Meeting the demand

Demand management offers a way to overcome variables, such as getting the right stocks to the shop shelf, which can hinder the simple law of supply and demand

  • E-Mail
By  Caroline Denslow Published  October 23, 2005

|~|Supplybody.jpg|~|Demand management offers a way to stop overstocking, which wastes money and clogs up warehouse space. |~|The art of supply chain management is often seen as getting goods from A to B as efficiently as possible. Equally important, however, is getting the right products there and in the right quantities. It is no good getting the right product to the shop shelf if there is not enough stock to meet demand. At the same time, too much stock can waste money and clog up warehousing space. Demand management systems are the best way out of this dilemma. Demand management is a set of business procedures and technologies that enable companies to recognise all the demands for products and services to support the marketplace. It covers forecasting, order entry, order promising, determining outside warehouse requirements, production balancing and spare parts. Before demand planning software was introduced to the market, forecasting was fundamentally a balancing act between competing blocs within the enterprise. For instance, the marketing department might set a high target because it wanted the product to be a success. The sales department, on the other hand, would come up wi-th conservative forecasts since they want to keep their sales targets low and manageable. Demand forecasting provides an impartial estimate that is based on statistical algorithms and mathematical formulas. “It studies the whole thing, it applies statistical formula, and it gives you the best quantity to purchase from your own suppliers so you can meet the demand,” says Christian Zakarian, senior project manager for SPAN-SNS, an IT provider of supply chain solutions. The concept behind demand planning is fairly simple. It’s about keeping a balance between supply (stocks) and demand (customers’ orders) — making sure that the company has enough supplies on hand to sustain its existing orders until the next delivery happens. At the same time, it is also about keeping supply costs down by avoiding overstocking. “You don’t want to be carrying inventory costs where something is lying in your warehouse, which is costing you money to store, and nobody is taking it,” Zakarian says. “Demand forecasting is about managing efficiency and minimising costs. It’s about getting the exact quantity that customers will be ordering for, say, the next two to three months,” he adds. One can see the attractiveness of an automated system that could offer an objective answer to demand planning, especially since the foundation — statistics and mathematical algorithms — required to build these systems has been in existence for more than 70 years. It was a British mathematician, Ronald Fisher, who first formulated a system involving number crunching to look for patterns — patterns that are then used to make predictions. The model, known as classic regression, is the basis of about 90% of all demand planning software today. Regression is fundamentally about taking multiple variables and then making assumptions about the relationships between them to identify trends that can either be positive or negative. For example, a regression st-udy of the rate of death among people between the ages of 20 and 80 would generally find that as people become older, the rate of death becomes higher. You could then predict that an 81-, 85- or 90-year-old person has a higher probability of dying than someone who is 80. In the case of demand management, regression takes into account data, like purchase orders and customer orders, and a range of variables, such as population, income and buying patterns, and perform certain calculations to determine whether sales have an upward or downward trend. Implementing demand management starts with a classification of the products or what Krishnan Sugavanam, general manager of Mantis Middle East — another supply chain management vendor — calls demand pattern. A product, he says, can fall under one of nine categories — new item, positive trend, fast moving, erratic, lumpy, slow moving, negative trend, dying, and obsolete — irrespective of what industry these items come from. ||**||Demand pattern|~|raobody.jpg|~|Narasimha Rao of Darwish bin Ahmed and Sons stresses the importance of grouping products.|~|Grouping products according to demand patterns can help a company determine where the majority of its items belong. Once that is done, the company can then generate test data in each category. Normally, during the trial-and-error phase, about 10% or 15% of actual data from each group is used for the forecasts. Multiple dummy predictions are created until the company is confident with the numbers presented. “We generate rounds and rounds of repeated iterations of forecasting to basically give the customer a comfort feel,” says Sugavanam. “We have to go through several trials and errors before we can say, this is the forecast that is acceptable. Once the client has accepted the results, we then ‘freeze’ the system and let it stabilise for a period of six months,” he adds. “It’s like fine tuning a car,” says Narasimha Rao, IT division manager of Darwish bin Ahmed and Sons, a UAE-based company whose line of business ranges from automobile and spare parts distribution to IT solutions provisioning. Darwish bin Ahmed and Sons, which is using a Syncron demand forecasting system, relies on the automated solution to support the planning needs of its heavy equipment spare parts business. “You have to tune the whole system — your demand, the pattern of the demands — all of that has to go through the first implementation period. That is key. To have a successful rollout you need to have a good implementation,” Rao adds. Although a demand management application is a standalone system, it is ideal to link it up with a supply chain application or an ERP (enterprise resource planning) application that the company has installed. By doing so, the company can ensure that there is only one version of data being used across its organisation. “Suppose you have an ERP system that maintains your orders and stock on hand. This information should be entered in a demand management software on a regular basis,” Zakarian says. There are four basic sets of information needed by a forecasting system: sales, current stock, current sales on hand and current purchases orders on hand. On top of this basic information, there