognition of your brand to the final competition.
By bringing in co-branding you increase competition: you give some reputation to
ISUS, which means you increase the power of a competitor.
Co-branding serves no value-added, only to improve the supplier’s profit
margin.
To pull out to break this strategy all high players have to do it (HP, Dell and IBM). This
game is a dynamic game with a cooperative part. Intel first wins the game and then
the high level players have to create a coalition to break it, otherwise it becomes a
prisoner’s dilemma.
46
Cooperative game theory is an interesting tool for the collaborative or strategic phase
prior to supply chain coordination, e.g. formation of alliances or the scope of joint
operations.
Shapley values or investigations of the existence of a non-empty core are only valid
for certain applications.
Biform games provide currently the base for cooperative models.
In static games all players make simultaneous moves. In dynamic games you play
several games.
It’s very rare in supply chain that you plan one static game once and for all. Usually,
you have dynamic games.
1 1 2 2
If you have a first static game with {𝑥 , 𝐴 }, a second one with {𝑥 , 𝐴 }, etc. Then
you can make your strategy contingent to the other games. Any strategy can be
contingent on the outcome of previous game. In static games, you have no
reputation, commitment etc. that is build during time.
In repeated games, the set of Nash equilibriums is much larger than in static games.
Indeed, it may include Nash equilibriums that are not found in static games.
Trigger strategies may induce coordination if the threat is credible. They make an
outcome (usually a poor result) subject of (automatic) retaliation. If you play the
prisoners dilemma for a long period, it’s more difficult to cheat. I know that from a
static point of view you will cheat. If you cheat, I will make something bad. That’s
possible as the game is repeatedly played.
We need to base this on something. Von Neumann invented the game theory
concept.
The folk theorem says that if you have a repeated game and you have a discounting
1 1 2 2
factor: {𝑢1, 𝑢2} for the first game and b{𝑢1, 𝑢2} for the second game where 𝑏 is the
discounting factor. If the discounting factor is 0, there is no dynamic game because I
don’t care about the future. If the discount factor is equal to 1 then you have a
problem in the formula because the sum does not converge. It means you give equal
consideration to today and to what will be happening in 100 years.
47
1 2 1 2
( )
𝑁𝑃𝑉 𝑋 , 𝑋 ... = 𝐸𝑢 + 𝑏𝐸𝑢 + …
If you run the game for an infinite amount of time and the players care about the
future then there is a possibility to make that equilibrium. I can make a strategy so
that the players will play the best strategy. Indeed, the folk theorem says that,
given an infinite sequence of payoffs 𝑟1, 𝑟2 ... for player 𝑖 and a discount factor
0 < 𝑏 < 1 a payoff profile 𝑟 is enforceable if 𝑟≥𝑢(𝑖){𝑠(𝑖), 𝑠(− 𝑖)} , if
● 𝑟 is a Nash equilibrium payoff vector for any game 𝐺, then 𝑟 is enforceable.
● 𝑟 if feasible and enforceable, then 𝑟 is a Nash equilibrium for the infinitely
repeated 𝐺 with average rewards.
Example:
Defect Cooperation
Defect (0,0) (20,-1)
Cooperation (-1,20) (10,10)
In period 1, player 1 declares the grim strategy. Let’s check if this makes sense for
player 2.
2 10
π(𝐶𝑂𝑂𝑃) = 10 + 10δ + 10δ … = 1−δ
π(𝑁𝑂𝑁 − 𝐶𝑂𝑂𝑃) = 20 + 0 + 0 = 20
20 is the temptation outcome.
2 δ
π(𝑁𝑂𝑁 − 𝐶𝑂𝑂𝑃, 𝐶𝑂𝑂𝑃, 𝐶𝑂𝑂𝑃, 𝐶𝑂𝑂𝑃) = 20 − 1δ − 1δ … = 20 − 1−δ
You need to compare the players discounted payoffs of COOP vs. NON-COOP. Is
10 10
COOP larger than NON COOP? 1−δ > 20. If 1−δ is smaller than 20, then the
grim trigger will not give a long-term cooperation.
You’ll never have (𝐶𝑂𝑂𝑃, 𝐶𝑂𝑂𝑃, 𝐶𝑂𝑂𝑃, 𝐶𝑂𝑂𝑃, 𝑁𝑂𝑁 − 𝐶𝑂𝑂𝑃). If I want to cheat, I
will do it now because of the discount factor.
