Price data forecasting

An additional complication in formulating the objective function is the quantification of uncertainty. Economic objective functions are generally very sensitive to the prices used for feeds, raw materials, and energy, and also to estimates of project capital cost. These costs and prices are forecasts or estimates and are usually subject to substantial error. Cost estimation and price forecasting are discussed in Sections 6.3 and 6.4. There may also be uncertainty in the decision variables, either from variation in the plant inputs, variations introduced by unsteady plant operation, or imprecision in the design data and the constraint equations. Optimization under uncertainty is a specialized subject in its own right and is beyond the scope of this book. See Chapter 5 of Diwekar (2003) for a good introduction to the subject. 

Many companies can be hired as consultants to provide economic and marketing information, or allow access to such information on a subscription basis. The information provided generally includes market surveys and technical and economic analyses of competing technologies, as well as price data and forecasts. There is not room here to list all of these companies, but some of the most widely used are... 

Most price forecasts are based on an analysis of historic price data. Several methods are used, as illustrated in Figure 6.4. The simplest method is to use the current price, as in Figure 6.4a, but this is unsatisfactory for most commodities. Linear regression of... 

Cost Indices The value of money will change because of inflation and deflation. Hence cost data can be accurate only at the time when they are obtained and soon go out of date. Data from cost records of equipment and projects purchased in the past may be converted to present-day values by means of a cost index. The present cost of the item is found by multiplying the historical cost by the ratio of the present cost index divided oy the index applicable at the previous date. Ideally each cost item affected by inflation should be forecast separately. Labor costs, construction costs, raw-materials and energy prices, and product prices all change at different rates. Composite indices are derived by adding weighted fractions of the component indices. Most cost indices represent national averages, and local values may differ considerably. 

All cost-estimating methods use historical data and are themselves forecasts of future costs. The prices of the materials of construction and the costs of labor are subject to inflation. ome method has to be used to update old cost data for use in estimating at the design stage and to forecast the future construction cost of the plant. 

Purvin and Gertz Provides quarterly forecasts of oil, gas, and fuel prices that are widely used in the oil industry. They have a 10-year archive of historic data and forecast prices of most fuel products as well as crude oils on U.S., N.W. Europe, Middle East, and Asia bases. 

Chemical Market Associates Inc. (CMAI) Maintains a large archive of historic data and future price forecasts for 70 commodity chemicals, including multiple grades, U.S., N.W Europe, Middle East, N.E., and S.E. Asia. Spot and contract prices are given for some compounds, and in some cases margins are also estimated by formula. 

Prices for some of the more common commodity chemicals are sometimes given in process economics textbooks. These prices are usually single data points rather than forecasts. They are suitable only for undergraduate design projects. 

Using projected market and sales data. Considerable effort must be expended to make a sales forecast and to establish a selling price. There is much more art than science in this task. 

In many cases the values of the data coefficients are obtained by statistical estimation procedures on past figures, as in the case of sales forecasts, price estimates, and cost data. These estimates, in general, may not be very accurate. If we can identify which of the parameters affect the objective value most, then we can obtain better estimates of these parameters. This will increase the reliability of our model and the solution. 

Consider the quick response (QR) approach. Under this scheme, the manufacturer has to receive the order only four months in advance. The retailer can collect data regarding sales of similar products at points closer to the upcoming season before placing an order. Intuitively, the ability to order closer to the season increases the possibility that more recent trend information or economic conditions can be used to better forecast demand. For this example, assume that data regarding demand for similar product enables the retailer to further refine the demand distribution estimates. What will be the impact of QRoii the retailer (For now, assume manufacturer prices remain the same.)... 

Market-driven demand management utilizes data from market and channel sources to sense, shape, and translate demand requirements into an actionable demand response bidirectionally from market to market. A true market-driven forecast is an unconstrained view, or a best estimate of market demand based on channel data. Demand shaping is based on campaigns to combine price, new product launches, trade and sales promotions and incentives, advertising, and marketing programs to impact what and how much customers will buy. 

Implementing FVA into the demand management process requires that forecasts be recorded and saved before and after each cycle. Having the capabilities to store forecast history by a stream of activities (e.g., consensus forecast adjustments, managerial overrides, price lift calculations, etc.) is critical to measuring the value-added, or non-value-added, contribution to the overall process. Utilizing the statistical baseline forecast as the default is the key to establishing a benchmark to measure the effectiveness of all the touch points in the process. Unfortunately, few companies capture the appropriate data, or the level of detail on a historical basis, to conduct FVA. This is an opportunity. 

A hypothetical product growth curve is illustrated in Figure 6.6. Clearly if the data available show exponential growth care must be taken in projecting over a long period. Techniques have been developed to modify the simple extrapolation but require extensive data. The projection of past data can be improved by combining with detailed user surveys of major product outlets to provide a more rational basis for estimated demand. All methods of assessing future markets assume the absence of major perturbations, and events such as the OPEC actions on oil prices in 1973 and 1979 can make nonsense of any forecasts. 

In forecasting price trends for established products the available data on prices and tonnages sold again forms the basis for extrapolation. Published information on many chemicals (e.g. US Tariff Commission reports, reports of Office Central des Statistiques de la CEE) enable a plot of log (price—adjusted to constant money values) against log (cumulative production) to be drawn. 

Having obtained forecasts of potential market share and price, the annual cash flow position for a selected plant size can be calculated using the capital estimate and operating cost estimates prepared from the literature or research data. The cash flows can then be discounted at the cost of capital to determine the NPV and the DCF rate can be estimated. The NPV and DCF for different plant scales can be calculated to determine the optimum size of plant and the competitive processes compared over expected project life. 

Projections of marketing data, including gross product demand, sales forecasts, and prices net of discounts. 

Semi-automatic software These are moderately priced but they require the user to have some basic knowledge of forecasting principles and fechniques. Here, fhe user has to select an appropriate forecasting technique based on the analysis of time series data. The software will then compute the optimal parameters for the chosen method using some measure of forecasf error. It also gives the forecasts and all the statistics, such as MAD, MAPE, MSE, Bias, etc. The software makes no recommendation as to which forecasting technique is appropriate for the given data. 

---