The evolution of demand forecasting techniques: From manual to smart-touch planning and forecasting

By Erik de Vos

When we look at the world of demand planning and forecasting, we see that over the last couple of years, some serious progress has been made in the available demand forecasting techniques. Nevertheless, it is safe to say there is still a lot of diversity in how companies have developed in this area. On paper, companies might prioritize enhancing their demand forecasting techniques in the long term, but the need for short-term success often takes center stage.  

We have access to more data, but strangely, it does not always translate into an improvement in forecast quality. There is this belief that we can manually extract all insights from the data, use fewer people, and still make the right impact on a forecast. Enter advanced tools with significant computing power, which, instead of revolutionizing forecasting, end up being used as storage spaces for manual forecasts. And, of course, we witness highly skilled planners spending their days inputting data manually. 

In this blog, we will explore the evolution from traditional manual forecasting to smart-touch planning and forecasting, evaluating the ideal balance between human expertise and machine intelligence to optimize forecast accuracy and efficiency.

No-touch planning & forecasting 

Now, with the numerous issues surrounding how to organize demand planning and forecasting, we hear terms like ‘no-touch planning and forecasting’ making their grand entrance. The idea is simple – when faced with problems arising from human involvement, why not automate everything to solve them? Let’s eliminate human bias, assuming humans will always make things worse, and machines will always make things better. The promise: more efficient planning, better forecasts for all situations, and a reduction in costs. 

But let’s pause these assumptions for a moment. Do humans always worsen the demand plan and forecast?  

At EyeOn, we believe that humans bring value to the forecast, especially where machines lack data. For example, in situations where new technology is being launched, or the entire world is shaken up due to a global disruption like COVID. Remember, exceptions will always exist, and the past does not guarantee the future. 

Similarly, can a machine always nail accurate forecasts? What is the one thing a machine needs for its job? Data. Often, the data is not available, not timely, full of mistakes, not centralized, disorganized, inaccessible, or people just do not want to put in the effort to clean it up. How can a machine work without valuable data? Finding the right balance between man and machine is key in transforming demand planning and forecasting.

 

human bias in demand forecasting techniques

 

Alternative demand forecasting techniques to no-touch planning 

We all agree on one thing: manual demand forecasting techniques are relics of the past. Typing in numbers and endlessly reworking them isn’t the way forward. It’s high time we acknowledged that continuously meddling with the forecast, unfortunately, tends to make it worse.  

But here’s the thing – fully automated, no-touch forecasting isn’t the silver bullet either. 

So, where does that leave us? Enter smart-touch planning and forecasting as the optimal vision. It’s an approach where we combine the strengths of the machine with human intelligence. 

Imagine a world where we automate the obvious, recommend the probable, and identify where human intervention is needed. In this approach, automation handles straightforward tasks, and for the less obvious ones, humans can get probable scenarios to allow faster decision-making in more fact-based processes. The planner still enriches the forecast, but strategically, where real value is added. The focus in the typical S&OP process will shift from compiling the forecast to deciding which forecast scenario will yield the best results.

 

how smart touch planning combines different demand forecasting techniques

 

Introducing smart-touch forecasting 

Let’s look at the core definition of smart-touch planning and forecasting first: any planning set-up that balances human expertise with the available machine intelligence. 

The available machine intelligence, what does this mean? When applying a forecasting technique, you will be bound by the technology you have available. Nevertheless, you can still make smart use of what you have. For example, it is smart to apply segmentation when you work with a statistical engine. It is smart to organize for value adding cleaning when using statistics. It is smart to ensure there is structured data on relevant demand drivers when you work with machine learning. When you do not organize yourself to get the best out of the machine first it will be difficult to claim you have a smart-touch set-up.  

Human expertise is the second part of the equation. Any demand forecasting technique will always encounter situations where you cannot leave it up to the machine to do the reasoning. However, human expertise needs some assistance to become value adding. In a smart planning set-up, you do not leave it up to chance, but you install methods that help the human. Help them to spot where intervention is needed, because the machine struggles there. Help them to reason which intervention is needed, by recommending potential interventions.  

Smart-touch forecasting in practice 

In our definition we added 3 core principles on top: automate the obvious, recommend the probable and flag where human intervention is required. Let’s use the example of promotions to make this more tangible. 

  1. Automating the obvious. Businesses that are in a promotional context will recognize that promotions will be often repeated. Even when the underlying parameters might change a bit, it is not so hard to find repetition in the promotional data. In a smart-touch set-up for promotions, we would automate the forecasting that relates to the obvious promotions. 
  2. Recommending the probable: you will always see promotional plans that have not been tried before. Potentially these do contain elements that are recognizable, but it is less obvious to fully rely on automation here. A machine can help by recommending the probable. It finds the best fit and can recommend a basis for forecasting, so you do not have to start from scratch.
  3. Flag where intervention is needed. Those who have worked on promotions before will recognize it can be a challenge to plan in time and not forget something. For example, what if we always have promotions in a specific period of the year but the plan for this year is completely different. A smart-touch set-up would organize to detect these deviations and bring the insight in an actionable way to the planner.

Looking at a situation with only promotions is one thing. Envision a set-up with multiple relevant demand drivers where the machine handles 80% of the effort and the other 20% requires human insight. In this scenario, we feed the machine with sales history, factual data (sell-out data, order book, contract book, customer inventory positions), plan data (promotions, portfolio changes, price changes, marketing plans), and market indicators (growth outlooks, market share evolution, raw material price outlooks).  

Trained to understand relationships, the machine synthesizes the information into one forecast considering all impacts. Of course, this setup calls for machine learning, going beyond the typical statistical engine. It requires a forecast code tailored to your circumstances and skilled individuals to fine-tune the code for optimal results. To make it smart-touch, it also involves supporting planners to intervene where they bring value. Providing insights, directions, options, and recommendations ensures that enriching the forecast adds value and is, therefore, smart.

 

comparison of different forecasting techniques

 

In summary, smart-touch planning and forecasting transcends mere forecast algorithms. It means implementing planning processes that derive decisions from forecasts considering all available and relevant data. It involves reshaping roles and responsibilities across the organization, with planners focusing on key forecast decisions and business functions providing valuable information for the machine. It necessitates structuring data in terms of flows, standards, and processes. Finally, it entails adapting planning tools, enabling humans to work efficiently with machine outputs and contribute value where possible. This is the essence of smart-touch planning and forecasting.

Ready to elevate your demand forecasting techniques and unlock the full potential of demand planning?

Despite some huge steps we have taken in demand planning, many planning teams remain stuck in the labor-intensive full-touch forecasting phase. It’s time demand planning teams unlock the true value of demand planning through smart-touch forecasting. 

That’s why we’ve created the smart-touch forecasting roadmap. 

In this roadmap, we’ll dig deeper into the hype surrounding complete forecast automation, advocating the crucial role of human expertise when applying statistical and machine-learning models. And we’ll provide you with practical steps to accelerate progress in your forecasting maturity level. Get your copy here.

future proof demand forecasting technique: smart touch forecasting

Search for