Faced with volatile markets where consumer demands shift ever more quickly and costs need to be controlled, improved forecast accuracy is paramount. An efficient and responsive supply chain is essential for competitive advantage, says Brian Davis.
Traditional demand planning systems give an indication of average sales of products for this time of the year, but don’t necessarily leverage current information in the supply chain — including customer inventories, point of sale and other information.
Companies like Unilever, Proctor and Gamble, Kraft and others are exploring new ways to forecast demand because consumer purchasing habits, retail and internet ordering processes and promotions are changing so frequently.
Terra Technology’s demand sensing system enables companies to gain supply chain visibility and improve manufacturing planning. “Demand sensing sits between demand planning and distribution planning, predicting customer shipments,” explains Robert Byrne, CEO of Terra Technology. “By getting this data more accurately, the supplier’s deployment, stock transfers and manufacturing schedules are in turn more accurate.” Demand sensing utilises pattern recognition mathematics to decipher daily demand signals and generate more accurate forecasts based on real world events in near real-time. Unilever’s Personal Care business managed to reduce forecast error by 25%-40% using Terra’s Demand Sensing solution.
Kraft Foods reduced its demand forecast error by 40%, cutting costs, improving service levels and reducing inventory in North America, and P&G cut its forecast error over 30% globally.
Making plans
Commercial demand planning systems have been around for about 25 years, and were supposed to solve the forecast error problem. But forecast error got worse because of many factors, including line extension in retailers, increasing product diversity, innovation, more promotions, special packs and evolution of consumer taste. Admittedly, forecasting is as much an art as a science.
Traditionally manufacturers have put in buffers of inventory to cope with demand variations. The majority rely on statistical analysis using Microsoft Excel spreadsheets. “Depending on the business, margins of error can be process driven,” says Andy Killick, business consulting manager for EMEA at Infor. “Spreadsheets are simply not designed for a collaborative multi-user process. Statistical analysis from spreadsheets tends to be general rather than fit for purpose, as each industry has a specific demand planning challenge.” Infor addresses the planning challenge with a series of end-to-end supply chain solutions for demand planning, advanced planning and scheduling, warehousing and transport management. Infor’s supply chain software is ERP agnostic, which is important as many suppliers have undergone acquisition and utilise numerous ERP systems, and it is vital to have total visibility of the supply chain.
In many cases, the manufacturer is not concerned with replenishment of the stores, but replenishment of the distribution centres.
Consequently, manufacturers supplying the retail space operate from two dimensions. They are market driven from a customer perspective, as handled by the account manager, and also manage from a logistical perspective. “Infor offers a demand forecasting system that allows the manufacturer to map the customer dimension, driven by sales, mapped onto their logistical structure in warehouses which often don’t have a one-to-one relationship with customers,” says Killick.
Collaboration stations
Infor SCM Demand Planning uses advanced statistical techniques based on Bayesian time series analysis.
As an umbrella system that operates across multiple ERP systems, this offers a single collaborative business process for higher customer service and less inventory, closer to real-time. AB World Foods has deployed Infor SCM, Infor SCM Demand Planning, Infor SCM Inventory Planner and Infor SCM Replenishment Planner as a foundation for managing its global supply chain.
Prior to implementing Infor’s SCM solutions AB World Foods relied on a spreadsheet based on retrospective performance over the previous two months. This process provided a flat forecast of demand and did not take into account different demand patterns for various types of products. Given that 75% of AB World Foods’ costs are supply chain related, phased implementation is having profound benefits across the business. “Our service levels for inventory planning have increased from 91-96 per cent, and stock holding of finished goods has been cut by about 20 per cent, saving over £2 million,” says Gary Brookes, head of supply chain.
Collaborative input has also improved across the business due to better demand forecasting. “Sales and marketing used to be out of step with the supply chain department. By integrating demand planning with our stock plans, we have significantly improved sales and risk reporting,” comments Brookes. Looking to the future, AB World Foods now plans to develop strategic demand forecasts for the next 12 and 24 months to guide stock replenishment action.
“During the downturn there has been increasing variability in demand,” says Tim Lawrence, lead of the supply chain team at PA Consulting. He cites the impact of the scrappage deal on the auto-industry and how supply actually outstripped demand.
“Consequently, the ability to forecast has been brought sharply into focus.” He suggests there are a number of ways to improve demand forecasting.
