Judgemental Forecasting Method

Judgemental Forecasting Method Forecasting is a significant tool for many different sectors as it makes predictions on the future by looking at historical data, present data and the analysing of trends. However, some business forecasting is not done at a good level, as some business people confuse it with goals and planning. Forecasting, Goals and Planning, these three differ significantly, Forecasting is trying to calculate the future a specific as possible, by using historical data, present data and the analysing of trend, Goals for business is that the business would like to happen for them in the near future. Goals are usually done with lacking any planning or forecasting, as the business looks at their competitors and they either want to match them or exceed them in the market. Planning is looking at the forecasting and goals and deciding the best action that will make the business forecasting match their goals. As the business world is moving more into analysing data, forecasting is and will be a vital part of decision-making for the management team, as the forecasting can help with short term, medium term and long term forecasting. When a business has a lack of past data or the business is launching a new product, the business can still use forecasting, and they will use Judgement forecasting. Judgement forecasting is the use of opinion, intuitive judgment and subjective probability estimates. Judgment forecasting has few methods that can be used to get the best statistical analysis and there are Statistical surveys, Scenario building, Delphi methods, Technology forecasting and forecast by analogy. The Judgement forecasting has increasingly been recognised as a science, and over the years the quality of Judgement forecasting has been improving as the approach has been well structured and efficient. But it is important to understand that Judgement forecasting has not been perfected as it still has limitations. Judgment forecast depend on human cognition which has limitations, For example, a limited memory may render recent events more important than they actually are and may ignore momentous events from the more distant past; or a limited attention span may result in important information being missed, or a misunderstanding of causal relationships may lead to erroneous inference.1 This example shows that human memory can affect the judgment forecast in a negative way, and misunderstanding can lead to wishful thinking or optimistic view which can lead to faulty forecast, and in the case of launching a new product, the marketing and salesman teams will have an optimistic view for their lunch so they will not forecast its failure Beware of the enthusiasm of your marketing and sales colleagues2. In the case of judgment forecasting without any domain knowledge and only a set of time series data is used, getting a forecast will be very hard, as in the Hogath and makridais (1981) in their paper, where they have examined around 175 papers where there was judgment forecasting, they have approached a result of that quantitative models outperform judgmental forecasts3, in their research they have seen that judgment has been linked with systematic biases and errors, as some people were looking for patterns and linking together clues where there was none as the process was random. Judgment forecasting has been compared to many different kinds of forecasting such as statistical methods, and many different types of research conclude different findings of the accuracy of the two methods. In the paper of Lawrence (1985) and (1986) where the paper compares the accuracy of quantities model and judgment forecasting, the paper has come to a conclusion that demonstrated judgmental forecasting to be at least as accurate as statistical techniques4, also in the paper show that the standard deviation of the error of the statistical method was greater than the judgment forecast error. The paper also shows that if judgment forecasting was added in the statistical method, better sets of forecasting can be predicted and the level of error would decrease. In the study by Makridakis S and Winkler R (1983) it shows that there are few ways to combine the judgement and statistical forecasting. In the study it says that there is two way to join the two forecasting methods, the first is Concurrent Incorporation where to get the final forecasting both methods will have to be used to get the averaging procedure. The second way is a Posterior Incorporation which includes the judgmental revision of statistically derived forecasts5   this second way tries to improve forecasting by allowing the judgement forecasting to see and access the results of the statistical forecasting. After many years of research in the area of forecasting, Judgment forecasting improves when greater domain knowledge and more up to date information included, therefore by using this new information, judgment approach can then be adjusted and producing an improved forecast. By using a well structured and systematic approach, it helps to decrease the undesirable effects of the limitations of the forecast. By well structuring the approach it will make the forecasting task clear, and a good understanding of the structure is important to avoid unclear and uncertain terms. The method that is well structured that can be used for the judgment forecasting is the Delphi methods. The Delphi method is the use of experts opinions and judgment in the specific field to predict the expectation in that field. The Delphi method is respect method as it only looks at the opinions of the experts in their field and allows them to be anonymous at all time, therefore there is not influenced by their social and political pressure in their prediction, and all experts opinions are weighted equally so no one prediction is superior to another. But like any other approach, the Delphi method also has its limitations, the method is time-consuming, therefore, the experts might be discouraged to join the study or they will not contribute fully at all time of the approach. Adding domain knowledge to the judgement forecasting can be used fully for the prediction. The knowledge of the