are other variables that need to be considered. One is basic forecasting or guaranteed sale, which is the minimum quantity that can be sold irrespective of what happens in the market. A company should also factor in trends, such as market acceptance of the product, and the seasonality of the item. Another variable that should be noted is user input — things like promotions, the opening of a new branch or the expansion to new territories — that can only happen once in a while and do not follow a specific pattern. But the most informative for demand planning software is an item’s sales history, which consists of past records of purchases and orders and other related information. It is by using this data that the application can establish the pattern for the item from which it can start its forecasts. If the item in question is an existing one, recovering the history of that product is easier. But what about new product launches? Where does the customer begin? “For new products you have two cases from which you can simulate history,” Zakarian says. “If the product is similar to an existing product, you can attach the history of the old item to the new one and forecast for it,” he states. Take for example, the Nokia Communicator 9500, which was launched last year. It is the latest in Nokia’s line of enterprise-grade mobile devices, and thus, during the immediate period after its release, had no sales history of its own to support its forecast. But given that it comes from an established set of Communicators, demand for the Nokia 9500 can be based on the history of previous models. For instance, the company can link sales data about the Nokia 9210, an older Communicator version, and treat it as part of the 9500’s sales history. “You have to included it in a group. It can’t be that from one month it is zero and the next it goes up,” Rao says. “You have to put it in one basket.” On the other hand, if the item is a new launch and does not have a link to any established products available in the market, the best thing to do is to rely on market research that the vendor normally has conducted before releasing the product. It will then take about six periods of data before the system’s forecasts for new products become reliable. ||**||Service levels|~|zakarianbody.jpg|~|Chri-stian Zakarian of SPAN-SNS forecasts better management practices.|~|As with any enterprise applications, demand management dictates that a company should establish service levels with the software provider to ensure the effectiveness of the system. Unfortunately, most of the companies in the Middle East that have implemented a demand management system have very low service levels or, worse, none at all. “What we find is that current service levels are as low as 40%, and what is worse is that most companies don’t even measure services,” Sugavanam says. A demand planning service level measures the percentage of the fulfilment of sales orders to customers, Zakarian says. Ideally, it should be above 90%, but it varies a lot depending on the type of item involved. Fast-moving goods need a much higher service level — about 99% — while slow-moving goods can have service levels close to 90%. On average, Sugavanam recommends 98% for all items. “If a company were to achieve 98% then it is absolutely doing well,” he claims. But no matter how good a company's demand management system is there are those that have suffered from bad forecasts, which ultimately caused them to lose a lot of money. One of the more famous forecasting fiascos happened five years ago and involved sneaker giant Nike. The debacle happened as soon as the company went live with i2 Technologies’ planning system. The demand management system underwent a year of implementation work at Nike, but after the company decided it was time to turn on the switch, the result was one massive supply chain mess. Nike’s experiment with demand forecasting caused it to lose more than US$100 million in sales — a massive loss caused by overstocking items that did not sell well and running out of stock for the more marketable ones — depressed stock prices by 20%, and earned the ire of Nike chairman, president and chief executive Phil Knight, who famously complained to analysts, “This is what we get for our $400 million?” There are definitely many reasons why forecasting systems fail. To start with, these systems are only as good as the data put in them and, because of the complexity of modern supply chains — where a company wants to collect information about multiple products from multiple customers and suppliers — more often than not the data is not accurate enough. “History must be clean. The data must be acceptable because we’ve seen a lot of companies that are not consistent with the way they are actually recording their transactions,” says Zakarian. User intervention — or the lack of it — can also lead to poor demand predictions. Some companies believe that the system can do everything on its own and, therefore, fail to monitor the whole process, especially data input, regularly. “The application is supposed to do everything. But it also needs maintenance. Some companies just let the system do what it has to do and forget to monitor every item sometimes. They should not let it run by itself,” Zakarian explains. In contrast, tampering with system-generated forecasts can also cause the application to fail. Some companies find it hard to accept and follow the calculations done by the system that they manipulate the data to something that they are more comfortable dealing with. The figures may not be accurate, hence leading to overstocking or shortages. “Forecasting happens to be unemotional. When you give forecasting, it’s done without passion,” says Arun Kumar Panneervel, consulting manager at Mantis Middle East. “In forecasting you have to be abstract; you can’t have your emotions getting in the way. Demand forecasting may recommend something that you may not like. If the forecasting software asks you to order so much, you may not like it, but you have to do it. Unless you let your system learn, it’s not going to work,” he adds. An effective demand management system requires a combination of accurate data and smart people. While the system helps provide accuracy because of its scientific nature, a company’s experience and knowledge about the market is also needed to weigh in how the forecasts fit with the bigger picture. ||**||

Add a Comment

Your display name This field is mandatory

Your e-mail address This field is mandatory (Your e-mail address won't be published)

Security code