Traditional demand planning
The main objective of demand planning is to help businesses prepare to meet future demand. The forecasts are largely based on historical, seasonal demand patterns, not current demand signals. In fact, the main driver for future forecasting is historic sales during the same time period in prior years. The inaccuracy of demand forecasts that are based on historical data alone has often resulted in gaps between past and present situations, which, in turn, create significant business challenges. For example, seasonal supply needs based on the events of previous years and on marketing trends can lead to overstocks and increased inventory costs, stock-outs and missed sales opportunities.
Gaps in traditional demand planning solutions
The absence of a structured way to refine the operational demand plan using consumption data is typically responsible for more than half of all near-term forecast errors. E2open’s machine learning (ML) enabled Demand Sensing application uses daily operational data such as sell-out, sell-through, on-hand inventory, open orders and planned shipments, plus market and external events information like time of year, weather and consumer sentiment. The result is a significant improvement in accuracy compared to approaches that rely on historic sales, with E2open Demand Sensing cutting forecasting error by more than one third. E2open Demand Planning can also leverage ML to reduce long-term error by more than 10%. Combined, it provides companies with the best possible forecast at any horizon, leveraging the most granular data to better predict executional, tactical and strategic demand.
Challenges and availability of consumption data
While consumption data, such as point of sale (POS) from retailers or distributors, has been available for decades, it is typically segregated across multiple companies and systems and not been normalized to make it available and easily consumable by demand planning solutions. E2open’s advanced analytics within Demand Signal Management and Channel Data Management applications remove this barrier by cleansing and normalizing consumption data to make it decision-grade and ready to be used to better predict demand.
The role and benefits of E2open’s advancements in AI and ML
E2open’s advancement in artificial intelligence (AI) and ML algorithms is a key enabler in the use of consumption data for next-generation demand prediction. E2open Demand Sensing uses supervised and unsupervised AI algorithms to systematically process orders, shipments, POS data and other signals such as weather, economic indicators and social media to improve short-term demand plans. These systems have been in global production for years with some of the world’s largest companies.
One of the questions asked in the 2019 E2open Forecasting and Inventory Benchmark Study is whether real-time data and ML can actually improve planning performance. The answer is a resounding yes, with the data illustrating the following benefits for ML-enabled short-term demand forecasting:
- Reduce demand forecasting error over time by 36%, with a forecast value-added of 2x over the traditional demand planning solutions
- Reduce extreme error by 55% compared to traditional demand planning solutions
- An average of 34% forecast error reduction in top movers, 30% forecast error reduction in new products and 17% average reduction in daily safety stock
Additional benefits—not mentioned in the study—that can be gained by implementing an ML-enabled, short-term demand forecasting solution:
- AI/ML provides operational efficiency through automation to eliminate manual planning intervention and focus on management by exceptions and data analytics
- AL/ML provides rapid return on investment (ROI) and is self-funded through operational efficiencies—including fewer transshipments and costly expedites—plus revenue growth from better service and more productive inventory
Predicting demand has always been a challenge, yet the pursuit of forecast excellence is strategic because it lays the foundation for performance improvements across the entire organization, including internal operations as well as those from external ecosystem partners. Reliance on historical data is failing to deliver the forecast accuracy manufacturers are looking for. Modern planning involves the continual process of filling the gaps between operational data and consumption data to better align supply and demand. The use of real-time data and AI/ML in planning at scale increases the agility to respond to shifting consumer demand, decreasing lost sales due to stock-outs, preventing losses from overstocks and excess inventory and improving manufacturers’ overall customer service levels. AI-powered demand prediction platforms lead to more resilient, effective and optimized supply chain management, delivering better business outcomes by reducing costs and improving organizational operations in a streamlined way, thus driving up productivity and growth margins.