Supply Chain Forecasting and Bullwhip Effect Evaluation Using Simulation Software

Supply Chain Forecasting and Bullwhip Effect Evaluation Using Simulation Software

Infineon Technologies AG is one of the world’s largest semiconductor manufacturers. In 2021, Infineon reported revenue of more than €11 billion with a workforce of 50,280 people worldwide. Following the acquisition of the US company Cypress Semiconductor Corporation in April 2020, Infineon is now a global top 10 semiconductor company.

The semiconductor industry in general is characterized by capital intensity and high demand volatility. Semiconductor demand is unstable, even before accounting for COVID-19, highly dependent on innovation cycles, and is prone to the bullwhip effect.

Supply chain engineers at Infineon have been using AnyLogic for many years because simulation is an effective tool for solving production demand and supply chain problems. At the AnyLogic Conference 2012, they talked about their project on investigating the bullwhip effect in their market. At the time, they combined agent-based and discrete event modeling approaches to build a model of their multi-tier semiconductor supply chain. The supply chain model helped them better adapt to demand fluctuations and reduce the bullwhip effect.


These days, the problems of volatility in demand and the bullwhip effect are even more challenging than they used to be. During the COVID-19 pandemic, automotive demand dropped significantly which resulted in too much inventory. Demand for cars fell because people were working from home and commuting less. Later, the market rebounded, and the increased demand coincided with a global computer microchip shortage.

In the chart below, you can see the correlation between the growth of the global economy and the semiconductor market growth. In 2008-2009, due to the global crisis and economic decline, the demand for semiconductors dropped dramatically. However, in 2009-2010, economic growth recovered, which immediately affected the semiconductor market. This is a great illustration of how the bullwhip effect works and why it is a major concern in the industry.

In 2020, during the COVID crisis, growth in a semiconductor market no longer followed global economic growth patterns. It did not vary as dramatically as GDP because demand for semiconductors fell in some areas while it increased in others. Nevertheless, reducing the bullwhip effect still has high potential.

Infineon reference market

*I-Ref-M = Infineon reference market = Total semiconductor US-Dollar-based market revenues excl. DRAM, NAND Flash, MPU. – Real GDP = Inflation-adjusted (real) Gross Domestic Product of all countries of the world; a total of local values converted within each case current US-$ exchange rates. World real GDP is from the chain-weighted index. Quarterly data (year-over-year growth rates)


In contrast to their 2012 project, the supply chain engineers at Infineon decided to apply system dynamics tools to study the bullwhip effect. They wanted to compare the new results with those from the 2012 study and, this time, to understand the impact of end-market scenarios on the bullwhip effect through the whole supply chain.

Simply put, the engineers wanted to look at the same problem from a slightly different angle. System dynamics is used primarily at the macro level, where general patterns are more important than small details. Using system dynamics tools helps apply systems thinking for the purpose of identifying feedback loops, understanding fundamental problems, and looking at their symptoms.

The engineers had three main goals:

  1. Simulate the recovery in demand from the COVID-19 crisis in the automotive semiconductor supply chain.
  2. Understand the impact of bullwhip effect for different end-market recovery scenarios.
  3. Provide a tool for evaluating collaborative efforts and habits.

To achieve their goals, the engineers completed four key tasks:

  1. Identification of end-market demand recovery scenarios: U-shape, V-shape, L-shape, etc.
  2. Creation of a system dynamics supply chain model in AnyLogic.
  3. Testing of model using historical data.
  4. Sensitivity analysis to see which parameters have the biggest impact on results.

The end-to-end semiconductor supply chain structure can be seen in the picture below:

Infineon end-to-end semiconductor supply chain structure

From right to left, the structure contains four echelons, each echelon describes a member of the supply chain:

  1. Echelon 1: OEMs (original equipment manufacturers)
  2. Echelon 2: Tier-1 supplier
  3. Echelon 3: Tier-2 supplier
  4. Echelon 4: Semiconductor supplier

All these echelons are on a globally aggregated level, which means that the OEM represents all the original equipment manufacturers on the global level, semiconductor manufacturers describe all the global manufacturers of semiconductors, and so on. As you can see in the picture above, the information flow of this supply chain propagates upstream, while the physical flow of products propagates downstream the supply chain.

The four echelons were specifically modeled in the simulation. Each echelon passes inputs through several control loops before outputting to the next stage. Different echelons have different parameters for the same components, including forecasting, capacity, work-in-process, stock, backlog, and supply line management.

In this supply chain model, we assume that the semiconductor supplier reserves are infinite because they are guaranteed by silicon supplier.

The basic structure of the system dynamics for each echelon had several loops:

The basic structure of the system dynamics for each echelon

For the supply chain model’s data input, the engineers used historical data. This provided a base for model verification and for scenario testing.

Light vehicle sales faced a harsh drop during the crisis while electric content per car is gradually growing. Both factors significantly affect demand in the semiconductor market.

Global light vehicle sales and Electronic content per car revenue projections

The end-to-end supply chain model has a simulation dashboard for choosing different scenarios and the variation of a wide range of different parameters when conducting scenario analysis:

The end-to-end supply chain model simulation dashboard


In the supply chain model results, we can see how demand drop affected other parameters.

Supply chain model results

After demand for semiconductors collapsed during the pandemic orders quickly rebound due to end-market demand recovery.

Semiconductor demand recovery

  1. The results of the simulation model show a clear amplification of the change in the end market for light vehicle sales. The more upstream in the supply chain, the larger the drop in the received demand signal during the crisis.
  2. The recovery phase in end-market demand shows high amplification of demand increase. The incoming demand for the semiconductor echelon exceeds end-market demand by about 40% with a doubled amplification compared to Tier-2.

Inventory recovery is challenging due to long cycle times and high demand during the recovery. The high demand coming from increased vehicle production and a greater number of semiconductors in new vehicles. The result is the semiconductor shortage.

Semiconductor inventory recovery

  1. Inventory at the semiconductor echelon rises due to the cancellation of orders from downstream supply chain partners. Inventory at the semiconductor echelon cannot be made flexible due to long cycle times.
  2. In the recovery phase of the crisis, the inventory level of the semiconductor echelon is insufficient. Due to capacity restriction and high demand from the downstream echelon, the inventory level only recovers slowly.

Sensitivity analysis for different echelons and parameters indicates a future direction for increasing overall supply chain performance. Different behavioral parameters show diverse influences on the backlog level towards Tier-2 suppliers, hence, the chip shortage for the whole supply chain. More up-to-date and lower time lag in information flow reduces the backlog level.

Semiconductor backlog and demand

As a result of the testing and analyses, the supply chain engineers from Infineon obtained multiple insights:

The case study was presented by Abdelgafar Ismail and Hans Ehm of Infineon at the AnyLogic Conference 2021.

The slides are available as a PDF. For comparison with other modeling techniques, read the 2012 Infineon semiconductor supply chain study.

Ähnliche Fallstudien

Mehr Fallstudien

Fordern Sie eine Broschüre mit industriellen Fallstudien an (in Englisch)