Estimate and Projection Show What is an estimate? An estimate is a value that is inferred for a population based on data collected from a sample of units from that population. Estimation is a technique that systematically adjusts the sample data to determine an estimated value for the population. For example, if our sample data shows that 51% of the sample are female, then the population value will be estimated to be 51% (as estimation is based on the assumption that the sample is representative of the population). An estimate is not a guess, it is a value based on sampled data which has been adjusted using statistical estimation procedures. What is a projection? A projection indicates what the future changes in a population would be if the assumptions about future trends actually occur. These assumptions are often based on patterns of change which have previously occurred. For example: Data collected about the total number of store locations for a retail chain over three years show an increase from 8 stores in first year, to 12 stores in the second year, to 18 stores in the third year. It could therefore be projected that if the chain continues to expand following the same pattern of increasing by half (50%) each year there will be 27 stores after the fourth year. A projection is not making a prediction or forecast about what is going to happen, it is indicating what would happen if the assumptions which underpin the projection actually occur. Comparison of Projections and Forecasts
How do estimates and projections differ? An estimate is a statistic about a whole population for a previous reference period which is based on data from a sample of the population, whereas a projection is a statistic indicating what a value would be if the assumptions about future trends hold true (often drawing upon past movements in a population as a guide for the assumptions). Return to Statistical Language Homepage Further information
Forecasting is the process of making predictions of the future based on past and present data. This is most commonly by analysis of trends. A commonplace example might be estimation of some variable of interest at some specified future date. Prediction is a similar, but more general term. Both might refer to formal statistical methods employing time series, cross-sectional or longitudinal data, or alternatively to less formal judgmental methods. Usage can differ between areas of application: for example, in hydrology, the terms “forecast” and “forecasting” are sometimes reserved for estimates of values at certain specific future times, while the term “prediction” is used for more general estimates, such as the number of times floods will occur over a long period. Risk and uncertainty are central to forecasting and prediction; it is generally considered good practice to indicate the degree of uncertainty attached to specific forecasts. In any case, the data must be up to date in order for the forecast to be as accurate as possible. In some cases, the data used to predict the variable of interest is itself forecasted.[1] As discussed in the previous chapter, functional strategies need to be aligned and supportive to the higher level corporate strategy of the organization. One of these functional areas is marketing. Creating marketing strategy is not a single event, nor is the implementation of marketing strategy something only the marketing department has to worry about. When the strategy is implemented, the rest of the company must be poised to deal with the consequences. An important component in this implementation is the sales forecast, which is the estimate of how much the company will actually sell. The rest of the company must then be geared up (or down) to meet that demand. In this module, we explore forecasting in more detail, as there are many choices that can be made in developing a forecast. Accuracy is important when it comes to forecasts. If executives overestimate the demand for a product, the company could end up spending money on manufacturing, distribution, and servicing activities it won’t need. Data Impact, a software developer, recently overestimated the demand for one of its new products. Because the sales of the product didn’t meet projections, Data Impact lacked the cash available to pay its vendors, utility providers, and others. Employees had to be terminated in many areas of the firm to trim costs. Underestimating demand can be just as devastating. When a company introduces a new product, it launches marketing and sales campaigns to create demand for it. But if the company isn’t ready to deliver the amount of the product the market demands, then other competitors can steal sales the firm might otherwise have captured. Sony’s inability to deliver the e-Reader in sufficient numbers made Amazon’s Kindle more readily accepted in the market; other features then gave the Kindle an advantage that Sony is finding difficult to overcome. The firm has to do more than just forecast the company’s sales. The process can be complex, because how much the company can sell will depend on many factors such as how much the product will cost, how competitors will react, and so forth. Each of these factors has to be taken into account in order to determine how much the company is likely to sell. As factors change, the forecast has to change as well. Thus, a sales forecast is actually a composite of a number of estimates and has to be dynamic as those other estimates change. A common first step is to determine market potential, or total industry-wide sales expected in a particular product category for the time period of interest. (The time period of interest might be the coming year, quarter, month, or some other time period.) Some marketing research companies, such as Nielsen, Gartner, and others, estimate the market potential for various products and then sell that research to companies that produce those products. Once the firm has an idea of the market potential, the company’s sales potential can be estimated. A firm’s sales potential is the maximum total revenue it hopes to generate from a product or the number of units of it the company can hope to sell. The sales potential for the product is typically represented as a percentage of its market potential and equivalent to the company’s estimated maximum market share for the time period. In your budget, you’ll want to forecast the revenues earned from the product against the market potential, as well as against the product’s costs.