Need a solution that will improve your near-time forecast accuracy by at least 40%?
Experiencing an increase in forecast error due to changes in market scenarios?
Are your existing planning solutions not giving great forecasting results at lower granularity?
Many demand planners are asking these questions because of demand volatility – huge spikes, deep troughs and everything imaginable in between. In today’s VUCA (volatility, uncertainty, complexity, ambiguity) world, all the basic assumptions around customer buying behavior are being challenged every day. And this has prompted many planning organizations to go back to fundamentals and rewrite the process of creating demand forecasts.
Most traditional planning tools rely heavily on historical trend and use methods like time series by aggregating past data from factors like product family, geography, customer segment, price, etc. to generate demand forecasts. However, in many cases, historical sales data is insufficient for predicting future demand, as the true demand depends on many real-time external demand factors and supply chain events that cannot be addressed using traditional methods alone.
To generate accurate near-term demand forecasts, it’s imperative to start considering external demand drivers and real-time events data. This is possible with demand sensing. Demand sensing enables reliable and accurate forecasting by leveraging advanced ML algorithms to collect, curate and model external demand drivers and real-time events data.
Here are five key differences between demand forecasting and demand sensing.
Demand forecasting is focused on various planning activities to help planners manage production levels, labor requirements, capacity, inventory, etc.
Demand sensing enables customer-centric supply chains by predicting what the customer wants and when they’ll want it. The goal of demand sensing is to reduce the total time elapsed from occurrence of external demand signals to the planner’s ability to respond intelligently to these signals.
Demand forecasting uses historical sales data to produce mid-term to long-term demand forecasts.
Demand sensing enables planners to generate near-term and short-term forecasts. Near-term can even mean hours, days or weeks, depending upon the supply chain and availability of data.
Demand forecasting techniques use historical trend and seasonality data aggregated at the product, customer, geography and price segment levels.
Demand sensing taps into real-time event and external signal information like weather-related risks, geopolitical situations, regulatory changes, and demand drivers like demographic information, market indices, etc. POS data is one commonly used dataset in demand sensing.
Demand forecasting uses market research and statistical methods like time series modeling. These methods are best suited to capture historical trends, cycle and seasonality.
Demand sensing uses artificial intelligence and machine learning algorithms to capture short-term demand patterns. This enables demand sensing to consider both internal and external demand drivers.
Demand forecasting at higher aggregated levels makes it easier to achieve higher forecast accuracy. However, when the same techniques are used to generate forecasts at a lower granularity, the same level of accuracy is difficult to achieve.
Demand sensing is all about generating forecasts at a lower granularity, even down to the SKU level, and delivers better forecast accuracy.
Companies equipped with demand sensing capabilities are more prepared to respond quickly to changes in market conditions and improve customer satisfaction by ensuring product availability in the market. In the coming years, companies that aspire to increase their market share and position will need to embrace a connected, real-time supply chain enabled through demand sensing to capture all influencing factors and keep up with the pace of change.
Bristlecone’s Demand.ai solution helps planners understand real-time market signals, external and internal demand drivers contributing to change in demand.
Demand.ai augments existing planning tools by adding automated external driver ingestion, selection of significant drivers and custom development of machine learning algorithms.