Way back in 1943, mechanical engineer Richard James was working to devise springs that could keep sensitive ship equipment steady at sea. He accidentally knocked some of the spring samples off a shelf and watched in amazement as they gracefully “walked” down instead of falling. Along with his wife Betty, James came up with a machine to coil 80-feet of wire into a two-inch spiral and Betty came up with the name “Slinky.” Initially, the sales were sluggish. By December 1945, Mr. and Mrs. James made 400 Slinkys for Gimbels Department Store in Philadelphia. The story did a demonstration for Christmas sales, and all 400 Slinkys sold within minutes.
Neither the James husband and wife nor Gimbels had access to high-tech artificial intelligence forecasting applications and had no way of predicting the immediate popularity of the simple toy. The Slinky is a good reminder that forecasting has always been difficult. Historically, forecasting relied on significant trial and error. With so many factors being uncertain, demand forecasting is not easy. It is also an inexact science and is short on hard and fast rules. No one’s got a crystal ball.
Knowing how many people will shop next month, if one popular product will suddenly fall out of favor and an obscure one will become the it-item that everyone is clamoring to buy is almost impossible. However, e-retailers can use various parameters to make informed decisions backed by extensive data analysis. While risk and uncertainty are part of every business, researching and making the best effort to properly determine the future demand or sales prospects for specific products will lessen problems down the line.
Understanding the geographical and seasonal demands of inventory and where to allocate advertising expenditure is important when making warehouse and shipping decisions. Factors like climate, for example, can shift regional demands for a product, especially for apparel retailers.
Swimwear likely sells steadily year-round in California and Florida but gets a strong bump in early summer in other parts of the country. Although a fulfillment center in one of these geographies might meet the retailer’s needs most of the year, having a sole distribution center in the Northeast might actually be the better option to be closer to the aggregate customer base over the course of a full calendar year, thus reducing overall shipping expenditures. At the end of the day — unforeseen circumstances notwithstanding – the best measure for forecasting future demand is sales history.
Data collection is key to demand forecasting. All business decisions should have basis in some analysis of numbers, especially data gathered over the long term. Having at least a year of sales tracked can provide a solid foundation for predicting how an eCommerce site will behave in the future, even as it grows.
Keeping track of customer information — from demographics to behavioral data — capturing and analyzing shopper data will help refine a business model. The better understanding of the customer, the easier to decide if an urban fulfillment center is a good investment.
According to the McKinsey Global Institute, “Artificial Intelligence technologies could eliminate many levels of manual activities in areas such as promotions, assortments and supply chain. AI will enable retailers to increase both the number of customers and the average amount they spend by creating personal and convenient shopping experiences.”
Determining which type of forecasting to use depends on the size of the business and the goal of the forecast. The specific need of the business should guide formation of a custom Demand Forecasting model, which will likely evolve with the business.
Passive demand forecasting – which supplies limited information for conservative growth goals – may be adequate for small and local businesses. Companies competing for expansion, however, require a more active approach, which includes knowledge of the external economic environment and a potential updated marketing strategy. Internal business level demand forecasting of each division of the company is often important for large companies, while many businesses are interested in forecasting at the external macro level to strategize for expansion, customer segments and risk mitigation. Many companies carry out a short-term demand forecast throughout the year, but effective planning and budgeting also obliges a long-term forecast for more than twelve to 24 months in advance.
The assortment of products available for an online business creates a hierarchy. Consequently, calculating demand relationships becomes more complex than observing a simple correlation. E-commerce businesses can benefit from a dynamic model of demand forecasting. Such a model has proven superior to a linear approach that does not account for multiple correlating sales patterns.
Although it might seem counterintuitive, having a limited number of SKUs will likely help a business grow, and successful forecasting eliminates as much excess inventory as possible.
Having too many choices can be confusing for customers — and increase costs for order fulfillment. More SKU’s increase cost to pick orders and more inventory means more storage space required. Having excess inventory on hand is never good for the bottom line. Every item sitting on a shelf in a warehouse is money invested with no current return.
A general rule is that 80 percent of sales comes from just 20 percent of SKUs. The more products available, the more capital needed to maintain stock — especially if it is spread out over multiple warehouses. Efficient urban fulfillment centers rely on businesses with a limited number of SKUs. Think niche rather than supermarket.
Regardless of the means used to create accurate demand forecasting, few products reach the success that Richard James’ accidental creation from the 1940s has reached, going on to sell more than 250 million units by the turn of the next century.
Article written by Brendan Heegan
Featured in Retail Minded.