)
Not many can look into the crystal ball and predict the future. But if one wants to manufacture right and sell all the products, the key is to forecast accurately. Thanks to global competition, demand is no longer certain for any business. Gone are the days of certainty, long product life cycles and loyal consumers. The overall environment today is dynamic. In such a situation, firms increasingly realise that understanding demand, planning demand and linking supply with demand pays. At the same time, if the supply chain forecast is wrong, the ramifications will be felt throughout the entire process.
This is why forecasting has assumed a significant importance, and more and more managers look to forecasting to reduce costs. Despite significant developments in the area of supply chain forecasting as well as IT, most organisations do a poor job of incorporating demand uncertainty into their production planning processes. Most often this is blamed on forecasting without realising the importance of selecting the appropriate forecasting technique. Managers need to identify first the firm-level variables, which cause variability in the supply chain. Once these are tabled, forecasting will be less uncertain in an uncertain environment.
Forecasting practices are characterised by some interesting insights about changes in techniques. Research indicates that in the 1980s, despite the growing availability of computer-based forecasting systems, companies continued to rely predominantly on subjective techniques. Since the mid 1990s companies have started using computer-based forecasting systems, yet surprisingly forecast accuracy has not improved even among those who use these models.
This gives rise to a range of questions: Are forecasts reviewed and agreed upon by key departments in the organisation? Are right statistical methods used in forecasting the demand for a product? What horizons and time period are used for both long and short-term forecasting? How are statistical and judgmental considerations combined? In a study conducted by the Great Lakes Institute of Management, Chennai, it was found that the most widely used method of forecasting is the sales force composite method. Causal and time series models have given way to rolling plans. With the changing nature of businesses and increasing complexity due to the changing nature of demand, this shift from quantitative to qualitative models is understandable. But what we found surprising was that even where causal and time series models would have been appropriate, IT-based sales force composites were used blindly. Forecasting is not owned as yet by any department, thus a consensus approach has yet to evolve leading to a budget-driven demand planning.
Not all demand is unpredictable: there are times when demand follows a predictable pattern. While auditing the forecasting processes of a lifestyle major, I found that the company used the time series technique for its vacuum cleaners and spare part requirement. The forecast error was so high that it gave up forecasting in favour of an ERP system where sales force composite forecasts were converted into rolling forecasts. This too did not meet with much success. To understand the problem we used data collated from Mumbai's Colaba market and found that the consumption data showed strong seasonality. No forecast would be accurate unless corrected for the seasonal trend and combined with the appropriate time series technique. Nor would the investment in an up-to-date information system be of much help.
Similarly, another company, a tractor major, for which we designed a forecasting model, had almost given up the causal method of forecasting and embraced the sales force composite method, even though the latter's accuracy remained a major worry. Tractor demand is closely related to what happens to agriculture. We had identified a causal model that was based on drivers of demand for tractors and provided a fair guide in planning sales. The company did not know how to convert causal forecasts into short-term forecasts for better operational planning and, thus, gave up scientific forecasting for judgmental methods.
On the other hand, agribusiness firms such as Bayer and Syngenta have been quite successful in this area. Along with their rolling plans, they also forecast the crop scenario for various regions. They use their sales forces to track changes in the cropping pattern, areas under different crops, procurement prices and rainfall. This data is then used to create an operational sales forecast on both a quarterly and a monthly basis. Not surprisingly Syngenta and Bayer have been able to minimise inventory in comparison to other players in the industry.
Indian firms seem to have lost direction. Their choice of forecasting methods seem to be dictated by supply chain requirements with little understanding of when, where, what and how to forecast. For example, we found that there is a tendency for small changes in customer demand to be amplified within a production distribution system. Upstream replenishment demand and physical shipments exceed the original order quantity. They are a result of orders moving up the supply chain levels, unplanned trade and promotion discounts, long lead-time and batch ordering.
Such business driven variability is further distorted as marketing and promotions create havoc with market data or demand trends. Most members of the supply chain stock during promotions and discounts, leading to a jump in demand. But each promotion is carried out in isolation vis-à-vis the rest of the organisation; and strategically to compete with key competitors. Keeping track of these spikes seems next to impossible, be it in the consumer durables sector or in an engineering company or for that matter agribusinesses. A forecaster sees the upward trend and forecasts high leading to inventory costs in the supply chain. Similar problems are posed when a full truck load becomes the norm due to the transport discount, and here again the jump in data can mislead forecasters. We also found that longer lead times meant higher demand amplification, poor forecast and excessive inventory cost.
In sum, firms often blame forecasts for the error when the real culprit is their own business practice. Forecasting methods can work when you are in apposition to track this business driven variability and then factor them into your forecast. Finally, it should be remembered that forecasting is an integrated exercise in which all levels of the supply chain are involved and are willing to share information which helps in increasing demand visibility within organisations as well increase the performance of forecast. IT thus is a critical tool. But, most firms show a significant level of dissatisfaction with the quality of IT.
We found that most firms lack extended enterprise functionality and open system architecture that can facilitate integration and collaboration and bring transparency across the supply chain. The current information systems are inward looking and miss linking across the boundaries of the organisation. Information systems in these companies also lack flexibility in adopting to the changing needs of the supply chain in terms of business models and processes. There is a lack of collaborative architecture in decision support software. These firms still use ordering-based information instead of flow-base. The lack of advance planning with process functionality hinders optimal supply chain allocations. Real-time communication between information systems transport and warehouse management systems and advance planning systems is absent. Information systems in these companies, thus, lack a modular, pen and internet-like architecture or "web-enabled ERP".
We found that firms that used forecasting successfully had developed not only cross-functional trust, but also cross-organisational trust with distributors and suppliers.
Thus, it is clear that forecasting as an exercise is more than using sophisticated techniques. These techniques will work effectively only when we create demand visibility across the supply chain. This calls for aligning marketing, promotions, discounts and other logistics decision with a clear purpose of creating demand visibility across the supply chain.
Rakesh Singh
Professor of economics & supply chain, Great Lakes Institute of Management, Chennai
Professor of economics & supply chain, Great Lakes Institute of Management, Chennai