48
Long-term player: δ is high. What you do tomorrow really has an impact on today. It’s
a culture of reputation.
If δ = 0, 9 then you’ll get 100 which is much bigger than 20. If δ = 0, 5, then you will
get 20 and that is not bigger than 20. For that discount factor, I will not be able
to play the grim trigger strategy.
The idea is to say that you do whatever you want because you have an independent
role but if you cheat on me you will have to consider that payoff of me never
forgiving you. Cheating carries the cost of difference in profit for the rest of the
game.
10 2 10 10
We need to compare 1−δ
to 20 − 1δ + δ ( 1−δ ). Assume δ = 0, 9 then 1−δ
is equal
2 10
to 100 and 20 − 1δ + δ ( 1−δ ) is equal to 100,1. What’s the conclusion here?
Player 2 will choose first not to coordinate and then coordinate. That’s a
problem? How do we modify the tit for that? If the π(𝐶𝑂𝑂𝑃) strategy does not
coordinate than you need to extend the relations for T periods. What is the
minimum T?
π(𝐶𝑂𝑂𝑃) = 100
2 𝑇 𝑇+1
π(𝑁𝑂𝑁 − 𝐶𝑂𝑂𝑃, 𝐶𝑂𝑂𝑃, 𝐶𝑂𝑂𝑃, 𝐶𝑂𝑂𝑃) = 20 − δ − δ − … − δ + δ 100
( )+ δ
𝑇
1−δ 𝑇+1
100 > 20 + δ 1−δ
100
δ = 0, 9
𝑇 𝑇+1
(
First Period: 100 > 20 − 9 1 − 0, 9 + 0, 9 100 )
Second period: 100 > 20 − 9 (1 − 0, 81) + 72↔100 > 92 − 9 * 0, 19.
When the discount factor is not too extreme, a finite game with a sufficient amount of
games is the same as an infinite game. The infinite game is also valid for a
49
finite game with a random number of games. Mathematically, the random thing
decreases the discount factor.
R 1 C NC NC NC
R 2 NC C C C
At the end of period 4, player 1 asks itself “should I forgive or should I hold grudge”?
He has to decide between,
π(𝐻𝑂𝐿𝐷 𝐺𝑅𝑈𝐷𝐺𝐸) = 0
π(𝐹𝑂𝑅𝐺𝐼𝑉𝐸) = 100
It says that even after a certain times of player 2 coordinating you’ll still punish
yourself and him by getting a 0 profit. If you forgive you could together earn a
100 profit. The grim strategy is not observable in reality because infinite
commitment is not possible. In reality, tit for tat is applied.
50
Debrief: Hamptonshire Express
Question 1
𝐶𝑈
What is the fill rate? It’s a service rate and it’s represented by ( 𝐶 +𝐶𝑂
). The fill rate in
𝑈
the exercise is 98%. How do we interpret that? Do we always cover 98% percent of
demand? No, but it’s the probability that you cover the demand. With 584
newspapers we expect to cover 98% of the demand. There is only 2% risk that you
don’t get a newspaper.
Question 2
Does Sheen shift something else than the mean demand when she changes her
effort level? The utility. Sheen is not only a profit-maximizer but also a
utility-maximize 𝑢(ℎ, π). In that case, you would have to include the profit but also the
effort term.
How many hours should she work? 4 hours. When she does that, how much would
* *
( *)
you stock? 𝑞 = 685. What is the fill rate? ϕ(𝑞 ℎ ) = 0,98. Do we serve more
customers? No, she shifted the mean and now she’s optimizing regarding that new
mean. That’s why the fill rate doesn’t change.
*
( *)
This would give Sheen a channel profit of π(𝑞 ℎ ) that amounts to 371,33. If we
compare this to the previous channel profit (331), she increased the profit by around
( *( *)) − π(𝑞°) = 40, 8). The cost of extra effort is 10.
40,8 (∆π 𝑞 ℎ
Effort is usually not measured as time. The effort is coming in and shifts the demand,
and so the mean of the demand too. It sort of increases or decreases the demand.
The effort is usually used when in a store you have different products, substitutes.
