Demystifying demand forecasting 

Demand forecasting anticipates future demand for a product or service, often making multiple predictions simultaneously. It plays a crucial role in business operations and supply chain management. In the former, it allows companies to make informed decisions about inventory, production and pricing. A precise demand forecast can help companies avoid overproduction, which can lead to excess inventory, but also helps avoid higher costs and stockouts, which can lead to unfulfilled demand and dissatisfied customers. This translates directly to supply chain management, as accurately predicting clients’ needs allows companies to plan their production and inventory levels to ensure that the right products are available at the right time. Demand forecasting is a cornerstone capability for integrated business planning with the potential to impact the entire value chain. For example, sharing demand forecasts with vendors and clients can improve coordination and collaboration, ultimately leading to better outcomes for all parties involved.  

Automating demand forecasting is increasingly important as businesses look for ways to improve the efficiency and accuracy of their supply chain; in this context, even improving small margins can lead to a substantial financial impact. An automated system assists in streamlining the time and resources for forecasting and permits quick analysis of large amounts of data, discovering patterns and trends which may not have otherwise been visible using manual methods. As a result of these implementations, daily forecasts are explicitly up-to-date, ensuring companies can make timely decisions.  

Automating demand forecasting at Metyis

At Metyis, product, inventory and pricing are central to the projects we undertake, which places demand forecasting as a pivotal component in our work for clients. Given the frequency at which we perform demand forecasting, it became imperative to hone and optimise the procedure, resulting in a decision to automate. This process is already underway, with the first step being to develop a codebase that utilises econometric, state-of-the-art demand forecasting models, allowing rapid installation and the ability to deliver demand forecasts to our clients more promptly. This codebase frees up a significant amount of time and resources, allowing greater focus on other tasks and improving efficiency, such as setting up deployment architecture and database connections.  

The codebase models are based on classic econometric theory, which entails making assumptions about the underlying data and the relationship between different variables. These assumptions allow us to make predictions about future demand using historical data. One of the critical features of these models is the ability to consider seasonality and external factors. Seasonality refers to seasonal patterns that may repeat over a specific period, such as sales going up during the holiday period. Another key feature is integrating external factors which might affect demand; these include economic indicators or discounts that can be incorporated into the models to produce more accurate predictions. 

We have partnered with a leading university in The Netherlands to ensure our econometric models are accurate and well-supported by a solid mathematical foundation. This collaboration also brings a high level of expertise and thorough analysis, resulting in reliable and robust forecasts. This partnership assures our position at the forefront of mathematical advancements within the field of demand forecasting and provides clients with the most evolved and accurate forecasts possible. 

Best of both worlds: from customisable models to tailor-made solutions 

The implemented models are highly customisable, allowing for adjusted parameters and coefficients to suit specific requirements and accommodate different industries or products. The purpose of automatisation is not to replace human expertise but rather to act as an aid; our experts will continue to interpret any such predictions before being shared with our clients. 

We have already implemented automated demand forecasting in a live project utilising the models in the codebase, providing real-world benefits and delivering a rapid solution to our clients whilst considering external factors and seasonality. Our demand forecasting experts further tailor the solution to meet the client's specific requirements, expedited by the codebase, saving both time and resources to deliver actionable insights and convey the impact of automating and streamlining the process.  

Next steps towards user integration  

Metyis has taken the first step in the automation process, providing a codebase to our data scientists and consolidating the range of our expertise and learnings into one place. The next phase of automating demand planning is the assimilation of forecasts into a user interface (UI), which will equip demand planners and business managers with the tools to interpret the forecasts quickly. By making the process more accessible, it can ease the change management process and allow for the straightforward integration of demand planning into daily undertakings, solidifying demand forecasting as a crucial aspect of business operations that can elicit further benefits from automation.