Chapter 4 – Philips Semiconductors (PSC)
4.1 Introduction: A Thoroughly Cyclical Industry
Why is the semiconductor supply chain always late in responding to changes in market demand?
Capacity investments in new factories, would typically start at the peak of the market, but only result
in additional production capacity in the market after several years, so when market demand was
already down again.
4.2 The Problem: Why Are We Always Late?
Philips measures its delivery performance with two indicators: RLIP and CLIP. RLIP stands for
Requested Line Item Performance, CLIP for Confirmed Line Item Performance. Line items are order
lines, so every part of an order that is for a specific product type. The RLIP measures how many out of
a hundred requests from customers for delivery of a product can be accepted, CLIP measures how
many out of a hundred accepted orders are delivered to the customer on time.
- RLIP is always lower than CLIP. This is because customers can always ask things you cannot
do for them, which lowers your RLIP. With CLIP you already promised that you could meet
the customer’s request.
- Delivery performance of both measures went out of the window in 2000, because the
company could no longer supply what the market wanted.
- There are hardly any delays, not on a monthly level of detail at least, between the drop in
demand and the resurgence of RLIP and CLIP performance.
Inventory peaked half a year later than the sales rate peaked.
Building a new semiconductor fab takes years.
The capacity PSC needed in 2000 became available when it was no longer required.
4.3 A Model Of An IC Manufacturing Supply Chain
The basic ‘physics’ of the supply chain at PSC can be described with three interconnected chains of
flows:
Flow of orders, or the order fulfilment process
The flow of products, or the production process
The flow of capacity, or the capacity management process
4.3.1 The Flow Of Orders
Once orders from customers have been entered in the order fulfilment IT-system, they move into
backlog, which is the accumulation of orders that still have to be delivered.
There is one KPI that is directly derived from this flow rate, which is “delivery performance”. This is a
proxy for the common indicators for the CLIP and RLIP.
4.3.2 The Flow Of Products
Once products are started up, they become WIP. It takes a considerable delay before they have been
assembled. ICs that come out of production are added to finished products.
,There are two KPIs specified on the basis of this flow: inventory and capacity utilization.
4.3.3 The Flow Of Capacity
In semiconductors, it takes very long to increase capacity.
4.3.4 Information Processing And Decision-Making In Production Planning
In order to determine how much production should be started, one needs to know:
How much future demand is expected to be
How large the discrepancies are between present and desired stock levels
How much capacity needs to be loaded to break even
In capital intensive industries, you do not just make what you think you can sell, but you will also
have to sell what you make.
Nemawashi: behind the scenes negotiations
4.4 Model Behavior: Understanding What Happened In The Real World
4.4.1 Model Calibration: Fitting The Model To Historical Reality
We have one very clear input signal: weekly demand rates, and four obvious output rates that tell us
about performance:
Delivery performance in terms of RLIP
Production start rate, in terms of wafer starts
Inventory, in terms of inventory value
Capacity utilization, in terms of factory loading
4.4.2 Intermezzo: Oscillation In Supply Chains
Oscillation: values of various performance indicators go up and down in a more or less regular
manner, and hover around some kind of average value, which they seem unable to stabilize at. When
you see oscillation, you can suspect that the underlying system is inherently unstable.
People in this industry know that their business is a cyclical one, but they assume that this cyclical
behavior is imposed upon them from the outside, from their customers. They do not see their own
production system as inherently instable, and therefore creating oscillatory behavior on its own.
4.4.3 Understanding The Key Feedback Loops That Drive Behavior
Two balancing feedback loops.
B1: The production rate adjustment loop
When the order backlog increases, this means that the order book grows. This means that the required
production will go up. This will lead to more work in progress, which will make a higher shipment
rate possible. More shipments will deplete the backlog. When the order fulfilment rate becomes
greater than the order rate, the order book will start to shrink and the production forecast will go down,
the WIP will decrease and the shipment rate will go down as well.
