Let your SC planning tools do what they were designed for.

We are often tempted to intervene in the planning process, convinced that we can optimise and improve the predictions made by our tools. However, there is a growing recognition that our interference in the planning process may, more often than not, add unnecessary complexity and uncertainty. The irony here is that the very actions we take to improve the system might actually inject more noise into the process, making it harder to predict and plan effectively.

To illustrate this, let’s consider the example of demand forecasting. Advanced tools today, powered by algorithms and vast amounts of historical data, can provide highly accurate predictions about future demand patterns. However, when human intervention occurs—whether it’s tweaking the numbers based on intuition, adjusting forecasts to reflect unproven assumptions, or simply overreacting to short-term market fluctuations—the value of these predictions can quickly diminish. By meddling with the output, we introduce more uncertainty into the process rather than eliminating it.

This tendency to over-manage or second-guess the system is not unique to the world of demand forecasting. In fact, it is a widespread issue in supply chain planning. We often believe that by taking "more control", we can exert better influence over outcomes, but in reality, our attempts to fine-tune the model may lead us to a worse place. This is especially true when dealing with complex and dynamic systems like supply chains, where the number of variables is so vast that even the smallest change can cause unforeseen ripple effects.

The root cause of this issue lies in how we interact with the data. The tools we use are designed to process large volumes of data and make sense of the inherent complexities. They are not simply systems that display numbers, but sophisticated models that are capable of predicting future outcomes based on empirical evidence and the relationships between variables. When we adjust these predictions without truly understanding the underlying model, we run the risk of overcomplicating matters. The tool, in its simplicity, may have been pointing us in the right direction, but through our intervention, we introduce bias, noise, or irrelevant data that can distort the outcome.

An example of this can be seen in the practice of safety stock calculations. Safety stock is intended to buffer against variability. Yet, when planners start adjusting these levels based on personal experience or anecdotal evidence, the result is often an overly conservative or excessively risky approach to inventory. In both cases, the system becomes less efficient, and the underlying principle of using data to guide decisions is undermined. Instead of letting the model tell us what it has learned from historical data, we assume that our own insight into the business environment is superior, and we alter the results accordingly. In doing so, we obscure the signal in the data and create additional uncertainty.

We must ask ourselves: when we intervene in these systems, what value are we truly adding? Are we genuinely improving the model, or are we simply introducing an extra layer of complexity and delay? It’s crucial to recognise the value of the tools we have at our disposal and allow them to do what they were designed to do. By over-intervening, we may be inadvertently causing more harm than good.

This problem is compounded by the fact that many supply chain tools are designed with a level of sophistication that is often underappreciated. These tools rely on statistical methods, machine learning, AI and optimisation algorithms, which, when left to operate as intended, can identify trends, relationships, and patterns that we may not immediately recognise. Human intuition, while valuable, is often based on limited knowledge or past experiences, which might not align with the complex dynamics of today’s global supply chains. The temptation to modify the data in ways we perceive as beneficial is understandable, but in many cases, it’s more effective to trust the algorithm and focus on improving the quality of the data fed into the system, rather than tweaking the outputs it generates.

A pragmatic approach would be to view these tools not as black boxes that need to be constantly scrutinised and adjusted, but as guides that help us to make better decisions. We should focus on their strengths—processing large datasets, identifying trends, and making accurate predictions based on the information available—while resisting the urge to override their insights with our own subjective opinions. This is not to say that human expertise is irrelevant; rather, it’s about knowing when to trust the machine and when to apply human judgment, and understanding that in many cases, less intervention is more.

As we look towards the future of supply chain planning, the lesson is clear: we must be more mindful of the "value we add" when we interface with the data. Our role is not to constantly adjust or meddle with the predictions and recommendations made by the tools, but to ensure that the data we provide to them is of the highest quality, and that we understand the context in which those tools are operating. By respecting the capabilities of the tools and recognising the limits of our own influence, we can create more reliable and efficient supply chain plans that are based on the best possible evidence available.

The best way forward may not be to strive for greater control, but rather to embrace the power of the tools and allow them to do what they do best. Our challenge is to stay humble, recognise the sophistication of the models we use, and allow them the space to operate effectively. In doing so, we can create supply chains that are more resilient, more efficient, and ultimately more successful.