Creating supply chain models or digital twins of your supply chain can be tremendously useful in understanding the breakdown of your logistics costs, productivities associated with logistics activities and the performance elements that make up the overall service level to the customer. It can also be very useful in establishing a firm foundation from which the future impacts of volumes changes, strategy changes or process changes, can be modelled and understood in advance.

However, it is easy to get it wrong and rather than support business decision-making, it can lead to misleading outputs and hence undermine business planning.

Here are 5 common mistakes when carrying out supply chain modelling that we have collectively observed at Bisham Consulting:

1. INACCURATE ASSUMPTIONS

Be mindful of basic foundational assumptions that have not been validated by operations or undergone insufficient due diligence. This can have, at worst, potentially catastrophic impacts and undermine operational performance of newly implemented supply chains.

2. INSUFFICIENT DATA QUALITY OR GRANULARITY

This must be of sufficient quality and again validated by operational fact holders. You should not use aggregated data as an input but raw transactional data of sufficient granularity, consistent with the analysis you are outputting. Using existing KPI’s, without checking and validating them first, as a short-cut to building your supply chain model can also lead to inaccuracy.

3. NOT TAKING VARIABILITY INTO ACCOUNT

This is often not accounted for, as historical averages are used instead. Variability occurs in lead times, volumes, (e.g. seasonality effects), shift performances. This needs to be understood and accounted for when modelling for operational change, otherwise a falsely optimistic conclusion will be made.

4. NOT APPLYING SENSITIVITY ANALYSIS TO THE OUTPUTS

Sensitivity analysis allows you to determine the risks associated with the model outputs and derived conclusions. Operational assumptions made around your future state will never be 100% accurate, so take an optimistic, neutral and pessimistic view and change your key assumptions by agreed factors e.g. +/- 10% , 20%,30% and see what impact this has overall. Then present outputs as a sensible range, rather than a single discrete number.  

5. NOT FACTORING OUT THE BENEFIT OF HINDSIGHT

When applying historically derived data, (which your model will be based on), you will need to account for the impacts of human real-time decision-making in a dynamic operational environment. There is a danger of building future state models that do not account for the benefit of hindsight and are hence over-optimistic in their findings. Your model may tell you that you can save 20% of logistics costs. However in a real-life situation you don’t have all your planning information in place well in advance and it is unlikely to be 100% accurate therefore operation disruption, so-called ‘noise’ occurs and will reduce expected benefits. Hence, a pragmatic view needs to be taken of this with contingencies built into your model to account for this.