Multinational food manufacturers are commonly powered by vast operational networks of production facilities and distribution centers to meet the demand of broad customer portfolios spanning various countries on different continents. Striving to build the ideal of strong local footprints in major established markets and expand into emerging markets, these manufacturers require a complex, yet efficient and flexibly structured supply chain.
Following key customers to new markets and growing hand-in-hand, leads to a border-crossing flow of goods and an increasing variety of products or SKUs. Contradictory to the industry-wide call for SKU rationalization, the long tail continues to grow. For example, in 2011, the consumer good industry grew, in part, because of low volume items representing 80 percent of the overall variety of products available to consumers carrying relatively high inventory costs (1).
As a result, well-planned operations that addressed key touch-points from long-term investments, to labor staffing and short-term maintenance plans for production lines were needed (2). Organizations based these planning processes on projections of how their markets would evolve and shape future demand forecasts.
In that regard, a 1996 study, which was part of the same year’s “Food and Drinks Industry Benchmarking and Self-Assessment Initiative” (FDBSI), stated that more than 48 percent of participating organizations considered their own forecasting processes poor and insufficient. This research was largely driven by the Leatherhead Food Research Associates and more than 50 food companies, including Campbell Soup Co., Kraft Jacobs Suchard AG, McKey Food Service Ltd., Scottish Courage Brands Ltd., Seaforth Corn Mills, HP Foods Ltd., Smithkline Beecham and Van den Bergh Foods.
The project members openly admitted they did not have an adequate idea of their respective markets’ futures, describing demand forecasting as a critical process (3). Indeed, recent studies such as “Forecasting Performance for North American Consumer Goods” by Robert F. Byrne (2012) show that during the past two decades overall forecast quality has increased, mimicking the overall esteem for accurate demand forecasting. It also demonstrated that there still are big challenges to modern forecasting methods, which are addressed below.
The driving factors behind this growing appreciation for, and understanding of, the demand planning matrix is hard to ignore, particularly in food manufacturing where the shelf-life of raw materials and finished goods are limited. Vigilance is essential on each of the three fronts of demand planning (short-, mid-, and long-term), but for very different reasons. The short-term demand forecast determines the production schedule two to four weeks out, often for production to stock. The mid-term steers raw material procurement and staging, and the long-term projection is indispensable for the assessment of potential investments in operational assets, as it shows where sales will go and thus investments should flow.
Twenty years have passed since the 1996 study and its industry self-reflection, and food companies are working hard on improving forecasting capabilities. Businesses across the food and beverage supply chain, from retailers and manufacturers to processors and growers, are taking on the effort of balancing their production and/or distribution operations with assessed demand. They have realized that well-planned operations with a focus on customer demand are key in offering high-service levels in a steady and efficient manner. Hence, planning processes that enable organizations to understand the exact shape of demand in regards to timing and quantity have been implemented industry wide.
Besides described strategic and tactical values, high forecast accuracy is often seen as low-hanging fruit to positively impacting efficiency and profitability key performance indicators. Particularly in the mid- and short-term, a solid forecast can have great benefit on inventory levels of finished goods and raw materials, improving operating measures such as Stock on Hand, Carrying Cost of Inventory, Inventory to Sales Ratio and eventually Perfect Order Rate.
Summarizing the above, a well-developed demand-planning process supports corporations in balancing the contradictory triangulum of increasing throughput and thus increasing asset utilization, positive customer experience by improving service levels, and decreasing cost by reducing inventory (2). For an example of this contradiction, one can look to businesses exposed to seasonality of demand, such as the confectionery industry.
Aiming for high-cost efficiency, candy organizations must find the right balance of investments in production capacity and inventory cost. They do so to avoid unnecessary downtime of their production lines and prevent high stock levels with risk of expired finished goods in periods of low demand, but still provide proper service levels to customers during high season.
To do so, an aggregate level forecast at the product group level, which most companies are able to perform, is not sufficient. It usually does not include strong enough information to support production and raw material sourcing planning. To support those functions and create the most valuable forecast data, it is necessary to break down the data to SKU level, as these details determine the exact ingredient and production requirements.