Depending on what advice the salesperson gives to the customer the demand is
going to be shifted. This is not observable such as time is in the exercise. You have
51
all kind of small efforts but they all refer to something that is costly. It’s an extra effort
you can make that will shift the demand. How to get the retailer to exert the optimal
effort?
Question 3
We switch from an integrated channel to a true supply chain with Anna, the supplier
and Ralph, the retailer. The question is here, in this new situation, for a given effort,
what would be the optimal stocking quantity? If the retailer can choose and Sheen
^
has a given effort, how much does he choose to stock? 𝑞(𝑤) = 516. Doing that he
^ ^
( ) ( )
gets a profit of π𝑅 𝑞(𝑤) = 62, 14 and Anna gets a profit of π𝑆 𝑞(𝑤) = 259, 6. In this
case, what happens with the channel profit? It shrinks because together they make
only 321,7.
What’s the fill rate now? 86%. Ralph doesn’t pay 𝑐 but 𝑤 to Anna to get his stocking
^
quantity 𝑞(𝑤). Ralph wants fewer articles because his overstocking cost is higher.
There is a lower fill rate since Ralph faces a higher overstocking cost, based on 𝑤.
Ralph’s profit function doesn’t depend on 𝑐 but on 𝑤. Now we have a lower profit and
a lower fill rate (proportion of satisfied customer is decreased).
What effort should Anna make? The optimal effort in the supply chain with wholesale
price contracts 𝑤 is ℎ (𝑤) = 2, 25 ℎ𝑜𝑢𝑟𝑠. What happened to her effort compared to
the original situation? It diminishes because her margin dropped.
The first effect of a differentiated channel (question (a) and (b)) is that retailer cuts
the service level and besides the optimal effort level for that contract is lower. If she
^
exerts 2,25 hours, Ralph will order 𝑞(𝑤, ℎ(𝑤)) = 491. He’s contracting his demand
because Anna shifted the demand again. What fill rate does that give? The same as
before, thus 86%.
^ ^
( )
π𝑅 𝑞(𝑤), ℎ(𝑤) = 57 𝑎𝑛𝑑 π𝑆(𝑞(𝑤, ℎ(𝑤)) = 262
That makes a channel profit of 319. What happened to Anna here? Before she had
(𝑝 − 𝑐) as margin and now she has (𝑤 − 𝑐) as a marginal benefit. Anna decreases
her effort level since her marginal profit decreases from 𝑝 − 𝑐 to 𝑤 − 𝑐. Her profit
levels out after 2,25 hours.
In question (d) we’re asked to find to optimal transfer price 𝑤 (whole sale price). What
can we get here? What do we get if we change this transfer price? What’s a good
transfer price? 𝑤 = 0, 46. What happened at that level?
Anna’s profit is computed with respect to the transfer price: (i) the higher the transfer
price, the higher direct transfer profit from the contract, (ii) besides she can exert
more effort, increase the demand and have a higher profit. The higher the transfer
price, the more money she makes.
What is the logic that applies to Ralph’s profit function? The higher the transfer price,
keeping the selling price constant, the higher the risk you make him carry. When you
increase this transfer price to the full price, then you have 0 quantity and his marginal
profit will be zero. Clearly here in this case, what matters is not to look at the
52
individuals because then we’ll have the two extremes, but we’ll have to look at the
sum of their profits, their joint profits.
What can we say about stocking and effort levels compared to an integrated
channel?
Both are maximized as well in the differentiated channel as in the integrated channel
but not in the same way. The prelude to the course is this question. Chains today are
almost exclusively differentiated. Integrated channels do nearly not exist anymore. In
the differentiated channel you need complicated coordination tools. Your task in
general, is to say that as maximize your own profit is not working well, you’ll need to
coordinate.
Question 4
In previous situation, they were not making so much money. Now Ralph is not only
facing the demand for the original newspaper but he’s also facing the demand for his
private label product. Customers only buy the Private if the Express is stocked out.
It’s the same problem as the Danone-Carrefour issue. It’s cheaper to make n°1
products, there is no marketing strategy but customers may buy it if they have a
stockout on brands.
