Demand-based production is making progress in food and beverage plants,
but optimizing the supply chain will require greater cooperation at the
Technology is optimizing merchandising efforts on behalf of food manufacturers in stores, but taking it to the next level hinges on retailers’ willingness to increase collaboration. Source: Trimble Corp.
Brotherhood and goodwill never have characterized the relationship between retailers and suppliers. The dynamic between grocers and food and beverage suppliers usually is particularly contentious, talk about collaborative planning and trading partnerships aside.
Some thawing of the traditionally frigid stand-off is beginning to occur, much to the relief of manufacturers who would like to improve product availability on store shelves while also reducing transportation costs and inventory levels. But the day when food processors can sharpen their forecasts sufficiently to accomplish those goals still is several years away. One food industry supply-chain executive recalls broaching the topic of real-time sales information from a retail customer. The conversation didn’t go far: “He said, ‘When I place an order, you will have product available,’” the executive relates.
Still, food companies are deploying more technology and higher degrees of sophistication to their store-level programs to improve sales predictability and reduce out-of-stocks. Considerable advances in automated forecasting models are occurring, though they remain a work in progress. “The forecasting function is still trying to determine best practices,” notes Sean Buckley, manager-demand process at Canandaigua, NY-based Constellation Wines US. “Is top-down forecasting superior to bottom-up forecasting? That’s up in the air. The only certainty is that you don’t want a purely judgmental or purely statistical model: combination forecasting is always a benefit.”
Better supply-chain visibility is beginning to help optimize production, though progress is dependent on degree of comfort between different groups within the organization. “When we started to look at how demand forecasting could affect our business, we were very much manufacturing focused instead of sales focused,” says Harold Upton, vice president-strategic business processes at Yuba City, CA-based Sunsweet Growers Inc. “Instead of keeping customers happy, production wanted long, efficient runs. The tension between operational efficiencies and producing to demand is an ongoing battle.”
Every manufacturer is attempting to improve demand forecasting. Depending on how finished goods go to market, the degree of visibility to the supply chain varies. For many, spreadsheets with historical sales are the only data source, and forecast accuracy depends on managers’ intelligence and intuition. For others, sophisticated forecasting tools are helping reduce inventory levels. For the largest brand managers, the brass ring is global data synchronization (GDS), though it’s unclear if many trading partners are willing to grab it.
Better data synchronization can drive down transportation costs, even if the synchronization issue is within the organization itself. In a recent study by the Grocery Manufacturers Association (GMA) and Food Marketing Institute (FMI), one manufacturer uncovered $2.2 million in wasted transportation costs because the system used to stage shipments was out of synch with its production system. Incremental product changes resulted in discrepancies in actual case weight and the weight reflected in the shipping database. Two additional pallets of products could be accommodated on inter-company transfers. The company projects additional savings of up to $3 million when all product weights are accurately cataloged.
The GMA/FMI study, “Synchronization-the Next Generation of Business Partnering,” outlines the brave new world of supply chain visibility envisioned by industry leaders. It attempts to present best practices and to complement expectations of soft savings from GDS with hard numbers. Retail benefits are based on results at Wegmans Food Markets; eight food manufacturers provided support and statistical input, including Coca-Cola Co., General Mills and Nestle. Those firms were among the 35 food companies that helped pony up $250 million in 2000 to launch Transora, the industry’s first stab at Web-based collaborative planning, forecasting and replenishment.
The race to the shelf is critical for leading brands, and data synchronization can save considerable time. Manufacturers recognize that the best margins are captured early in the rollout game, before their new product is copied and commoditized. In the study, Wegmans was able to get new items onto the shelf seven days faster, a 23 percent improvement. GDS helped the supermarket chain enter data about new items into its system in half the time.
The bigger winners are likely to be food companies, the study notes. “One manufacturer reduced its speed to shelf from four-to-eight weeks to two weeks, a 67 percent improvement.”
