Zum Inhalt gehen
Automotive Hub-Web-Banner-1
Supply chain

AI-based simulation for automotive supply chains: Precise forecasts despite high variety


Consumers today enjoy having options when purchasing a vehicle. However, as the number of ordering options increases, predicting demand becomes increasingly complex. Automotive expert Maid Jakubovic explains how cloud-based technologies and innovative forecasting methods with artificial intelligence (AI) now help plan production efficiently and precisely, despite an almost unlimited number of variants.

Mr. Jakubovic, when the first vehicles came onto the market, it was still easy to buy or sell a car. There were only a few variants!

That is true. There were just eight different variants of the first production vehicle, the Benz Velo. The Ford Model T already had 60 more. The range of options was limited. Tesla CEO Elon Musk initially offered the Model 3 in only six colors and two tire sizes, totaling 12 variants. Now, there are more options, and the number of variants has increased to 120. Tesla was therefore initially able to literally sell what was in stock. 

Currently, the configuration of a BYD, Lucid or Tesla is still comparatively straightforward because there are not as many variants as we are used to with combustion engines. With the Mercedes S-Class, the number of variants is practically infinite. Will there also be this variety for e-vehicles in the future?

The e-vehicle market is currently changing from a seller’s market to a buyer’s market. From a global perspective, the built-to-stock and built-to-order markets in the automotive sector are roughly balanced. In Germany, for example, customers like to configure their vehicles individually and are willing to accept longer delivery times. In China and the USA, on the other hand, they prefer to take it straight from the dealer. The art of a car manufacturer now lies in being able to serve both markets reliably and quickly.

How can you determine which vehicle configurations customers want in the future?

Two to three years before a new model is built, automakers start developing scenarios based on specific assumptions. This is not easy: some vehicles are made up of 50,000 individual parts. For a single side mirror, there can be numerous options, including different colors, blind spot assistants, and winter packages (with built-in heating). Different wiring harnesses are needed for these options, making up to 100 variants possible for a single side mirror. It’s similar to car doors, which can vary in color, upholstery, switches and can also contain speakers for a sound system. However, car manufacturers can always draw on past experience with similar models and incorporate this into their forecasts.

Capgemini has now developed the “Capgemini Order Simulation for Automotive” forecasting tool in co-innovation with a premium vehicle manufacturer as part of the joint strategic initiative with SAP in automotive cloud solutions, which can do both – incorporate historical data and develop future scenarios. What advantages does the solution offer?

There are three key advantages:

  1. The number of variants is not limited: In current SAP systems, more than 2,000 variants are not possible without workarounds. The new tool is not limited to a specific number of variants, as cloud offers almost unlimited computing capacity.
  2. Detailed planning: production planning is based on the smallest unit of data, at order and part number level. Advantage: Production and set-up times can be planned in detail, allowing the forecast to be continuously improved.
  3. Forecast plus scenarios: The new tool can incorporate both historical data and scenarios into its forecast and continuously refine it. Ideally, orders for comparable models can be included in the forecast. If this is not possible, assumptions are made, and this forecast is taken to the market. After the first weeks of a model’s market presence, the forecast can be further improved, and production adjusted accordingly. Machine learning helps by identifying outliers.

How long does it usually take for a custom-configured vehicle to be delivered?

It varies, but up to ten months is not uncommon. The current benchmark is three months, although individual details such as the color can still be changed six weeks before the start of production. However, the solution from a premium manufacturer is based on legacy software, which must be maintained individually and can hardly be upgraded in terms of performance. The industry cloud solution Capgemini Order Simulation for Automotive is based on a modern infrastructure. If necessary and desired, we can adjust our forecasts on an hourly basis.

For more information we recommend a short explanatory video and an expert discussion from Hannover Fair 2024.

Exchange with an expert:

Would you like to address this topic, or have you already gained experience with it? Please write to us! We look forward to exchanging ideas with you.

Unsere Expert*innen

Maid Jakubovic

Managing Business Analyst

Anke Rieche

Global Automotive Program Lead
Anke ist eine Business Development Expertin mit 20 Jahren Erfahrung in den Bereichen Software, Infrastruktur und Beratung. Als hochmotivierte Teamplayerin mit ausgeprägter Kundenorientierung hat sie sich einen Namen für die Entwicklung und Umsetzung von Markteinführungskonzepten gemacht, insbesondere im Zusammenhang mit den SAP-Plattformen S/4 HANA und Intelligent Enterprise, vor allem im Automobilmarkt. Anke ist davon überzeugt, dass Automobilzulieferer und OEMs durch den Einsatz der Automotive Cloud-Lösungen von SAP, einschließlich der gemeinsamen Entwicklungen von SAP und Capgemini und der Co-Innovation mit Pilotkunden, neue Dimensionen der Agilität und Geschwindigkeit erreichen können.