~
What happens now to the orders? The stocking quantity drops to 𝑞(𝑤) = 409 and the
~
effort level remains at ℎ(𝑞(𝑤)) = 2,25 hours. In this case, if you don’t change the
price or revenue, the order changes but it has no impact on the effort level. Why did
his fill rate go down? He goes down to 73% because Ralph has an incentive to
stockout because of the Private.
~
()
π𝑅 𝑞 = 106, 57
~
()
π𝑆 𝑞 = 212, 9
This makes a channel profit of 319,47. Why doesn’t he make more stockout?
Because he’s only selling Privates on only 40% of lost sales. He doesn’t want to
switch completely to the sale of Privates.
The understocking cost of Ralph is decreasing which implies he’ll order less.
In last question, the retailer charges a slotting fee par stocked unit (here 0,03$). In
the USA if you want to list one SKU you pay around a million dollars. In Belgium,
slotting fees are lower but they are annual. If you go to Delhaize, they’ll list if for
15.000 but they can ask for the supplier to launch campaigns for example.
For which product types are the slotting fees the highest? Alcohol and tobacco. Each
tobacco producer wants to list his products on one shelf. Yet, retailers make the slots
such that it does not correspond to the supplier number of products. It creates the
market value to buy the leftover slot. The slotting fee is not linear. If we come to fruit
and vegetables then there are no slotting fees. There is no brand on those tomatoes.
When the brand name is weak then the slotting fees go out as well. Ralph is putting a
slotting fee because Anna has a branded product.
~
What’s the impact of that slotting fee? The stocking quantity 𝑞 will decrease and the
~
()
effort level ℎ 𝑞 will decrease as well. To Anna, the slotting fee is a fixed retailer cost.
53
For her, the profit margin of selling one article went down. We have a two level
decision; her effort level goes down because she’s making a smaller margin
(marginal profit decreases), thus the mean goes down which implies Ralph
decreases his stocking quantity.
The optimal stocking quantity drops to 349. What will happen to the profit? It is bad
for Anna. What about Ralph? He gets 10 dollars directly and has more stockout. It’s
getting worse and worse to the supplier of the branded product, which makes the
fixed investment. Ralph has a 300% profit increase.
Question 5
We introduce the buyback contract to coordinate the supply chain. How does it work?
Before we just had the wholesale price and now we have the buyback value 𝑏. Here
the retailer can return the article at a price 𝑏. What’s the relationship between these?
𝑏≤𝑤
How about 𝑐? We have to distinguish 𝑣 and 𝑏. 𝑣 has to be lower than 𝑐. 𝑣 is new
money coming it whereas 𝑏 is just coordination between the two players, there is no
fresh money coming in.
0
In (a) 𝑤 = 𝑤 = 0, 8.
^ ^ 0
( )
Π 𝑞, ℎ → 𝑤 = 0, 8 𝑎𝑛𝑑 𝑏 = 0, 76
If we take away all the risk from Ralph, this is when 𝑤 is equal to 𝑏, then he orders
infinitely.
Qualitatively, 𝑏 transfers part of the risk of unsold inventory from the retailer to the
supplier. 𝐶𝑂 goes down, which means the fill rate goes up (fill rate formula).
What happens to Ralph profit? It approaches 0. You can give him a little bit so he
runs the show with following contract 𝑤 = 0, 95 𝑎𝑛𝑑 𝑏 = 0, 93.
If there is a franchise fee that Ralph will have to pay, what will the wholesale price
be? If Sheen requires Ralph to pay a fixed fee, I’ll put 𝑤 to be 𝑐. What does he do
then? He has the true profit, how much does he stock? He sees the marginal cost
and the fixed fee doesn’t changes anything so he decides to stock the optimal
amount. In the franchise case, 𝑤 is equal to 𝑐 and 𝑏 is equal to 0. Anna only makes
profit on the fixed fee. If the fixed fee depends on the mean, she’ll increase her effort
level.
54
Debrief: Mattel Supply Chain Organizations
What are the trends in the toy sales? We are buying more and more toys because
women get children later, when they have more money. There is an increase in
overall expenditures but Mattel now has a challenge because the growth is in the
higher age ranges and they have large market shares for smaller age ranges. Today
the big money from the toy industry is coming from cobranding (Disney): toys of
Despicable Me that come out at the same time to movie is. The product life cycle is
getting shorter and shorter. Two movies a month are launched per month for kids.
