Digital models of value chains or entire companies support management decisions on matters ranging from strategy to medium-term planning and day-to-day operations. A good example is margin optimization in complex chemical value chains.
Production networks, often referred to as “Verbund,” are multistage value chains that span one or more sites and feature combined production processes at individual stages and multiple outlets to the market, including for intermediate products. In such a network, decisions have to be made constantly in the short, medium, and long term. What is the most profitable application for a specific raw material or an intermediate product (allocation decision?) How does this change in shortage situations or when the price structure shifts? What should be the price of the final product in order to realize a profit across the entire value chain (pricing decision)? How do I design customer contracts so that I can pass on changes in raw material prices more effectively (customer management)?
These decisions are tricky, primarily because value chains in large companies span multiple business units. And their interests may differ considerably. Prices might cover costs for one unit, but result in losses for another. Decisions in the interests of the company as a whole must therefore be based on a consolidated view of the entire chain.
Today, decisions like these are still often made in frequently held coordination meetings in which experts from all the units hurriedly interpret data compiled from many sources to come to a conclusion.
New technological capabilities
The technical capabilities have now changed. With the help of familiar programming languages such as R or Python, it is possible to create statistical models that can realistically represent an entire value chain. Such models enable
- Transparency of product and value flows.
- Simulation of different scenarios involving raw material and product prices, capacities, etc.
- Calculation of an allocation or a pricing decision that leads to an improved overall margin across the value chain.
The main advantage of such a model is its optimization function, that is, its ability to not only represent, but also improve reality. Artificial intelligence comes into play here (as well as in individual model elements for transparency and simulation). Digital models make it easier for decision-makers to analyze data. They help them concentrate on the really important questions. And even if the models don’t necessarily determine a decision, they limit the option range to those that are truly relevant.
Faster, cheaper, and easier
But it’s not sheer performance alone that makes such models so attractive. Nowadays they can be generated quickly (in around eight weeks) and inexpensively. What’s more, they are visually appealing and relatively easy to use. As such, they help the numerous departments involved in decisions at a company (production, supply chain, sales, finance) find a common language.
All this is achieved by applying generally established techniques to “evergreen” challenges of the chemical industry. What makes the overall solution innovative is a third factor: The project approach used to develop the statistical model of the value chain.
The approach makes the difference
In traditional IT projects, someone would inquire into the requirements for business operations. A specification would then be drawn up on this basis, and a programmer would convert it into a tool. This sequential approach is not only time consuming, but prone to error due to multiple handovers. There is also no way of knowing whether subsequent users will actually like the results.
A value chain model should always be developed in an “agile” project approach. In the process, chemical experts, data analysts, and programmers work hand in hand from start to finish. A prototype of the model is created at a very early stage so that the users can practice and refine their requirements. The programming is done in several sprints. After each one, the users assess the status of the work and suggest improvements. As a result, fundamental changes can be made to the model, even at a later stage.
The possibilities of data analytics extend far beyond the control of individual value chains described here. The vision is to model an entire company, with its product and value flows and most important profit drivers. This would make it possible, for example, to rapidly estimate how the divestment of individual products, businesses, or production sites would affect the profitability of other units and of the company as a whole.
Please feel free to contact me if you would like to learn more about digital models of value chains, companies, or company networks.
We would like to thank Dr. Yorck Dietrich for his valuable contribution to this article.