[2] Forecasting HorizonsLong term forecasting tends to be completed at high levels in the organization. The time frame is generally considered longer than 2 years into the future. Detailed knowledge about the products and markets are required due to the high degree of uncertainty. This is commonly the case with new products entering the market, emerging new technologies and opening new facilities. Often no historical data is available. Medium term forecasting tends to be several months up to 2 years into the future and is referred to as intermediate term. Both quantitative and qualitative forecasting may be used in this time frame. Short term forecasting is daily up to months in the future. These forecasts are used for operational decision making such as inventory planning, ordering and scheduling of the workforce. Usually quantitative methods such as time series analysis are used in this time frame. Categories of Forecasting MethodsQualitative ForecastingQualitative forecasting techniques are subjective, based on the opinion and judgment of consumers and experts; they are appropriate when past data are not available. They are usually applied to intermediate- or long-range decisions. In the following, we discuss some examples of qualitative forecasting techniques: Executive Judgement (Top Down)Groups of high-level executives will often assume responsibility for the forecast. They will collaborate to examine market data and look at future trends for the business. Often, they will use statistical models as well as market experts to arrive at a forecast. Sales Force Opinions (Bottom up)The sales force in a business are those persons most close to the customers. Their opinions are of high value. Often the sales force personnel are asked to give their future projections for their area or territory. Once all of those are reviewed, they may be combined to form an overall forecast for district or region. Delphi MethodThis method was created by the Rand Corporation in the 1950s. A group of experts are recruited to participate in a forecast. The administrator of the forecast will send out a series of questionnaires and ask for inputs and justifications. These responses will be collated and sent out again to allow respondents to evaluate and adjust their answers. A key aspect of the Delphi method is that the responses are anonymous, respondents do not have any knowledge about what information has come from which sources. That permits all of the opinions to be given equal consideration. The set of questionnaires will go back and forth multiple times until a forecast is agreed upon. Market SurveysSome organizations will employ market research firms to solicit information from consumers regarding opinions on products and future purchasing plans. Quantitative ForecastingQuantitative forecasting models are used to forecast future data as a function of past data. They are appropriate to use when past numerical data is available and when it is reasonable to assume that some of the patterns in the data are expected to continue into the future. These methods are usually applied to short- or intermediate-range decisions. Some examples of quantitative forecasting methods are causal (econometric) forecasting methods, last period demand (naïve), simple and weighted N-Period moving averages and simple exponential smoothing, which are categorizes as time-series methods. Quantitative forecasting models are often judged against each other by comparing their accuracy performance measures. Some of these measures include Mean Absolute Deviation (MAD), Mean Squared Error (MSE), and Mean Absolute Percentage Error (MAPE). We will elaborate on some of these forecasting methods and the accuracy measure in the following sections.[3] Causal (Econometric) Forecasting Methods (Degree)Some forecasting methods try to identify the underlying factors that might influence the variable that is being forecast. For example, including information about climate patterns might improve the ability of a model to predict umbrella sales. Forecasting models often take account of regular seasonal variations. In addition to climate, such variations can also be due to holidays and customs: for example, one might predict that sales of college football apparel will be higher during the football season than during the off-season. Several informal methods used in causal forecasting do not rely solely on the output of mathematical algorithms, but instead use the judgment of the forecaster. Some forecasts take account of past relationships between variables: if one variable has, for example, been approximately linearly related to another for a long period of time, it may be appropriate to extrapolate such a relationship into the future, without necessarily understanding the reasons for the relationship. One of the most famous causal models is regression analysis. In statistical modeling, regression analysis is a set of statistical processes for estimating the relationships among variables. It includes many techniques for modeling and analyzing several variables, when the focus is on the relationship between a dependent variable and one or more independent variables (or ‘predictors’). More specifically, regression analysis helps one understand how the typical value of the dependent variable (or ‘criterion variable’) changes when any one of the independent variables is varied, while the other independent variables are held fixed. Figure 3.1: Example of regression analysis.
Demand Patterns When we plot our historical product demand, the following patterns can often be found: Trend – A trend is consistent upward or downward movement of the demand. This may be related to the product’s life cycle. Cycle – A cycle is a pattern in the data that tends to last more than one year in duration. Often, they are related to events such as interest rates, the political climate, consumer confidence or other market factors. Seasonal – Many products have a seasonal pattern, generally predictable changes in demand that are recurring every year. Fashion products and sporting goods are heavily influenced by seasonality. Irregular variations – Often demand can be influenced by an event or series of events that are not expected to be repeated in the future. Examples might include an extreme weather event, a strike at a college campus, or a power outage. Random variations – Random variations are the unexplained variations in demand that remain after all other factors are considered. Often this is referred to as noise. Figure 3.2: Diagram of trend, cyclical, and seasonal demand patterns.Time Series MethodsTime series methods use historical data as the basis of estimating future outcomes. A time series is a series of data points indexed (or listed or graphed) in time order. Most commonly, a time series is a sequence taken at successive equally spaced points in time. Thus, it is a sequence of discrete-time data. Examples of time series are heights of ocean tides, counts of sunspots, and the daily closing value of the Dow Jones Industrial Average. Time series are very frequently plotted via line charts. Time series are used in statistics, signal processing, pattern recognition, econometrics, mathematical finance, weather forecasting, earthquake prediction, electroencephalography, control engineering, astronomy, communications engineering, and largely in any domain of applied science and engineering which involves temporal measurements.