Car manufacturers have stopped due to their home-made bullwhip

Introduction

This article is based on a series of interviews with specialists and employees of companies in the European supply chains of Automotive. In addition, information is used from a monthly survey[i]  that was done amongst a few dozen companies active in those chains. The graphs are based on a simplified supply chain model that they built in System Dynamics for the situation around semiconductors in Automotive.[ii]^,[iii]

Semiconductors

The model comprises a supply chain that runs from Tier5, via Tier4, -3, -2, -1 to the OEM, and via the Dealer ends at the Consumer. The production process for Semiconductors consists of many steps, which makes the lead time from the so-called wafer start in the  fab (the semiconductor factories in Tier4), via the Backend (Tier3, where wafers are cut into individual chips that are packaged), the Components (Tier2, e.g. transistors, diodes or complex chips) and Electronics (Tier1[iv], for example navigation systems or parking sensors) to the build-in just-in-timein the assembly line of the OEM can be more than 3 or 4 months.  Tier5 consists of suppliers of raw materials for the fabs. Because chips develop fast and thus age fast, stocks are kept low.

Figure 1. Model structure

As the margins are low in the fabs, they always work at a high utilization rate. Capacity expansions can take up to five years and require a lot of capital. For that reason, Tier3 customers at an early stage have to make contractual agreements for their share of the Tier4 capacity and provide a planning well in time. Shortly before the planned assembly time, the OEM calls of the required Electronics from Tier1.

When the OEM went into lockdown, not only the call-off was cancelled, but also the planning and the capacity claims in Tier4 were allowed to expire. In response, Tier4 started producing for other markets. In parallel some other problems manifested themselves. When the OEM production started to rebound, the required capacity had to be assigned from scratch. This caused a delay of multiple months before the first extra products came out of the fabs.

Lockdown in Automotive

In March 2020, most OEMs in Europe closed down their assembly lines because of Corona infections and triggered by closures in China. In the same period, the consumer willingness to buy collapsed. Due to the general lockdown, traffic intensity sharply declined and an economic crisis developed. The OEMs did not continue production, because continuing in the previous crisis resulted in large stocks of unsellable cars. Also most of the dealers closed their doors. After one-and-a-half month, dealers and assembly lines started up again.

Figure 2. Consumer demand from 1991 till 2019 is based on the 11mma of reported car registrations[v], brought forward 90 days.  As of 2020 this has been replaced by an estimate based on media and interviews. Index 2019 =100.

During their lockdown the OEMs did not order much raw materials. Tier1 saw her sales to Automotive decline with 70% and had to reduce production. Tier1 also reduced her stocks in line with turnover, with some delay, by reducing her purchasing. As a result, Tier2 saw her Automotive sales plummet, and partially had to close as well. Also Tier2 wanted to reduce her stocks, for as far as intended for Automotive. This closure and stock reduction caused a big production decline in Tier3 companies, etc., etc.

When the assembly lines wanted to restart, the OEMs started ordering more raw materials. Tier1 supplied from stock, rapidly moved her production in higher gear and ordered raw materials as input for production and to bring her stock back up. Tier2 moved production up and ordered extra from Tier3 for production and for stock. Tier3, 4 and 5 did the same. In summary, the demand shock of March/April 2020 overlapped with a capacity shock. Together they caused a temporary demand shock and stock adaptations amongst the layered suppliers, after which the OEM quite rapidly ramped up again. As from the beginning, it was clear for the experts that this would lead to a classic bullwhip. [vi]

Survey

Although the number of answers in the Survey, and in particular about inventories, is too small for significant conclusions, it does give an insight in what is happening.

Figure 3. Average sales (blue) and average stock (red) of Tiers 1-4 in the survey of January 2020 – February 2021. Index 2019=100.

The survey results can be compared with the average sales and stock in the model:

Figure 4a. Average sales in Tier 1 – 4 in survey (red) and model (blue).

Figure 4b. Average stock in Tier 1-4 in survey (red) and model (blue).

In both survey and model the sales level first goes up, then down, then up again, whereby the timing is quite similar and the absolute level is reasonably similar. Stocks in both survey and model first go up, then down. Recent stock moves in the survey are not reproduced in the model.

Figure 5. Average sales (blue) and average stock (red) 1991 - 2027 of all the Tiers in the Automotive industry (model). Index 2019=100.

This figure shows that over a 30-year period the stocks in the model followed sales, even in the credit crisis, but now it is taking an opposite direction.

