Today’s supply chains are challenged by product-line proliferation, shrinking product lifecycles, increased complexity, and massive amounts of data. To add to that list, customers are becoming more demanding and unpredictable; internet-fueled trends are changing on a dime, and promotions and sales are subject to weather, new-product introductions and the economy.
To be ready for what tomorrow might bring, companies must understand how changes in supply chains affect demand. Only then can they ensure that the products customers want are available at the right time.
Demand sensing, as the name implies, is essentially the art and science of picking up on short-term trends immediately, in order to better predict what consumers will want. Demand is impossible to predict perfectly because it’s affected by an infinite number of known and unknown variables. According to a study by KPMG and the Economist Intelligence Unit, only 22% of companies’ forecasts were accurate within five percentage points. On average, forecasts were off by 13%, causing a significant effect on share price.
There are lots of purists who stick to forecasting approaches that use weighted averages and other traditional methods that are easy to understand. This approach might feel comfortable and accessible, but it can only yield predictions based on historical demand patterns. It lacks crucial online and external data that can provide a near real-time view of demand and what influences it.
Today, there’s a better way. As with forecasting the weather, it’s possible to train intelligent demand-sensing systems to predict the seeming chaos of consumer demand with a high degree of accuracy. These hybrid demand-sensing systems apply multiple forecasting techniques and data types.
Hybrid demand sensing is more robust, reliable and sustainable than simple forecasting alternatives. It not only benefits from applying different planning approaches, but also gets stronger and more responsive as additional human intelligence and data sets are factored in.
A huge benefit of demand sensing is that it immediately incorporates short-term trends into the forecast. Instead of having to use the same forecast in a 60 or 90-day window, planners gain insights that empower them to continually fine-tune forecasts using the latest sales data. This ability to react faster and more frequently to demand changes leads to higher profits and service levels, and less waste.
Here are some examples of how demand sensing is helping companies turn insights into profitable actions:
TireHub, a distributor of replacement tires, incorporates local PoS data into its demand-sensing model, which is enhanced with machine learning. Twin challenges of SKU proliferation and seasonality had been making it hard for TireHub to optimally position its inventory. Today, it not only plans for predictable demand fluctuations (such as snow tire sales peaking in winter), but also the complexities of local demand, by factoring how many of the different car models are sold locally and regionally. Using this model, TireHub has been able to fully automate much of its replenishment planning, improve business outcomes, and build capacity to serve 70,000 points of delivery in just 18 months.
There are a number of ways to sense demand, and each new insight can speed reaction time and boost profits. The biggest returns from demand-sensing processes are in three areas:
Short-term forecasting with sell-in data. The easiest way for companies to start sensing demand is to use the most granular historical data available. This usually involves analyzing daily sell-in/ship-to demand data using short time horizons and adjusting the forecast accordingly. This type of demand-sensing factors in shipment history, which is readily available in most supply-chain planning or enterprise resource planning (ERP) systems. Some planning tools include short-term statistical forecasting, to improve forecasts’ responsiveness to ongoing demand changes.
Incorporate sell-out data. When sensing demand, it’s important to brainstorm all the possible, useful data sources that stand to improve the forecast. Downstream sell-out intelligence such as customer, PoS or channel data, for example, can help identify demand trends, give early warnings of problems, and close the gap between the plan and what’s actually happening in the supply chain.
Add external data and demand causals. Demand sensing can and should also integrate demand-correlated variables to create a robust forecast capable of responding to a wide range of future events, from the known to the unknowable. These include stock market fluctuations, competitors’ promotions, viral social media trends, new-product introductions, weather and other external factors.
Putting all three pieces together — the sell-in and sell-out data along with relevant demand casuals — provides the most complete, joined-up picture of demand possible. It also lays the foundation for highly automated demand sensing, which frees up planners to apply business knowledge to further improve forecasting and customer service.
Countless internal and external variables impact demand variability, and these will only increase in the huge period of change that lies ahead. The best tool for managing variability and guaranteeing high service is inventory. Demand sensing helps companies use inventory as optimally as possible. It extracts the important signals from the noise to sharpen forecast accuracy, improve short-term demand visibility and minimize inventory — all while improving service to customers. Companies that invest today in the tools, processes and skills to boost their demand-sensing capabilities can truly be ready for whatever tomorrow might bring.
Robert Kaufholz is director of solution design at ToolsGroup.
Enjoy curated articles directly to your inbox.