Sometimes, change happens when nobody's looking.
Broad industry initiatives to promote concepts like the "demand-driven supply chain," for example, may generate a lot of heat and light at first, only to be dismissed as largely hot air after the initial hoopla dies down. Then, quietly, without fanfare, companies really do start to change the ways they operate.
While no one is attempting to achieve the "consultant's dream" of one unit produced at the factory for every item scanned at the store, a significant number of companies up and down the food and CPG supply chain, from growers and processors to retailers, are making healthy progress toward aligning their production and distribution activities more closely with demand.
This shift may reflect a sea change in thinking across the industry, but it is probably as much the product of another trend: the easy availability of high-powered software that makes it possible for companies of all sizes to predict, model and shape demand in a more accurate, detailed and timely fashion than was feasible for even the largest, wealthiest corporations a few years ago.
Sophisticated demand planning and forecasting engines, whether offered as standalone software packages, or modules within larger supply chain management or enterprise suites, are helping many companies today dramatically change the ways they look at and respond to demand.
For starters, these software tools enable organizations to dramatically reduce forecast error, leading to improved performance in a variety of key measures, from inventory investment and order fill rates to transportation costs-thanks to fewer expedited deliveries, for example.
But beyond the immediate, obvious benefits that flow from dramatically improved abilities to anticipate demand, down to the SKU, customer or even specific retail location level, companies are discovering that the flexible tracking and analytic capabilities embedded in these planning programs provide a host of capabilities to help managers better understand their business and the marketplace-allowing them to strategically direct the enterprise with greater intelligence than ever before.
More than just a crystal ball, good demand planning systems can serve as all kinds of useful instruments: barometers, microscopes, early warning detectors. They help companies become aware of changing trends in the marketplace, in real time.
For some companies, notes John Bermudez of Oracle, Redwood Shores, CA, which offers the Demantra demand planner, the systems are doing double duty as "a very streamlined and real-time version of sales and operations planning. They're using it to capture updated sales forecasts from each salesperson, so that they're automatically aligning their sales budgets with the demand plan," he observes.
"We've added capabilities in Demantra to do very accurate promotion planning based on trade funds provided, so that salespeople can actually use the system to track their trade spend against a forecast in real time. Then they can make adjustments sooner rather than later if, say, a promotion goes better than expected, so they won't end up overrunning trade fund budgets and negatively impact profitability."
Growing use of demand forecasting systems is sparking greater adoption by many companies of other, more conventional network optimization tools, Bermudez adds, such as inventory optimizers, expert scheduling systems and others.
"Companies are realizing that if you start with a much better forecast that is far more granular and specific to various products, geographies, and customers, then all your planning operations are going to get better," he points out.
John Shaw, IT director at Litehouse Foods, Sandpoint, ID, says that demand planning was the initial tool implemented in its supply chain management suite. Its success created the foundation for integrating two additional downstream systems, advanced planning and scheduling, to manage its production operations and inventory planning and replenishment, to manage transfers of inventory between its two factories and two warehouses. In aggregate, these supply chain management tools have had a powerful impact on the manufacturer's bottom line.
In addition to providing more accurate input to systems used for planning and managing daily operations, demand planning and forecasting systems are proving highly effective for long-term strategic planning and business intelligence.
At Yuba City, CA-based Sunsweet Growers, for example, use of Zemeter Demand Forecasting from Supply Chain Consultants, Newark, DE, in conjunction with other tools, has enabled the growers cooperative to strategically channel the fruit from its prune harvest each year to the most profitable mosaic of potential product applications, based on forecasts that are updated throughout the year as sales trends develop.
"We use our demand planning process to do a lot of our strategic planning as it relates to product utilization and where sales will go," explains Harold Upton, vice president, strategic business processes at Sunsweet Growers. He says the co-op's beefed up planning process, based around the demand forecasting tool, enables the organization to manage virtually every aspect of its operations better, from working with growers to managing production lines, down to hitting specific targets for limiting overtime in its plants.
"It's hard to pinpoint exactly how much Zemeter itself has contributed, but over the last five years our return to growers-the ultimate measure of our success-has doubled. In the last two years we've set back-to-back record highs in terms of profitability to our members. Installing the system alone didn't cause all this, but clearly one result of using it is that people are making better decisions and that would have been very hard if not impossible without this kind of strategic planning tool," Upton comments.