What’s the traditional Mattel toy: Barbie.
Mattel would like to have all items sold by the web, directly to the customer. Toys are
seasonal products (Christmas). In the video game industry the demand is more
smoothed. Kids that grown up will buy the video game when they want to and not
wait till Christmas. Moreover, they have no marginal costs and the manufacturer sells
directly to the customer. This is the best situation to make money. That’s not the case
in the traditional toy industry.
Where are the toys physically made? Mostly in China. That means we have long
supply chains, you need to know how to deal with that. Media products are not made
in China as well.
How about the cost structures? What is the share of costs? Licensing costs are a big
part. If you remove the licensing costs you have a third of margins, 80% of product
and 20% of freight. The wages in China are lower but do increase (50% per year)
whereas in Mexico the wages are much higher. What’s the link with the rest of the
story? If labor costs increase, whereas the freight is fixed, you will have to think if you
produce everything in China or if you want to produce some in closer markets,
Mexico for example. It’s better to consider having some part of the capacity close to
the market and save the bulk in China. Indeed, China’s wages will continue to
increase.
Question 1
55
We have multiple toys but they are independent, which means we can use the
Newsvendor model.
( ) ( )
π 𝑞1, 𝑞2, …𝑞𝑛 = π1 𝑞1 + π2 𝑞2 + … + π𝑛(𝑞𝑛) ( )
When we find the FOC of this,
δπ𝑖
δπ
δ𝑞𝑖
= 0+ … δ𝑞𝑖 (𝑞𝑖) + 0 = 0
0
We have separable profits. We have a function for each 𝑞𝑖 .
𝑛
If there is a capacity limitation, we have for each product a limitations ∑ 𝑎𝑖𝑞𝑖≤𝐾.
𝑖=1
What does it change?
(
∏ 𝑞1, 𝑞2 ... 𝑞𝑛 )
𝑛
𝑠. 𝑡. ∑ 𝑎𝑖𝑞𝑖≤𝐾
𝑖=1
𝑞𝑖≥0, ∀𝑖
How do we interpret the lambda? It’s a penalty, in euros per hour, of not respecting
the constraint. This number turned out to be 60. It brought down the capacity of
34.234 to 22050 and the production of 564.700 to 400864. Only 45 items are
produced. When capacity is too expensive, we start not to produce items that have
low margins. Once you have that penalty you take only the most profitable items and
that’s why the profit didn’t dropped that much regarding how much capacity you lose.
Question 2
^ ^
You get two sets of decision variables: 𝑞𝑖 = 𝑞𝑖 + 𝑄𝑖 where 𝑞𝑖 is the in-house
production and 𝑄𝑖 is the outsourced production.
^
− ∑ 𝑐( 𝑞𝑖) − ∑ 𝐶 𝑄𝑖
𝑖 𝑖
( )
𝑛
^
𝑠. 𝑡. ∑ 𝑎𝑖𝑞𝑖≤𝐾
𝑖=1
^
𝑞𝑖, 𝑄𝑖≥0 ∀𝑖
56
If you have a fixed cost per unit, you will produce the item either in-house or
outsourced. But that’s not working as we have a capacity limitation. Outsourcing
doesn’t need to be cheaper one by one but it also competes with the machine
capacity.
− ∑ 𝑐𝑖( 𝑞𝑖 − 𝑄) − ∑ 𝐶 𝑄𝑖
𝑖 𝑖
( )
𝑛
𝑠. 𝑡. ∑ 𝑎𝑖(𝑞𝑖 − 𝑄)≤𝐾
𝑖=1
Now all products are made again. What happens to the marginal value of capacity, λ?
It goes down because of the outsourcing option availability. Is it likely to have lower
in-house production costs than in outsourcing productions costs? They are both
located in China so they’ll be quite the same: salary, equipment etc. are the same.
Just for some things were you need a specific machine the costs may differ.
Question 3
We have capacity limitation and outsourcing but now we take into account the supply
chain. We need retailers to sell our toys. What do we expect? Overall profit will go
down as we have an imperfect coordination mechanism. The retailer will order less
than Mattel wants them to. The coordination will go down and we will have to split the
profit in a very specific way. It’s not very flexible.