“First, get closer to demand.” In the retail sector access to point of sale, RFID, and online purchasing data, means information is able to be fed more rapidly up the chain, so real demand is reflected more accurately. For example, when an Apple iPhone is activated, the information is relayed immediately to Asian suppliers who can gauge which models are favoured and react rapidly.
Ready for casting
Oliver Wight partner, Les Brookes, also suggests that getting closer to real consumption demand is more important than ever — “This demands more collaboration with trading partners up and down the extended supply chain.” He stresses the importance of principles like demand flow costing, where systems (from companies like Red Prairie) utilise electronic point-of-sale (EPOS) data to assess demand patterns over the short-term. But he warns against the potential for amplification in this scenario, “As sales at a remote location can be misinterpreted further down the line, like a Chinese Whisper, resulting in significant overstock.” Historically, demand in the auto-industry, for example, was fairly predictable. But in the recession there is a real need for consensus-based demand planning rather than just using OEM forecasts.
“There is a need for suppliers to have their own intelligence to factor in true demand, as amplified demand can lead to overstocking,” says Brookes. He points to the Forrester Effect, which says the further you are from forecasting real demand in the supply chain, the greater the potential for amplification.
Consensus-based demand planning, rather than a statistical based forecast, allows companies to factor in market intelligence, customer intelligence, economic and market share assumptions. Systems including JDA Demand Planning, SAP APO Demand Planning and Oracle’s Demantra, all support consensus-based demand planning.
Nevertheless accurate demand forecasting also requires culture change, as the process demands higher visibility throughout the supply chain.
“Optimisation requires improvement in the accuracy of demand plans, moving more intelligence around sales and marketing activities. But this results in improved service, minimises costs and reduces working capital.”
Worst case scenario?
ERP vendor Epicor has introduced a statistical forecast system from Business Forecast Systems, called Forecast Pro in Epicor 9. The system can generate multiple ‘what if’ scenarios. Epicor also features a fully embedded multi-constraint optimisation for taking forecast demand, actual sales and optimising production plans.
“Historically companies have had to balance what they need to make with what’s most efficient based on confirmed sales orders or available capacity,” says Adam Prince, senior director of product marketing at Epicor. “Companies used to make finished goods but are increasingly creating components or semi-finished goods for completion when the order is confirmed. Their best guess comes down to demand forecasting.”
Previously, Epicor incorporated Smart Forecasting for this purpose, and still offers it on the Vantage product line. More recently, Forecast Pro has been introduced on the Scala line. Prince suggests: “For manufacturing highly configured items, manufacturers will need a strong intermittent demand capability. Whereas for commodity or CPG suppliers, where there is only minor configuration of the product, such as the packaging, a statistical algorithm such as Holt-Winters exponential smoothing is useful for demand forecast.”
“Organisations that have multiple sales and production plants also need different multi-level statistical models, with aggregation capabilities. So historic demand can be compared with plant availability using APS or a supply chain optimisation engine to smooth out demand and utilise manufacturing capacity effectively,” he says.
Straight to the source
Daylight Supply Chain Services has developed a web-based supply chain engine which automatically balances demand and supply. According to managing director, Tony Hardy, Daylight middleware will integrate with any ERP system, so suppliers can truly see what clients demand and then make empowered decisions on how to service that demand, driving critical mass production for supply optimisation.
“In some cases we’ve managed to reduce client stock by 70 per cent in the confectionery industry, reduced pipeline stock by 30 per cent in the pharmaceutical sector, and reduced supplier made readies by 35 per cent, as well as improving logistics delivery frequency by 25 per cent,” he says. Latest developments include a supplier web connect and demand management vehicle for small suppliers and a new interactive price variation dimension in the core Daylight system.
Daylight originally developed the system to manage supplies to Chesapeake, a paper board and packaging supplier with 40 operations worldwide. The company’s suppliers deliver on a cycle of up to three months, while its customers demand products within 10 days. Consequently, material needs to be sourced quickly and cost effectively in a complex supply chain. The internet-based middleware integrated with Chesapeake’s existing sourcing and stock systems, analysing behaviour patterns and anomalies within the supply chain, and recommending actions to avoid stock shortages or overstocking, saving the company £1m in the first 12 months.
Better integration of demand forecasting and supply chain modules promises to reduce the guesswork in planning, cut costs and overstocking for many industries, not simply those involved in CPG.
