A commercial jet and a jar of salsa are inverted images of the supply chain: Millions of parts come together to make a relative handful of passenger planes, while only a few raw ingredients may be needed to produce millions of jars of salsa.
But just as fingers are pointed at production because of the long delays in delivering Boeing’s 787 Dreamliner, manufacturing takes the fall when not enough product is on the shelf or too much is in the warehouse. Poor demand forecasting is the real culprit, but production is the easy target.
“Manufacturing gets blamed for the variance, when in fact it is caused by sales,” grimaces Joe Bermudez, senior director-planning product strategy for Oracle Corp.’s Demantra unit. “Retailers tend to be brutal to manufacturers who shortchange them on product for a promotion, and manufacturing gets blamed.” When the job is to make the finished goods or source copackers to meet demand, there’s nowhere to hide when forecasts miss the mark.
Manufacturing has been whipsawed for so long by cranky forecasts, it seems like a normal state of affairs. The age of ERP was supposed to change matters, but that hasn’t happened. An independent study commissioned by the Grocery Manufacturers of America (GMA) concluded the error rate on monthly item level shipments in 2002 was 34%, 44% on a location level. Six years earlier, the error rate was 23%.
Food production scheduling and forecasting software has been widely used for more than 20 years, points out Olin Thompson, food industry strategist at St. Paul, MN-based Lawson Software Inc., but older programs were oriented toward the plant and were separate from companies’ business systems. In a marketing masterstroke, SAP bundled those plant systems with business programs to produce headquarters-oriented ERP. The problem, Thompson explains, was that the programming was oriented more toward business needs and failed to keep up with evolving needs on the production side.
“What SAP did was brilliant,” agrees Darryl Landvater, cofounder of RedPrairie Collaborative Flowcasting Group. “Creating a complete suite of integrated products was a very appealing message. The bad news is that the integration of the system was extremely complex.” A bigger failing, adds Oracle’s Bermudez, was that the programs couldn’t produce forecasts, only estimates. That shortcoming created the opportunity for a host of demand-forecast “point solutions” that are cobbled onto ERP systems.
The recession has made the supply chain an inviting target for inventory optimization. Unfortunately, distribution patterns are becoming more complex, making it more difficult to avoid starving the pipeline. Retail out-of-stocks are in the 5-8% range, experts say, and the ratio can balloon to 15% during promotions. Add to that the varying needs of retail and foodservice accounts and product returns for perishable items, and the food and beverage challenge becomes particularly daunting.
Access to store-level sales data could improve matters, and projects like a Kraft Foods-Sam’s Club collaboration that gives the manufacturer visibility to inventory levels in the stores and DCs help cut replenishment from weeks to days. But examples like this are few and far between. Working with Wegmans Food Markets and seven food and beverage companies, GMA and the Food Marketing Institute quantified the mutual benefits to manufacturers and retailers from “data synchronization” three years ago. A celebratory “synchronization summit” was staged, awards were passed out, and GMA has not uttered the “S” word since.
In fact, today’s sales don’t answer the more pertinent question of tomorrow’s volume, points out Lora Cecere, vice president at AMR Research Inc. Retailers do a poor job of forecasting, she adds, and when they involve their suppliers, they often hamstring them. Vendor-managed inventory is back in vogue, says Cecere, but grocers are apt to limit the supplier’s role to warehouse management while retaining the critical order-development responsibility.
Retailers can’t produce accurate forecasts because often they lack granular SKU data, seconds Jerry Shafir. His products are a case in point: Chelsea, MA-based Kettle Cuisine makes meals-to-go soups sold in supermarkets. More than 50 varieties arrive in bulk bags, and retailers place them in self-service bins. At the end of the day, the bins may be empty or half full, but in either case, they are sanitized and there is no record of which ones customers loved and which they ignored.
A former manager at Boston’s Legal Seafoods, Shafir founded Kettle Cuisine in 1986 and relies on slow cooking and fresh ingredients to produce his soups. Managing perishable raw materials is complex, he says, but it’s child’s play compared to demand forecasting. “You get a few cold, wet days, and demand changes dramatically,” says Shafir. “It’s disruptive.”