Demand planning systems come in a variety of flavors, but combine the same basic elements, notes Sujit Singh, Supply Chain Consultants' chief operating officer.
"There is a statistical engine running historical data to create forecasts; a collaboration engine to take input from the people that are talking to customers; and a demand shaping engine, to help companies model and shape demand through promotions."
"All the programs attempt to merge market information, sales history and collaboration with sales and marketing and through a variety of algorithms, come up with the best single number to drive downstream planning processes," adds Chris Taunton, director of product management and supply chain planning for CDC Software, Atlanta.
In terms of reducing forecast error, their greatest impact tends to be on lower level, or SKU-level forecasting, Singh comments.
"Most companies have a good handle on higher level forecasting. If they sell 500 SKUs divided among 10 product families, they can tell you what their forecast will be at the product family level with a fairly high degree of accuracy, probably within 90-95 percent.
"However, from a production planning and procurement planning perspective, those high level numbers don't mean much. What is really critical for operations is to get forecast numbers at a very detailed level. That's where we provide the most help to companies, in taking their accurate, high level forecasts and extending that accuracy down to the SKU.
"At this highly detailed, SKU level, we've seen many users achieve from 30 to 50 percent improvements in forecast accuracy," Singh says.
Most programs allow users to bring in a wide variety of factors as inputs to forecast calculations. In addition to sales history, which provides the bedrock for all projections, data that may be factored in range from price history and weather predictions or patterns, to syndicated and other POS data, warehouse shipments, competitive activity and "any other causal factor a company might want to incorporate because it has good reason to believe that factor to be leading in nature," Singh comments.
Promotional calendars, of course and promotional histories are also key inputs.
Despite the ability most programs afford users to input their own lift factors tied to specific causal events and to adjust forecasts calculated by the systems, the programs are meant to predict SKU-level demand based largely on sales history with a high degree of autonomy-that is, without user interference.
As they sift through sales histories, the programs' algorithmic engines are designed to recognize patterns and apply different formulas to the calculations based on the patterns detected. This includes making determinations of which patterns are statistically significant and which are not, Singh points out.
"When you're planning to use a system such as Zemeter, one of the major advantages is for you to let the software do the heavy lifting. Instead of reviewing and adjusting each item, people get the most benefit if they let the program handle the bulk of the forecasting, while users concentrate their time on dealing with exceptions."
One beauty of today's highly intelligent forecasting systems is that they incorporate built-in processes for continuous improvement, adds Karin Bursa, vice president, marketing for Logility Inc., Atlanta, which includes a comprehensive demand planning system within its Voyager supply chain suite.
"The systems actually get smarter over time, because as the data on real demand vs. forecast feeds back into the systems, the data they're working with and the way they apply various formulas keeps getting better, enabling management to keep making better, more informed decisions."
By the same token, notes Antonio Boccalandro of Dallas-based i2 Technologies Inc., the systems provide extensive audit capabilities, enabling users to document and annotate any changes made to the forecast, so these can be reviewed later and judged as to whether they were accurate or not, thus helping the human elements in the equation to continually improve their forecasting performance as well.
A large part of their value as a strategic tool lies in demand planning system's abilities to simulate scenarios and perform what-if projections, notes Taunton of CDC and others.
"The ability to model demand by distribution channel and product class, as well as by particular items, combined with capabilities to model seasonality and promotional activity, allows companies to become very proactive in how they can influence and shape demand in different markets," says Bursa.
"One thing that's interesting with these tools for scenario analysis, is that companies are starting to look at things more holistically, at the supply chain as a whole," Boccalandro of i2. "So from projecting an increase of 10 percent in sales, for example, a company can examine in a very detailed way just how such an increase would affection their production or logistics capability, and so identify potential challenges or bottlenecks before they occur. They might determine that certain vendors might not be able to supply necessary materials, or a packaging line might not have enough capacity. Then they can take steps to accommodate these potential shortfalls, before they occur."
"By using a good demand planning system, companies can greatly increase their forecast accuracy, but that's not the end point," adds Singh. "For every business, there's a triangle of considerations: increasing customer satisfaction, decreasing costs, increasing throughput. If you think about them, there is an inherent conflict among these three areas, and the trick is finding the most successful balance. A good demand planner helps companies achieve this balance, so they can improve customer service, increase asset utilization by up to two-three percent and reduce overall inventory, all at the same time, and not only that, but do it in a sustained way, month after month."