Soft drink distributors, snack food companies, bakeries and other food companies with direct store delivery (DSD) systems also stand to gain from greater collaboration and synchronization. In the GDS study, Wegmans calculated savings of $163,000 in DSD backdoor operations, primarily from less spot-checking of goods.
Demand forecasts and supply availability are helping Sunsweet Growers stabilize production labor requirements. Source: Supply Chain Consultants.
Electronic eyes and ears
With club stores and mass merchandisers demanding more vendor-managed inventories, manufacturers are trying to leverage the point-of-sale transaction data that is part of the deal. Scan-based training (SBT)-disparaged as consignment selling by many-is being mandated by retailers such as Grand Rapids, MI-based Meijer’s, a chain with a product mix similar to Wal-Mart Super stores. Smith Dairy Products in Orrville, OH, was among the suppliers required to switch to SBT, getting paid for what is actually sold instead of what is delivered to Meijer’s back door. SBT forces Smith to share the cost of product shrink, but it also has helped to boost net sales and reduced inventory levels because of its access to sales data, according to Scott Gift, Smith’s regional sales coordinator. “It’s been eye-opening to see what’s really happening at the store shelf.”
Smith Dairy taps into a subscription service from Prescient Applied Intelligence, a West Chester, PA-based firm that pulls sales data from retailers’ extranets, cleans the raw data and relays it to manufacturers. “The initial pushback to SBT was in the area of shrink,” admits Jane Hoffer, Prescient’s president and CEO. “Beverage vendors in particular do not want to do scan-based trading, but other categories, particularly those with DSD, are realizing the benefits of wider adoption.” Detailed, accurate sales data helped Smith gain a full day of production, for example. “The level of detail results in 100% accurate invoices and gives route people the ability to spend more time merchandising,” she claims. “We’re seeing companies using SBT as a way to get in the door with retail accounts they otherwise wouldn’t sell.”
One of her firm’s services is the Consumer Zoom Demand Network, a data-exchange service designed to slash out-of-stocks by giving suppliers a transparent view of product movement. Empty shelves are killers for retailers, Prescient research confirms: half of consumers surveyed say they will visit a different store if their grocer is out of stock when they need an item. By far the most critical replenishment item is milk, followed by bread and eggs.
Just as an army needs boots on the ground after an aerial attack, manufacturers need bodies in the stores to complement data-capture technology. Managing the mobile worker is the focus of Trimble Mobile Solutions, a Cambridge, MA, firm. A sales blip can be forecast when sales reps win extra in-store displays, says Trimble Vice President Cyndee Hoagland, but if the displays aren’t actually in the aisles, the sales surge won’t occur and too much inventory will be in the pipeline. Putting auditors in the stores to verify promotional programs is one of Trimble’s services.
“Speed to shelf is one of the areas where we have had a great impact,” adds Hoagland. “What is the compliance level on end caps in a rollout? Does the supplier need more merchandise on the street?” No matter how sophisticated demand forecasting models become, verification of behavior is required.
Trimble equips delivery trucks with GPS boxes with Bluetooth wireless connectivity to help companies manage that element of distribution. This year it will bring GPS to handhelds used by DSD route salespeople. Algorithms built into the handhelds already calculate the impact of promotions and historical item movement in a given store; now clients like Philadelphia’s Tasty Baking Co. will know how long their reps will spend on each aspect of merchandising. “It becomes a tool for training,” says Hoagland.
Greater visibility to the store is the first step in solving supply-chain problems that manufacturers don’t realize they have, adds Hoffer. “We see examples of claimed deliveries by independent route operators where the goods actually were sold off the back of the truck,” she says. Those situations become obvious very quickly when you track shrink by item and by store.”
Making sure product is available when consumers want to buy is the main benefit of retail-information systems, and “the algorithms for managing the shelf have become quite sophisticated,” Hoagland says. She cites a client who factors in the impact on daily store sales when monthly welfare checks are issued.