B2: The capacity adjustment loop
,When the order book grows, the forecasted capacity required will also grow, albeith with a
considerable delay. With yet another considerable delay this will lead to capacity additions. The
increase in capacity will lead to more shipments, which will deplete the order book. A smaller order
book leads to a lower capacity forecast, hence to a lowering of capacity.
These two loops make sure that demand and supply stay in sync.
4.5 Policy Analysis
How to deal with cyclical order patterns and changing capacity?
- Technical complexity
- Organizational complexity: many things that make sense to do technically, are not feasible
organizationally.
4.5.1 Operational Performance Improvements: Saying No To Customers
At the level of daily operations, problems start when customers place orders that you cannot really
handle.
4.5.2 Tactical Performance Improvements: Shortening Planning Delays
Planning takes place on tactical level. Two kinds of planning: production planning and capacity
planning. Both of these processes have delays associated with them.
The chain becomes more flexible when planning delays are shorter. Flexibility simultaneously creates
more agility, which is good, and more volatility which is not good.
4.5.3 Strategic Performance Improvements: Lower Utilization And Shorter Acquisition Of
Capital Investments
At strategic planning level, it is all about capacity and money.
Shorter acquisition delays has pro’s and cons. One obvious cost is that of higher production costs,
since building new fabs takes time and building them faster will cost you more.
This policy helps to increase agility at the expense of stability. Production can now ramp-up sooner
and peak at an even lower level, but production capacity is also reduced more aggressively, leading to
more amplification later on.
4.6 Conclusions: The Curse Of Cyclicality
Capacity is so inflexible and costly because it is state-of-the art and very complex, as is common in
innovation driven industries.
, Chapter 5 – ASML
For ASML the order intake in the beginning of 2000 was much higher than it had ever been before,
and there was just no easy way in which all these customers could get their machines on time.
5.1 Introduction: Order Status Visibility During Capacity Shortages
5.1.1 The Problem: Real-Time Visibility For Order Status For Businesses
Consumers can see nowadays a lot of order information and there is a lot of real-time visibility. Real-
time order visibility no longer wins you the customer order, it just ensures that you qualify as a
supplier.
5.1.2 Desirability And Inevitability Of Real-Time Order Status Visibility
All this visibility is wonderful and sensible use of IT, but it remains nevertheless mainly limited to
consumer markets. In B2B-markets, real-time order status is less widespread. This raises the question
if companies should invest in IT that will use this innovation from the consumer world for their own
business transactions as well.
Why should such wonderful use of IT be not desirable? The nature of transactions in B2B settings is
different from B2C environments. These differences become exposed when there is a capacity
shortage, leading to longer and less reliable deliveries. When this happens to consumers, they
complain, or cancel their orders, or just put up with it. Business customer would act differently.
Typically, their orders are for repeat purchases, they will often have a pipeline of pending orders
needed to maintain their own order pipelines at the required level and partly for speculative reasons,
so-called shortage gaming, they will start asking for more then they really need. This is called demand
rate amplification, duplicate ordering or phantom ordering.
In the lithography industry, shortages are an inescapable fact of life. The bullwhip effect causes
demand fluctuations to be higher upstream in the value chain, and production equipment
manufacturers are typically far upstream indeed.
5.2 Intermezzo: Upstream Order Amplification
Throughout the industry and throughout time, there is a tendency for fluctuations in demand to
increase as this demand is translated upstream in the value chain.
5.2.1 The Forrester-Effect In Supply Chains
Jay Forrester was able to confirm that the company’s policies actually amplified any oscillations that
might arise.
Forrester effect = bullwhip effect
5.2.2 The Destabilizing Effect Of ERP Inventory Control Algorithms
Forrester’s analysis showed how delays in perception and misaligned decision-rules to manage single-
echelon stocks create instability in supply chains that consist of multiple stocks.
You need to buy enough to:
- Compensate for your expected sales rate