However, it must be noted that an accurate understanding of future demand on product group level can assist in investment decisions on operational assets, though the quality of this high level forecast is enhanced by recognizing the origin of the demand on a SKU and customer level. The accumulation of detailed forecasts will indicate the optimal location for long-term investments to increase asset utilization and drive superior customer service.
To facilitate the development of such detailed forecasts at the SKU level, a variety of forecasting software is available to support demand planners, either as stand-alone platforms or subsystems of entire enterprise resource planning (ERP) ‘ecosystems,’ such as advanced planning and optimization (APO) for SAP.
These systems typically use historical sales data to provide a baseline forecast, not requiring any user involvement. The seasonality of products is recognized and automatically applied to the statistically projected demand figures. If the demand is expected to change substantially, regardless of sales history, software systems require users to choose from various statistical models, such as exponential smoothing, to shape or skew the forecast.
Given the necessary data, available planning software will readily forecast the seasonal demand levels for confectionery products around the holiday season. This does not, however, come without human interaction with, and interpretation of, the data inputs and outputs.
Regarding planning for promotions, within the group of FDBSI members, supplier companies experienced an increase of short-notice promotions, leading to heavy strains on supply chain processes (3). They described promotions as the largest controllable event that can happen between retailers, suppliers and growers that represents one of the main reasons for shortfall or surplus of stock and manufacturing inefficiencies leading to poor customer service and customer dissatisfaction.
Despite an increase of overall forecast performance since the FDBSI study, recent studies show an upward trend in forecast error, which is attributed to the “rapid pace of promotional activities and product innovation” (1).
Innovations are, of course, the lifeblood of a healthy consumer market, and manufacturers are driven to the meet the demand for innovation. As a result, sales volume of innovations —products introduced less than 12 months ago — continue to rise. Likewise, the unending requirement to increase the velocity of off-take drives promotional activity, as indicated by promotional volumes climbing eight percent in recent years.
For example, in 2011 promotions represented nearly one-quarter of the overall sales of such companies as Kraft Foods, ConAgra Foods Inc., Kimberly-Clark Corp., Unilever, General Mills, Inc., Procter & Gamble and Campbell’s. The effects on supply chains prove the increasing complexity must be managed to ensure excellence in execution.
To this point, consideration must be given to the supply chain from end to end — a promotion or an introduction of a new chocolate chip cookie in retail stores does not only affect the production of these cookies, it also affects the suppliers of cookie dough or dough ingredients, chocolate chips and packaging, and the suppliers that manufacture these inputs. The consideration of time in this matter is of great importance as well. For example, in a perfect world the members of the chocolate chip cookie supply chain would know about the unusual increase of demand resulting from innovations or promotions well prior to scheduling raw material sourcing.
Despite the outstanding benefit statistical forecasting systems offer to modern corporations by taking on the data heavy lifting, it is apparent human input is required to enable the software to draw a more realistic picture of high demand periods that are not based on market information extrapolated solely from historical patterns.
It is these non-experience based demand changes, perhaps through new promotions or new demand because of product innovations that remain the biggest challenges in the demand planning process. Hence, it is data indicating this, as refined by human input, which offers the biggest value to today’s forecasting cycles.
In addition to applying internal downstream data from what is called “demand shaping processes,” demand projections provided by customers, manually entered and interpreted by forecasting professionals, are highly valuable inputs that increase forecast quality. Academia describes the demand shaping process as steering customers’ demand in a strategically valuable direction for the supplier by introducing products, adjusting prices, running promotions and adjusting the products (i.e. increase/decrease package sizes, changes to ingredient decks to accommodate changing customer demand to organic products), or adjusting the product placement in the marketplace (4).