Syndicated scanner data provide useful insight into item movement, though it can be fairly dated and comes at a cost prohibitive to many companies. Welch Foods Inc. spends up to $10 million a year to review IRI and Nielsen numbers, and a glimpse at promotion results in 2008 can be less than illuminating when forecasting a 2009 promotion. It also can be overwhelming: “The problem with scan-based data is the same as it was 15 years ago: the sheer magnitude of it,” says Bill Harrison, president of Demand Solutions.
Multi-source data managementVisibility to inventory and movement through the supply chain usually requires multiple technologies. Both Demantra and TAKE Supply Chain (formerly ClearOrbit) claim Welch’s as a client, with both relationships tracing back five years. At that time, Welch’s had just scrapped a decade-old ERP system and deployed Oracle, former CIO Larry Rencken explained at Food Engineering’s 2004 Food Automation and Manufacturing Conference. Welch’s wanted to establish “very collaborative relations” with 30 key customers, rather than trying to be “all things to all people,” Rencken said. To do that, the right merchandise had to be in stores such as Sam’s Club and Wal-Mart at the right time, price and quantity, and that would require “supply chain excellence,” he said.
Demantra, which was independent of Oracle at the time, was charged with developing a forecasting model to meet those objectives. Forecasts based on inaccurate information on shipped and received product, inventory levels and finished goods location would be worthless, however, and TAKE supplied the bar-code scanner and other hardware and software to address that. “We empower the user on the floor and in the office to have confidence that the data are being validated,” says Solutions Architect Pat Anderson of TAKE. “Something as simple as data collection is crucial to the process.”
“Data integrity,” says Ray Gosselin, Welch’s chief information officer, “led to better accuracy, timeliness and visibility across our warehousing and distribution operations.” The automated order fulfillment and inventory system is in place at three of the cooperative’s four plants/distribution centers, though not at its copacker network.
Without accurate information on inventory levels, production scheduling and run rates, forecast programs are dangerous tools, reflects Lawson’s Thompson. But if the basic numbers are correct, “computing power and software engineering can deliver what we always knew was possible,” he says. In that event, “the data should go to tons of decision-makers throughout the organization, including guys on the line in the plant.”
In a previous life, Thompson headed Marcam Solutions, an ERP firm providing a plant-oriented system called Prism. The software was used in 130 Kraft plants to schedule raw material deliveries and deal with operational constraints in producing finished foods in the quantities and locations needed to meet sales forecasts. Tying the demand forecast to the production forecast wasn’t practical a decade ago, but programs like Lawson’s M3 Analytics now deliver that kind of business. “Taking the mountain of information and delivering it as something meaningful to all the organization’s decision-makers” is a major advance, he says.
The shortcomings of front-office ERP opened the door to point solutions for supply chain and warehouse management, according to Demand Solutions’ Harrison. These modules are strapped onto a company’s central ERP system to manage forecasting data and crunch the routine numbers while kicking out the exceptions that require human review. “You’re smarter than the software,” reasons Harrison. “We give you a number but also tools to evaluate it.” Many organizations develop their own models and crunch the number in a spreadsheet program, “but at some point, the house of cards falls in on you,” he adds. His forecast-management module incorporates moving averages, time series and a score of other forecasting tools, and then lets users select the model that best predicts demand, based on historical data.
A moving average drove forecasting at Brea, CA-based Ventura Foods until 2008, when management determined a point solution was necessary. Ventura’s manufacturing network sprawls across 12 sites plus a major sunflower oil refinery in Mankato, MN. It has a foot in private-label and branded retail, foodservice, ingredients for other food manufacturers and copacking, producing shortening, margarine, salad dressing, mayonnaise and sauces. Production is a hybrid of constrained lines that make to stock and flexible lines that make to order, according to Prashant Sanghvi, director of logistics.
A hybrid approach also is taken to forecasting, resulting in “a best of breed system that has its own challenges,” Sanghvi says. Although he formerly worked with another supply-chain software firm, he opted for Demand Solutions’ software as a compromise between cost and functionality. “Sometimes you can get by with 80% of the functionality and customize the remaining 20%,” he explains. Besides, “complex solutions don’t work in industries with short lead times like food.”
A make-to-inventory model is a pipedream in food, he believes. Ventura’s foodservice orders have to be filled in multiple plants to minimize distribution costs, while retail orders must be made to stock. “It’s physically impossible to make to order for retail promotions,” he insists. While the new forecasting system has only rolled out to two plants, the early returns are encouraging: an 8-10% improvement in finished goods turns and 10-12% gain in raw material turns, plus fewer inter-plant transfers and fuller truckloads that are delivering seven-figure cost savings. “We’re getting better flow through the warehouses and have gone a long way in removing obsolescence and inventory damage,” Sanghvi says.
