Supply chains are the backbone of the global economy. Weak economies require efficient supply chains to prosper on thin margins; whereas robust economies require healthy supply chains to support growth. Accurately predicting demand is one of the largest factors influencing supply chain performance because it affects decisions that drive between 80 to 90 percent of the costs in the supply chain, including procurement, manufacturing, distribution and inventory. Billions of dollars have been invested in upgrading communication and computing infrastructure with the latest technology, yet most of the industry still relies on planning mathematics developed decades ago.
For years, the industry achieved incremental gains in forecast accuracy through the use of increasingly complex statistical models. While these improvements might have been perceived as “good enough” in the past, this is no longer the case.
Today we live in a fast-paced world where manufacturers and retailers face risks stemming from uncertain economic, climate and political pressures. It is a connected world where shopper preferences are increasingly shaped by mobile technology and social media. The time has come to rethink supply chain performance and challenge decades-old assumptions that unnecessarily restrict supply chain efficiencies and constrain the economy. Let’s look at some assumptions we’ve long assumed to be correct.
People quickly learn that revenue depends on product availability, so the tendency to raise safety stock levels is a natural reaction to market volatility. Yet this is at odds with shareholder pressure to free cash flow and improve return on working capital by reducing stocks. The answer is to ensure availability without increasing inventory, or better yet, while reducing inventory.
Manufacturers are using new mathematics designed for volatile markets to analyze current supply chain data, gain visibility into shifts in demand and quickly respond to new market expectations. For example, during the H1N1 “swine flu” pandemic several years ago, one tissue manufacturer detected a corresponding lift in demand as the virus spread across the country and raised production to ensure on-shelf availability in affected markets at a time when competitors stocked-out.
Whether shifting schedules for underperforming items or stepping up production to capture a lift in sales, an accurate view of future demand provides the means to manage optimal inventory levels to balance business needs.
Today’s volatility has rendered past shipments an increasingly poor predictor of future demand. A recent study encompassing roughly one third of all North American consumer packaged goods traffic found that forecast error has increased since 2009, with weekly error rates now at 53%. In response, leading manufacturers are turning from statistical time-series systems built on the assumption that history will repeat, to pattern recognition systems based on the certainty that data is dynamic.
To sense changes in demand patterns, these systems require much more data. Fortunately, there is no shortage of current data in the supply chain. What have been missing are the applications to systematically use this data to improve supply chain efficiencies and achieve better business outcomes.
By sensing demand, manufacturers can now predict future sales in tune with rapidly changing market dynamics, giving them the insight to plan and build what they know will sell instead of what they hope will sell. Findings from the above study show that companies that sense demand achieved a 40% reduction in forecast error.
Initially, demand sensing technologies were focused on improving near-term operational planning and execution, within 4-6 weeks. The same pattern recognition techniques are also available for long-range planning across multiple months or years, again cutting forecast error by up to 40%. As an automated process, this frees planners from time-intensive tasks such as individually fitting statistical models for each item/location. By providing a better starting point, planners can now focus on discontinuities—areas that require knowledge outside of forecasting systems, like promotions or new products.
Automating demand prediction also helps balance inherent tensions within the S&OP process. Demand sensing forecasts are free from conflicting incentives and temptations to game the system by artificially inflating or depressing forecasts. Companies with mature, automated planning processes use demand sensing forecasts as an impartial view of demand to gain consensus and ensure that resources are most effectively used.
Traditional collaboration efforts have good intentions but tend to be time-consuming, requiring significant commitments from both manufacturers and retailers. We see the automated use of retailer data as the future of collaboration. Most large retailers already make their data available and expect to receive tangible value in return; it is up to the manufacturers to use this data in ways that provide mutual financial benefit.
An IndustryWeek article from April 2012 highlights that Procter & Gamble and Kimberly-Clark use downstream data to improve forecast accuracy with compelling business benefits. Retailer data, including warehouse withdrawals and point-of-sale, is processed by multi-enterprise demand sensing to create better forecasts that reflect current supply chain realities flowing all the way to the store. These updated forecasts are published daily to planning systems for execution. The result is fully-automated and scalable collaboration that improves stock availability, makes promotions more effective and lowers cost to serve.
Investments in accurate demand prediction yield strategic planning and execution advantages, allowing companies to consistently outperform their peers. As demand sensing becomes a standard business practice, it supports the global economy in good times and bad. Until such time, those who take the lead will have a clear competitive advantage.
Robert F. Byrne is president and CEO of Terra Technology, a provider of supply chain solutions for consumer products companies.