Given the necessary data, available planning software will readily forecast the seasonal demand levels for confectionery products around the holiday season. This does not, however, come without human interaction with, and interpretation of, the data inputs and outputs. – Felix Koch, Barry Callebaut USA, LLC
As these inputs are usually driven by marketing and sales teams, the integration of these teams’ demand-specific knowledge in the forecasting process is crucial to driving precision in the projection of future sales volumes on customer-to-item level. This is one of the most important steps in closing the gap between the commercial end of a company’s business and its supply chain organization. Some scholars, such as Chase, maintain it is sales and marketing’s responsibility to gather all information in regard to sales promotions and marketing activities that influence the customers to purchase their product, and to consequently share these insights with the forecasting function to allow for the necessary adjustments to the internally used demand projections.
Tying sales and marketing teams’ performance to forecast accuracy can increase and support the needed communication. Modern sales and operations planning (S&OP) processes often apply accurate forecast data as demand thresholds for capacity management. To ensure proper service levels for all customers, the forecast provided by “Customer A” serves as a ceiling to that demand, for example, on a monthly basis. This process and its benefit of steady supply must be thoroughly explained to customers to create the appropriate sense of value and appreciation. This enables suppliers to commit to increasing volumes while maintaining high service levels, and also pushes operational planning improvement throughout the entire supply chain.
At the same time, every modern corporation that holds a supplier role within the food manufacturing supply chain strives for high service levels and needs to make efforts to introduce the demand forecast data it receives into its operations planning cycle. With this data, suppliers can determine which customers need what product in what timeframe and steer their operations in an efficient, but more importantly, customer beneficial way.
Customers will be willing to share downstream data if suppliers are able to illustrate how that data will be used to build value for both parties. To illustrate this, suppliers may rely on research that continuously demonstrates the strong correlation between demand visibility, supply chain performance and customer service. With the size and complexity of customers’ demands on the rise, increasing effort should be made in the forecasting process. The higher the purchased volume of one customer, the more this customer’s demand behavior will affect the supplier’s supply chain processes.
A monthly forecast consensus meeting or call can be used to analyze, discuss and agree upon the projected demand figures for significant customers. In these meetings, the mid- to long-term forecast (six to 18 months) should receive the most attention, including discussions of upcoming promotions and introductions of products. Since a statistical forecasting system will not grasp the pending demand of new products (because of the absence of historical data) it cannot function as an early warning system, and thus the timely communication of innovations is key in making new products a success by ensuring proper supply to the marketplace. Communication processes as such are supported by plenty of available CRM software, which, just like monthly meetings, formalize and structure communication and data exchange outside of the usual email traffic.
Within the candy sector, it is common to incentivize internal functions that possess critical insights in regards to demand (such as marketing, R&D and sales), to share this knowledge in a formal and structured process with demand planning functions.
To increase the quality of the baseline statistical forecast, this practice is followed by Barry Callebaut in addition to closely cooperating with top volume customers in regards to future demand behavior. The company uses customer- and item-specific demand data to manage production and stock positions, such as determining slow-moving items. Furthermore, Callebaut utilizes the detailed demand forecast in combination with macro economic data to make well-founded investment decisions for new production lines to increase its local production footprints.
For example, Callebaut’s production network expansions in Monterrey, Mexico; Chatham, Canada; and American Canyon, CA, will increase the overall production capacity of the North American Organization by 10 percent. The Monterrey and Chatham facilities were completed in Q3 and Q4 of 2016, American Canyon will be completed in Q3 2017.
To be clear, a strong forecast goes hand-in-hand with cross-functional communication within, but also between companies. Manufacturers that expect high service levels from their suppliers should push to be involved in their suppliers’ forecasting process.
Thus, in the progression toward the purely statistics-based software crystal ball that enables corporations to produce what they envision as demand, companies are able to produce what they sense they can sell, by both receiving and applying internal and external downstream data. It is proven that a solid statistical forecast based on historical data serves as a perfect foundation for a high-quality forecast. On top of this foundation, a high-quality demand projection can be constructed by adding downstream data from internal and external sources to formulate the demand in a more precise fashion. Only this enables a company’s supply chain organization to fully benefit its customer and itself, while also mastering the high complexity of global supply networks.