Modeling and Results

Consumer demand was indexed at 100 units/day. The model is demand-driven: the consumer “pulls the cars out of the chain”. This means: without demand no sales. Goods travel downstream; orders travel upstream. Stocks are low, so can quickly become critical.

Figure 6. Stock in Tier1 (blue), 2 (red) and 3 (green)

Stocks will continue to be volatile for quite some time.

Figure 7. Production in the OEM assembly line in period 2020-2025 (model), with index 2019 =100

OEM Production shows a deep dip (1) during the lockdown in March/April 2020, and a second dip (2) early 2021, caused by lack of semiconductors. According to the model there will be additional dips (3) in 2022 and (4) in 2023. Whether the short extra dip takes place in the model at the end of 2020 (between 1 and 2) depends on the stock levels.

Thought experiment

Between restarting in May 2020 and stock-out in January 2021 the OEM draws an estimated 5 months’ sales (index 2019[vii]) from the chain, in comparison with the April level, while at the beginning of the supply chain no extra product enters. All stocks flow to the OEM, until the moment in January 2021 that Tier1 is empty and the assembly lines have to stop. This means that there must have been an amount of inventory equal to some 5 months’ of sales in the semiconductor supply chain.

Figure 8. Sketch of the semiconductor supply chain as a tube that connects warehouses with product (blue balls). Duet o the temporary stop of the fabs for automotive products, an empty bubble is moving through the tube, emptying all warehouse it passes under way (red arrow).

This also means that the 2020 backlog of the fabs in comparison with the OEM (the empty bubble in above picture) must have been bigger than the warehouses (which are equal to the stocks minus Work in Progress minus Goods in Transit), because otherwise Tier1 would not have been empty in January.

Indirect effect

The closure due to lack of semiconductor parts also impacts other chains: the ordered paint raw materials, rubber, glue, shock absorbers and car doors are accumulating in their respective supply chains, and in their turn are causing temporary upstream closure of factories, which move through the chain as bullwhips and which will cause new shortages and price volatility in a few months’ time.[viii]

Caveat

The model is based on interviews with a limited number of companies in the supply chains, public data and surveys. The shape of the various oscillations strongly depends on the average stock levels, the circumstances and behavioral and will therefore deviate from the model. More research and data will be needed to make truly reliable predictions. The authors do not accept any liability for the effects of business decisions that will be taken based on this article.

Conclusions

1. According to our model, the unbalance in the Automotive supply chains will continue for a number of years. The current closure will not be the last this year. At the moment, all chains are heavily stressed and regularly run out of stock. The number of Force Majeure announcements grows every week.
2. The survey showed sales results for January 2021 that were on average above 100%, with some companies at 140%. This is the upward peak of the bullwhip, which is now going down fast. The OEMs did not produce above their 2019 level.
3. In times of shortages, one often sees stock piling and this is also what we hear in interviews. Stockpiling increases volatility and raises prices. After a stockpiling peak one usually sees a demand dip, often followed by a second peak.
4. During the pandemic, companies that used to forecast their turnover based on extrapolation of historical sales figures, switched to sentiment, or vested their hope in AI. It is advisable to explore deeper in the chain, multiple steps upstream and downstream, including the end market, while combining this with an assessment of the economy as a whole and the usage of scenario’s.
5. The semiconductor chain consists of many links, with high utilization rate, low stocks and long lead times. She is like a tight string, susceptible for disturbances.
   1. Long lead times are always a potential source of strong volatility and pose a big risk for pork-cycle type waves.
   2. Capacity extensions in fabs take years and cause the well-known pork cycle in semiconductors, but this cycle seems less relevant for the current discussion. The plans of the Biden government to solve the problems in automotive by investing 27 B$ in fabs in the USA, nor the announced expansions by TSMC will solve the current crisis.
   3. Due to the high utilization rate, there is little opportunity to catch up on a lost production-hour and shortages will continue to exist for a long time.
6. The bullwhip can be dampened. Companies should behave in an anti-cyclic way by making slower stock adaptations and by using all unused capacity to build stocks. In general stocks are the solution to get out of this crisis. In due time, when the chain has stabilized, the stocks should be brought back to lower levels in slow fashion.
7. For years, in production and supply chain management, the Toyota Production System (TPS) is used for stability and optimal efficiency. Protection against major disturbances in such an environment is possible with so-called resilient supply chains, a holistic approach that goes further than classic risk management and is a combination of chain cooperation, agility, culture and chain structure.[ix] Examples of cooperation are info-exchange between the Tiers and Chain Managed Inventory (CMI). i.e. stock that serves the need of the whole chain.

Footnotes