Robert Byrne scoffs at any suggestion that real-time sales can positively impact production schedules. “No one is going to be able to make to order in the food industry,” flatly states Byrne, president and CEO of Norwalk, CT-based Terra Technology. “Anyone who tries will be constantly out of stock, particularly for their slower moving items.”
A relative newcomer to the demand forecasting area, Terra relies on a pattern-recognition model to help companies improve the accuracy of short-term sales forecasts. “For most food companies, safety stock is 35-40 percent of their inventory,” Byrne explains. “That’s their hedge against forecast error. If we can cut short-term forecast error in half, that means 20 percent less safety stock.” Last spring, Procter & Gamble Co. began using the model in Europe and will roll it out globally in the next two years. By improving forecast accuracy one to six weeks out, P&G expects to reduce inventory levels 10 percent.
“Many of our clients have case fill rates of 99-plus percent; customer service levels already are high,” says Byrne. “Inventory reductions are the issue.”
Manufacturers’ forecasts based on previous promotions and other sales variables are delivered on handhelds to DSD route salespeople for in-store merchandising, but adjusting production schedules based on actual sales remains a challenge. Source: Trimble Corp.
The upside of terror
Oddly enough, the Bioterrorism Act of 2002 may have set the table for mid-tier food firms to use automated demand-forecasting tools. Terra’s predictions feed into ERP planning programs that drive scheduling, and ERP systems have become almost universal as companies automate recordkeeping to meet track-and-trace requirements. “The Bioterrorism regulation has been a great lead-in for ERP,” agrees Rebecca Gill, vice president-marketing at ERP vendor Technology Group International, Toledo, OH. “Regulations are pushing more and more computer usage, and small- and mid-sized companies are starting to embrace ERP planning and scheduling.” However, ERP usually comes up short in forecasting demand, and much of Gill’s work involves integrating specialized forecasting models into the ERP scheduling function.
One of the leading model providers is Logility Inc., Atlanta, which is trying to incorporate more POS data in its collaborative planning and forecasting modules. Karin Bursa, vice president-marketing, characterizes data availability as “sporadic,” with “pockets of business focused on sharing data and having it in place. We’re still a few years away” from having enough store-level information to improve manufacturers’ production planning and scheduling, she says.
For Logility clients selling through distribution, collaboration with retailers isn’t an option. Constellation Wines may be the world’s largest wine company, selling more than 2,500 SKUs produced in 10 US facilities, but GDS is a non-starter. Scanner data is sketchy at best and modest in coverage; in some states, virtually all sales are through small independent shops. “Collaboration is primarily internal, not external,” notes Buckley. For now, the main benefit of better sales forecasting is more efficient line scheduling, with fewer changeovers. “With high-speed bottling lines, you don’t want to tear them down every two hours to run a different product,” he points out.
With a four-year lag between planting and harvesting usable grapes, winemakers rely on long-range forecasts for sales and production. Despite a shortage of real-time retail data, producers also are trying to optimize inventory levels and adjust packaging lines accordingly. Source: Constellation Wines US
Buckley’s forecasts are posted to Constellation’s ERP system, “and that’s where all the number-crunching goes on,” he says. The process is similar at Sunsweet Growers, the cooperative that accounts for two-thirds of the world’s supply of prunes. For Sunsweet’s Upton, all elements of the supply chain are considered, from the members’ trees where the plums grow to the person who consumes the finished goods. Sunsweet contracts with growers for acreage, not tonnage, making the amount and condition of raw materials another variable in demand forecasting.
Disastrous harvests occurred in 2004 and 2005, and Sunsweet shifted production to premium products, heavily promoting dried fruit and prune juice for women’s health in the UK and other markets. The 2006 harvest rebounded, and production shifted toward volume-oriented commodities. Better forecasting has been a boon in managing production labor by smoothing out daily schedules. Overtime dropped to 12 percent of labor costs, half their previous share, according to Upton. “If you stabilize the work environment and your scheduling, it’s a huge benefit for the company and the workers.”