time series and further information which explains the historical performance of the series can have a minor influence on the forecast or a huge impact on the variable of the data. The domain knowledge represents the un-modelled module of the series. The un-modelled module is very important as it can be included into the statistical forecast to reach better results for the forecast. Many studies have been looking at judgement forecasting with the addition of domain knowledge, a study by Brown (1996) which looked at earning per share forecasting. The study shows that the forecasting of the management team was more accurate than the analysts predictions and the statistical model forecasting. In the study, it shows that the inside information which is the domain knowledge of the firm lead to the accuracy of the management team forecast. In the study, it showed that it did not matter if the statistical model was complex or simple as the management team and analysts got a higher accuracy level because of the domain knowledge the management team holds. In a study by Sanders (1992) where it compared the preference of judgement methods to statistical forecasting, the study compared both methods by the use of an artificial time series. The study looked at 38 business students, the students were thought some different ways of statistical and judgement forecasting and every student had two-time series and past data. The task for the students was to use all the information they had to forecast the next 12 steps ahead. The students were given one week to produce their judgement forecasting, then they were given statistical forecasting of the series, and then they were asked to review their forecast and do any adjustment if needed. The study has used the mean absolute percentage error to assess the forecasting results, and the mean percentage error was applied to calculate the level of bias in the forecast. The results of the study have similar results as the past studies did, as statistical methods outperformed judgment forecasting in all -time series but not the low noise step function. And the more complex the data pattern got the worse the judgement forecast became. The study clearly shows that the statistical methods had better forecasting in the high noise level data, and an increase in noise level has worsened off the judgement forecasting, the study says this is due because as the high noise increases it becomes harder for an individual to detect any kind of patterns. While judgement forecasting didnt perform well during a high noise, it did significantly well in the low noise function. Looking at the bias in the study, it shows that at a low noise series the judgement revision bias is low in the series, while for a high noise series it increases the bias in the series. The main point of the study by Sanders (1992) is that judgement amendments with statistical methods can have great advantage for a low noise series with a specific data patterns, and it will do better when statistical method are applied blindly to a time series, also at a low noise series the judgement revision bias is low in the series, but in a high noise series the judgment forecasting is not the right approach comparing to a statistical forecasting and in some instances the bias level in the judgement forecasting was greater than the statistical forecasting in a high noise series. Sanders approach of the judgement forecasting is not overwhelm approved in the forecasting filed, as it has many critics wondering about its efficiency, as the sanders approach for judgment forecasting does not use the experts opinions on the field that is going to be forecasted but uses the opinion and judgment of normal people who may have not have studied the field and have a small knowledge about it, therefore, there judgement would not be the best to use to create a prediction from it. Judgemental forecasting is an important tool in the business today but it has to be used right, as some business people confuse it with goals and planning. When doing a judgmental forecasting the aims and the purpose of the forecasting have to be clear and well structured to get better results. But like any forecasting method, Judgemental forecasting has its limitations and it is up to the person who is performing the forecast to make sure they are at a minimum. To get a better prediction it is important to try and increase the domain knowledge of the series as it has been shown in the Brown (1996) study, as the management team outperformed the statistical analysis due to the inside information of the firm and because they are the experts in that field. Also to improve the judgement forecasting as it has been shown in the Sanders (1992) have found if judgment forecasting is done with a revision of statistical methods, the forecast can be more accurate in a low noise series and with a less level of bias. Judgmental forecasting is not a perfect method to predict the outcome of a specific time series but it is a good point to start. References: Rob J Hyndman, (2013) Chapter: Judgmental Forecasts: Beware of limitations. Book: Forecasting: principles and practice R. Fildes and P. Goodwin (2007b). Good and bad judgment in forecasting: lessons from four companies. Foresight: The International Journal of Applied Forecasting 8, 5-10. Hogarth Makridakis, 1981 R.M. Hogarth, S. Makridakis Forecasting and planning: An evaluation Management Science, 27 (1981), pp. 115-138 Lawrence et al., 1986M. Lawrence, R. Edmundson, M. OConnor The accuracy of combining judgmental and statistical forecasts, Management Science, 32 (1986), pp. 1521-1532 Makridakis S and Winkler R (1983) Averages of forecasts: Some empirical results. Mgmt Sci. 29, 987-996 Sanders, 1992  N.R. Sanders  Ãƒâ€šÃ‚   Accuracy of judgmental forecasts: A comparison   Omega, 20 (3) (1992), p. 353 Brown, 1996 L. Brown Analyst forecasting errors and their implications for security analysts: An alternative perspective Financial Analysts Journal, 52 (1) (1996), pp. 40-47 Lawrence, M., Goodwin, P., OConnor, M., ÃÆ'-nkal, D. (2006). Judgmental forecasting: A review of progress over the last 25years. International Journal of Forecasting, 22(3), 493-518.