Go with the flowFor manufacturers selling exclusively through retail outlets, replacing traditional forecasting models with “flowcasts” based on daily store-level sales could virtually eliminate out-of-stocks and magnification of forecast miscalculations through the supply chain, particularly during promotions. Landvater and his partner, Andre Martin, pioneered the integration of distribution centers, factories and raw material suppliers for Abbott Labs in the 1970s into a forecasting system they called flowcasting. “It was never our objective to start a software company,” says Martin, but a program had to be developed to bring the concept to consumer packaged goods. The two recently created a joint venture called RedPrairie Collaborative Flowcasting Group with warehouse management specialist, RedPrairie Corp.
Customer orders and warehouse withdrawals leave manufacturers vulnerable to “the bullwhip effect,” according to Martin. A dip or up-tick in retail sales gets magnified over a period of several weeks, resulting in demand forecasts that overstate the change by a factor of as much as four.
Flowcasting requires “a completely different relationship between the manufacturer and the customer,” Landvater concedes, though he insists one has been forged by Sam’s Club and Kraft, which piloted flowcasting in early 2008. The retailer enjoys fewer out-of-stocks and more reliable forecasts, while Kraft has less safety stock and has integrated production planning with future demand.
Martin predicts retail-based forecasts could supplant demand calculations at every point in the supply chain, extending to the manufacturer’s raw material ordering. That sounds utopian, particularly to food companies that don’t deal exclusively with retail distribution. Still, knowing what actually was purchased in a store yesterday has value. AMR’s Cecere points out it takes five times longer to determine what sold off the shelf when relying on DC shipments than when a DSD sales rep captures that information.
Regardless of how reliable demand forecasting models get, they must be translated into a production schedule, argues Mike Gay, industry manager-CPG at Milwaukee’s Rockwell Automation Inc. “Manufacturing is just a black hole to ERP,” he says, “and the genealogical records of food products can’t be handled by ERP.”
Siemens PLM Software’s Daniel J. Staresinic agrees. Just as parts management drove the aerospace industry to marry manufacturing systems to ERP, food manufacturers are migrating to product-oriented systems that address the “transactional complexity” of production and complement demand forecasting. “One little error in production assumptions magnifies,” says Staresinic, worldwide director-consumer products and life sciences at the Milford, OH business unit, “and then you have a massive reconciliation at the end of the month.”
Forecasting errors are everybody’s headache, and whether they result in lost sales or excess inventory, manufacturing has a stake in better demand forecasting tools.
For more information:
Lora Cecere, Bill Harrison, Olin Thompson, John Bermudez, Andre Martin, Mike Gay, Daniel J. Staresinic, Pat Anderson.
Masters of the chainWith manufacturers of all types struggling with a sour economy, the value of demand-driven forecasting and supply chain visibility is at a premium. If those cylinders are firing, production schedules can be fine-tuned, and excess inventory and product returns can be minimized.
Electronics companies are the leaders in this area, according to AMR Research Inc., which issues an annual ranking of the top organizations in supply-chain execution, based on revenue growth, ROA, inventory turns and other criteria. The best of the rest in AMR’s top 50 list includes seven food and beverage companies. Those firms, and their ranking in the overall list, are:
1. Pepsico (9)
2. Coca-Cola Co. (13)
3. Unilever (22)
4. Nestlé (28)
5. Kellogg (36)
6. SABMiller (42)
7. General Mills (50)
For more information:
Lora Cecere, AMR Research Inc., 617-542-6600
Bill Harrison, Demand Solutions, 314-333-5918
Olin Thompson, Lawson Software Inc., 651-767-4397, email@example.com
John Bermudez, Oracle Corp., 781-744-0000, firstname.lastname@example.org
Andre Martin, RedPrairie Collaborative Flowcasting Group, 450-437-3616, email@example.com
Mike Gay, Rockwell Automation, 414-382-2000
Daniel J. Staresinic, Siemens PLM Software, 513-576-2184, firstname.lastname@example.org
Pat Anderson, TAKE Supply Chain, 512-231-8191