Like many food companies, Sunsweet previously used spreadsheets with historical sales data for planning, feeding the results into an SAP system. “ERP offers forecasting modules, but those systems are geared for transactions and volumes of data,” says Tom Leonarski, senior consultant-food at Supply Chain Consultants, Wilmington, DE. “To do effective planning, you need a model that incorporates business rules on lead times and other constraints to come up with a workable forecast.”
Leonarski’s firm developed Zemeter, the forecasting tool that feeds into Sunsweet’s ERP. Leonarski also implemented a finite scheduling program at Mother Parkers Tea & Coffee Inc. during a 32-month automation project.
“When we had fewer than 500 SKUs, a group of experienced individuals could work through the coordination of packaging and production,” explains Will Kappel, vice president of supply chain at Mississauga, Ont.-based Mother Parkers. “When you produce over 3,000 blends and 1,400 SKUs, you absolutely need an automation system.”
Atlanta’s CDC Software has a stake in both the ERP and model forecasting worlds: its Ross Systems division delivers ERP solutions, and last year CDC acquired JRG Software, a firm that enjoyed success providing scheduling software to several mid-tier food companies. “ERP is not designed for synchronization, and ERP systems typically know nothing about changeovers,” says Chris Taunton, CDC’s director of supply planning products. “When you present master schedules from an ERP system to the production floor, the typical reaction is, ‘We can’t do this.’”
CDC and a growing number of software providers are moving to Web-based programs available on a subscription basis, minimizing the need for IT support for on-demand services such as CDC’s scheduling application. CDC also is expanding collaborative forecasting to the supplier side. “Allowing preferred suppliers to see what your demand might be in two weeks through a trusted supplier portal could cut one or two days in your supply schedules,” points out Taunton.
Splintering supply chains add yet another level of complexity to managing multiple brands and promotions. “In 1990, 90 percent of consumables were purchased in supermarkets,” notes Rory Granros, director of process industry marketing at ERP vendor Infor, Alpharetta, GA. “Within five years, it’s projected that 50 percent will be sold in non-traditional outlets.” Determining where demand is coming from, designing the right products for different channels and effectively managing promotions and new-product introductions will tax even the most sophisticated automation systems, Granros says.
An example is Organic Valley of Farms, a 650-farm cooperative that applies Infor’s Adage ERP system to manage enterprise data at 50 manufacturing facilities nationwide. “We also are implementing demand and supply planning models for constraint optimization, which will allow us to better balance supply and demand,” says Organic Valley COO Louise Hemstead. Besides the extensive record keeping needed for certified organic products, the cooperative gets involved in formula management and item-pack management. Product diversity is notable: the company markets geography-specific products, including Texas Farms Organic and Rocky Mountain Pastures organic milk.
While variables beyond historic volumes are being built into demand forecasting models, seasonality and promotional events are wild cards that require human intelligence to override automated programs, maintains Rob Wiersma, director of food & beverage industry strategies at St. Paul, MN-based Lawson. When retailers are prepared to lift the veil on item movement through their checkout lanes, transparency and data synchronization will simplify inventory buildup for promotions and other events. Until then, manufacturers will focus on synchronizing as much of their in-house efforts as possible and temper forecasts with reason to optimize production planning and scheduling.
For more information:
Chris Taunton, CDC Software Inc., 770-351-9600
Rory Granros, Infor Global Solutions, 678-319-8000
Rob Wiersma, Lawson, 905-752-6507, email@example.com
Karin Bursa, Logility Inc., 404-238-8338
Jane Hoffer, Prescient Applied Intelligence, 610-719-1600
Tom Leonarski, Supply Chain Consultants, 302-738-9215, firstname.lastname@example.org
Rebecca Gill, Technology Group International, 248-363-5491, email@example.com
Robert Byrne, Terra Technology, 203-847-4007, firstname.lastname@example.org
Cyndee Hoagland, Trimble Mobile Solutions, 617-285-4362, cyndee_hoagland@ trimblems.com