There can be two kinds of compliance. There can be forced compliance – the retailer
has to be delivered what he ordered – or voluntary compliance – if the supplier’s gets
a negative profit to produce one more item, he can choose not to deliver it, he’ll
de-list it. Both of them exist but they have different contracts. Voluntary is usually the
most realistic.
Depending on what compliance regime you solve you have different values. The
number of outsourced items will decrease. What’s the coordination loss? It’s the
integrated profit compared to the differentiated profit. The coordination losses is
about 200.000$, and the ration is about 8%.
Question 4
What kind of contract is this? It’s going in the direction of a revenue sharing contract
because they do not take the salvage value. In fact, it is a real contract. Is it likely
they take physical toys with revenue sharing? No it’s not. The supplier cannot verify
how much you sell from the Barbie doll and on the complementary products.
The contract here says that for one market, I will set a recommended retail price and
then we’ll fix a share of them. Where is the biggest market for Mattel? The USA.
Mattel’s profit goes to 1,4 whereas the retailer keeps the same profit as before thus
total profit drops. We delist 4 products. Conceptually what’s happing in this situation?
We had before complete freedom, degree of freedom was 50. Now we have only one
degree of freedom anymore, ∝. Doing that, I have a problem fitting everything. With
57
voluntary compliance, it doesn’t work. It is an inferior contract than the wholesale
price contract. If we pay this for 2 periods, what will happen in period 2.It’s not very
robust unless Mattel recommends the prices.
Question 5
Here we step in to the real revenue sharing contract. What is our intuition now? This
mechanism is supposed to be good because in theory it leads to full coordination.
What is an acceptable transfer price? The first level of acceptability is feasibility.
What does it means that both parties compare to the default situation in question 3?
This means the retailers
If Mattel proposes something, they will chose something that increase their profit
under the condition that the retailer’s profit doesn’t decrease.
π𝑆(𝑞(ß, 𝑤))
𝑛
𝑠. 𝑡. ∑ 𝑎𝑖(𝑞(ß, 𝑤))≤𝐾
𝑖=1
3
π𝑅(𝑞(ß, 𝑤))≥π𝑅 (win-win constraint)
𝑞(ß, 𝑤)≥0
ß, 𝑤≥0
𝑤∈[𝑝, 𝑐]
I could also do it differently: not optimizing the supplier’s profit but the retailer’s profit
subject to what I want. This should lead us to the same point.
The only reason why you have no full flexibility (nobody get zero), is because of the
reference point. You’re not reasoning from 100 to 0 but in a much more narrow way.
Question 6
You have the supplier, the license holder, Mattel and the retailers. The system, here,
must be coordinated by Mattel. Rather to have a variable fee, Disney wants a fixed
fee. What is the intuition here? It increase the profitability of Mattel so the under
stocking cost is diminished. This increases the willingness to have higher quantities.
That’s a good idea. Mattel wants to distribute this over the whole supply chain to
benefit of this.
The retailer’s profit is kept at the acceptance level. Disney is considering accepting
and Mattel does the offer. The lower reservation level of the lumpsum payment is the
profit of the licensing costs Disney had before. The maximum is the maximum
acceptable by Disney. Is it likely to happen? Disney has to accept it. What’s in it for
Disney? Disney gets rid of the risk of the downstream market. The risk is removed;
before anything is sold they are paid. Disney will become a cobrander. A lumpsum
fee removes that issue. The upside game is that if you make more than you though
and you pay a lumpsum fee, you can make much more money. The production
58
quantities are verifiable by Disney. This creates too much risk for Mattel so he will not
accept this.
Question 7
I can sell items with more premiums if I can get a buyback contract with a transfer
price no higher than the current retailer prices. This is an assortment value. You
come in to the store and you have a big shelf with only Barbie products. If one
product is not sold, the retailer could decide to get rid of it. But doing that they would
replace a Barbie product by an ugly doll. Therefore, the retailer wants a buyback to
not remove the item from the pink shelf. Let’s make a Barbie corner but then I want
buyback because doing himself a Barbie corner is way too risky.
We shifted back a local risk from the retailer to a global risk at the supplier. The
quantity increases and all products are listed. The total profit reaches 9 million. The
buyback contract is coordinating but it will cost something in the profit of the supplier.
59