[4] In the following, we will elaborate more on some of the simpler time-series methods and go over some numerical examples. Naïve Method Simple Moving Average Example Some relevant notation: Using the following table, calculate the forecast for period 5 based on a 3-period moving average. Solution Weighted Moving Average Example
Solution Note that if the sum of all the weights were not equal to 1, this number above had to be divided by the sum of all the weights to get the correct weighted moving average. Exponential Smoothing There are two versions of the same formula for calculating the exponential smoothing. Here is version #1: Ft = (1 – α) Ft-1 + α Dt-1 Note that α is a coefficient between 0 and 1 For this method to work, we need to have the forecast for the previous period. This forecast is assumed to be obtained using the same exponential smoothing method. If there were no previous period forecast for any of the past periods, we will need to initiate this method of forecasting by making some assumptions. This is explained in the next example. Example
In this example, period 5 is our next period for which we are looking for a forecast. In order to have that, we will need the forecast for the last period (i.e., period 4). But there is no forecast given for period 4. Thus, we will need to calculate the forecast for period 4 first. However, a similar issue exists for period 4, since we do not have the forecast for period 3. So, we need to go back for one more period and calculate the forecast for period 3. As you see, this will take us all the way back to period 1. Because there is no period before period 1, we will need to make some assumption for the forecast of period 1. One common assumption is to use the same demand of period 1 for its forecast. This will give us a forecast to start, and then, we can calculate the forecast for period 2 from there. Let’s see how the calculations work out: If α = 0.3 (assume it is given here, but in practice, this value needs to be selected properly to produce the most accurate forecast) Assume F1 = D1, which is equal to 42. Then, calculate F2 = (1 – α) F1+ α D1 = (1 – 0.3) x 42 + 0.3 x 42 = 42 Next, calculate F3 = (1 – α) F2+ α D2 = (1 – 0.3) x 42 + 0.3 x 37 = 40.5 And similarly, F4 = (1 – α) F3+ α D3 = (1 – 0.3) x 40.5 + 0.3 x 34 = 38.55 And finally, F5 = (1 – α) F4+ α D4 = (1 – 0.3) x 38.55 + 0.3 x 40 = 38.985 Figure 3.3: Solution for Exponential Smoothing Version 1Accessible format for Figure 3.3 Here is version #2: Ft = Ft-1 + α(Dt-1 – Ft-1) Example Accessible format for Figure 3.4 Seasonal Index Example
Using these calculated indices, we can forecast the demand for next year based on the expected annual demand for the next year. Let’s say a firm has estimated that next year annual demand will be 2500 units.
Forecast Accuracy MeasuresIn this section, we will calculate forecast accuracy measures such as Mean Absolute Deviation (MAD), Mean Squared Error (MSE), and Mean Absolute Percentage Error (MAPE). We will explain the calculations using the next example. Example
Here are what need to do: Step 1: Calculate the error as et = Dt – Ft (the difference between the actual demand and the forecast) for any period t and enter the values in the table above. Step 2: Calculate the absolute value of the errors calculated in step 1 [i.e., Abs (et)], and enter the values in the table above. Step 3: Calculate the squared error (i.e., et2) for each period and enter the values in the table above. Step 4: Calculate [Abs (et) / Dt] x 100% for each period and enter the value under its column in the table above. Solution
MAD = The average of what we calculated in step 2 (i.e., the average of all the absolute error values) = (5 + 6 + 7 + 6) / 4 = 24 / 4 = 6 MSE = The average of what we calculated in step 3 (i.e., the average of all the squared error values) = (25 + 36 + 49 + 36) / 4 = 146/4 = 36.5 MAPE = The average of what we calculated in step 4 = (7.94% + 10.17% + 12.96% + 9.23%) / 4 = 40.3/4 = 10.075% End of Chapter ProblemsProblem #1Below are monthly sales of light bulbs from the lighting store.
.
Solution
Problem #2Demand for aqua fit classes at a large Community Centre are as follows for the first six weeks of this year.
.
Solution
Problem #3Sales of a new shed has grown steadily from the large farm supply store. Below are the sales from the past five years. Forecast the sales for 2018 and 2019 using exponential smoothing with an alpha of .4. In 2015, the forecast was 360. Calculate a forecast for 2016 through to 2020.
.
Problem #4Below is the actual demand for X-rays at a medical clinic. Two methods of forecasting were used. Calculate a mean absolute deviation for each forecast method. Which one is more accurate?
Solution
Which of these terms refers to forecasting what will happen before it actually happens?To prognosticate means to predict something or at least hint at what will happen in the future.
What are different methods of forecasting?There are three basic types—qualitative techniques, time series analysis and projection, and causal models.
What are the four types of forecasting?Four common types of forecasting models. Time series model.. Econometric model.. Judgmental forecasting model.. The Delphi method.. What is forecasting based on?Forecasts are based on opinions, intuition, guesses, as well as on facts, figures, and other relevant data. All of the factors that go into creating a forecast reflect to some extent what happened with the business in the past and what is considered likely to occur in the future.
|