Demand Forecasting Reaches The Supermarket Shelves
Most demand planning systems are aimed at manufacturers and wholesalers. But sophisticated, automated demand forecasting systems have even reached the store level. Plano, TX-based Retalix, which provides food retail software systems, for example, introduced DemandAnalytX, which allows retailers to forecast daily demand for each SKU based on the same kinds of factors and calculations incorporated in the demand planning systems used by their suppliers upstream, including sales history, seasonality, pricing, promotions and special events.
The program actually combines forecasting with store-level order development, notes Gil Roth, executive vice president, supply chain development.
"It calculates, based on the forecast, variability in the forecast, and other factors such as shelf life, item cost, interest rate and more, how much product the store should have on hand on any day and how much to order, based on a perpetual inventory, to maintain that level."
For many retailers, the biggest challenge in implementing the system is developing the discipline of maintaining a perpetual inventory at the store.
For the software vendor, applying automated forecasting to the store shelf involved other challenges.
"At the retail level, there is a much higher level of statistical noise compared to the warehouse level, where product is ordered in large quantities and demand patterns are relatively smooth," Roth explains. "If you look at demand for different items at the store level, you will find huge deviations. You may have slow movers that are sold only once or twice a week, or a handful of customers may be responsible for a significant percentage of demand for a particular item."
To deal with this greater level of uncertainty, Retalix developed its own algorithmic tools to allow the system to better differentiate between real patterns and deviations that have no predictive value.
Among the benefits retailers realize from the system, Roth adds, are three that are easy to quantify.
"First is reduction in out of stocks. Our customers typically experience anywhere between a 50-60 percent reduction in out of stock events. This typically translates in to a one-three percent increase in sales. Second is reductions in inventory. These average from 12-20 percent, but in some departments like tobacco are much larger, 25-30 percent."
A third benefit that stores almost don't notice is reduction in perishables lost to spoilage, Roth adds.
"Stores can expect to reduce what they throw out by around 25-50 percent," he says. -C.C.
Extend Forecasting Benefits With Inventory Optimization
Once they've enhanced short-term forecasting through the use of a sophisticated demand planning engine, a growing number of companies are seeking to extend the benefits by integrating a similarly advanced inventory optimization program.
"A major driver of safety stock is forecast error. So once companies have this very accurate forecast, their next question is, how do I monetize that accuracy and take down my safety stock in a structured way that reflects the changes in my forecast error?" comments Robert Byrne, CEO, Terra Technology, Norwalk, CT. "What you're really measuring when you set safety stock levels is variability, so demand forecasting and safety stock calculations are closely related.
"Historically," he adds, "most companies have set inventory targets based on rules of thumb-this is a high volume item, we need about two weeks inventory on hand; this is a low-volume item with lots of variability, better carry three weeks of inventory."
Terra's Multi-Enterprise Inventory Optimization system, introduced last year, systematizes these calculations by measuring forecast error related to lead time on a daily basis.
The system takes as inputs three basic kinds of uncertainty: demand uncertainty, manufacturing uncertainty and transportation, in addition to desired service level to customers and traditional lead times. Combined with historical item movement data, it automatically calculates optimal inventory levels for each SKU, which can be viewed over any time period and compared to actual inventories on hand and in the supply chain.
Similar to sophisticated demand planning systems, Terra's inventory optimizer is designed to deliver its results with minimal human interference, leaving human intelligence free to deal with exceptions.
"The program allows users to do some root cause analysis. If, for example, an item is flagged because last month the safety stock level was 100, and now it's 150, you can drill down into the data to understand what happened, whether it was due to forecast error, a change in lead time, or perhaps schedule compliance fell off a cliff," Byrne points out.
The system also allows users to explore what-if scenarios.
"You can pose questions such as, how would my inventory levels change if I changed my customer service level? Or if I cut my cycle time in half at a certain plant, how would that affect safety stock levels?"
One of the unusual features of Terra's inventory optimizer, Byrne notes, is that it takes into account not just inventory at the warehouse and plant level, but throughout the supply chain, including regional DCs, work-in-progress, and raw materials, in addition to warehouse and plants. -C.C.