With more than 90% of executives acknowledging that forecasting needs to look different in 2021, it’s clear that the days of relying on weighted averages and other historic data to identify patterns are long gone. Demand sensing improves accuracy, but producing reliable forecasts requires the right type and quantity of data. As company leaders set out to better use forecasting to optimize their product mix, they should follow these steps.
Demand planning isn’t easy, particularly when you’re relying on last year’s static spreadsheet numbers and applying them to this year’s SKUs. According to a 2020 survey from McKinsey & Co., 91% of the C-suite wants forecasting to be more capable in 2021. Whether a product is out of stock or overstocked, it’s bad news. The former costs $634 billion in lost sales per year, while the markdowns necessary to move excess product end up amounting to $472 billion annually. Companies run into these issues when they rely solely on time-series methodology, which uses prior sales history to create a forecast.
Time-series forecasting is often good enough to predict mid- and long-term demands, but it doesn’t allow sufficient accuracy for short-term planning. Throw in rapid demand fluctuations caused by unforeseen circumstances such as the pandemic, and your organization might completely miscalculate what’s on the horizon.
Deepen Insights With Demand Sensing
Most demand planning theories rely on weighted averages and other historic data to identify patterns. They’re comfortable and accessible methods, but they lack critical pieces of raw data — whether it’s online or external — that can paint a more complete picture of real-time demand and the factors that influence it. Demand sensing is an attempt to take those other factors into account, and it’s based on the premise that the most recent developments in demand can inform the near future.
Demand sensing improves accuracy because it’s sensitive to sudden, immediate, and real-time demand fluctuations that wouldn’t be on the radar of a traditional forecast model. It can also help companies understand whether a planned event that positively impacts one SKU negatively impacts demand for another higher-margin SKU and results in a net loss of revenue. Picking up on these nuances is not easy, particularly if your company has hundreds of thousands of products across multiple locations, sales channels, and distribution systems.
Most companies focus their attention on top-selling items, while the rest get lost in the shuffle, but there’s a major opportunity in the middle layer, where optimizing the product management strategy throughout the year can improve your company’s bottom line considerably. According to a study by Retail Systems Research, around 65% of retailers run into stockout issues repeatedly with their most popular products and categories. Another 63% contend with the opposite problem: excess inventory in slower-moving categories.
Depend on the Right Data
Producing reliable forecasts with demand sensing requires the right type and quantity of data. For example, you need sufficient historical data to strengthen confidence levels for seasonal or cyclical trends. You also need to track demand based on end-client requirements instead of internal company orders, which may have been inaccurate. It’s also important not to confuse correlation with causation. External events might appear to influence demand but have no actual bearing on it. Making these assumptions will skew forecasts considerably down the road.
Fortunately, access to data and analysis can result in remarkably accurate forecasts. As you set out to better use forecasting to optimize your product mix, follow these steps:
- Optimize decisions around inventory flow. Historic data can’t tell you everything you need to know about future demand fluctuations. Instead, accurate forecasts will come from AI-driven demand sensing algorithms that can learn from other items and stores.
Use these tools to exact detailed information from parallel marketing events, and you’ll have a wealth of information about how a new product or service will react to a variation in price, demand trends, and more. Pull transactional data from your CRM, internal SC systems data (e.g., point-of-sale and inventory data), and data on demand events regressors (e.g., holidays, promotions, store schedules, etc.) and sales/margin price changes. For example, when ThroughPut worked with a Fortune 500 paint and coating company, we used demand sensing to:
# Connect historical sales data
# Segment sellable products based on grouping, geolocation, and product mix
# Load and test additional regressors against demand
# Extract existing cyclical/seasonal trends
# Expose demand correlations
# Generate forecast and operations plans to match market demand with high service levels while keeping the operations expenses under control
With these steps, the company optimized decisions around inventory flow between manufacturing, warehousing, and retail locations and saw superior results.
- Conduct an accurate flow analysis. Modern management tools like an AI-driven demand sensing solution result in more accurate, automated pull-based replenishment, creating a highly profitable push-based replenishment. These algorithms are constantly learning and improving, and they can offer real-time views into sudden demand fluctuations. Accurate flow analysis means:
# Offering customer-centric demand management
# Managing demand variability for new product introductions
# Optimizing forecast accuracy and hitting revenue goals
# Limiting stockouts and excess inventory
# Leveraging internal and external factors for ideal stock levels based on events, trends, patterns, and seasonal fluctuations
# Anticipating accurate orders
# Planning better for demand-driven replenishment
For example, Europe’s largest retail food chain recently relied on help from our team to implement a data-driven review of fulfillment and shipping practices and optimize replenishment and allocation down to individual SKUs. By conducting an accurate flow analysis between warehouses and improving the shipment handling and transportation processes, the company saw $20 million in annual savings, cutting trucking costs by 10% and increasing top-line revenue by 1%.
- Monitor real-time fluctuations in purchasing behavior. Demand sensing isn’t a standalone forecasting method, but it’s a highly effective way to capture real-time fluctuations in purchasing behavior that inform and improve existing predictions. These solutions extract daily data from point-of-sale systems, warehouses, and other external sources to highlight increases or decreases in sales and evaluate the significance of each fluctuation and divergence.
REI, for example, uses predictive rules and demand sensing to retrieve the necessary amount of in-store and warehouse inventory to meet its customers’ needs based on purchasing behavior and seasonal variables.
The complex and intertwined network of modern supply chains makes it impossible to rely on a single source of truth across upstream and downstream operations, but a demand planning strategy that combines available data with advanced technologies such as AI can offer organizations a competitive advantage. A holistic, intelligent demand sensing solution will help companies achieve a sound balance across the shop-floor and top-floor operations, delivering the right products in the right numbers to meet ever changing customer demand.
Commentary by Ali Hasan Raza. Here’s what you’ve missed?
Most Powerful Countries.
Most Influential Countries.
Most Forward-Thinking Countries.
Most Entrepreneurial Countries.
Best Countries For Cultural Heritage Influence.