Where does the data for such a simulation model actually come from?
We sometimes actually ask ourselves this question. Of course, it clearly depends on the phase of the project. The earlier we are involved in a project, the more abstract the data usually is. And you have to distinguish between greenfield and brownfield projects.
In a greenfield project, this is usually data that cannot be generated from anywhere, but is in principle prepared or automatically generated on the basis of assumptions. This applies in particular to order data. The corresponding questions are linked together:
- What product range will I have in the future?
- Which customers will order from me?
- In which order structure will the customers order from me?
- How do I then produce these orders?
- How do I ship these orders?
And, crucially, the more accurate I want to simulate, the more accurate the data has to be.
One example is the supply chain: If I am simulating across an entire supply chain, then it is probably enough for me if I know roughly the lead time per location in my supply chain. So I know that if I send an order to the Cologne site, it will take two days for the order to be processed and it will then be transported to my receiving warehouse.
However, if I want to simulate a distribution system down to the conveyor level, then of course I have to simulate the individual conveyors. I need to know how they behave and what physical properties they have. And then I need the corresponding data.
In a brownfield project, it would of course be ideal if I could take the order data from an SAP interface – for which, by the way, there are now certified SAP interfaces for simulation – and project it into the future using factors, for example, by saying: In the future, for example, I will have 20% more customers, I will get 50% more orders, I will have a larger product range, and so on.
And from this I generate the actual data, which of course makes me much more accurate than if I simply generate it from scratch.
After receiving the data, the question then arises: Can we process it 1:1? But we will answer this in our next article.
By the way, our self-developed tool SimAssist supports data analysis and helps to understand and visualize the correlations in your data.
More info at: https://simassist.de/lp_simassist-datenanalyse/.