Fichas
7. Forecasting
Fichas del conjunto (22)
- 1Factors of forecastingForecasting depends on the length of the period that is being forecast. The impact of trends should influence decision-making, for example, capacity operations. Trends, seasonal influences, variations and peaks can challenge a forecast.
- 2Simple overview of methods
- 3Explain naive forecastsA naïve forecast employs no specific forecasting model. It simply takes the previous period as a starting point, looks at moving averages and assumes that seasonal items will be the same as in the last period.
It is mostly used only for comparison with the forecasts generated by better (sophisticated) techniques.
• A demand with trend graph shows the increase or decrease of demand per interval of time. It is not necessarily a first order (‘straight’) line. - 4Variations and operations strategies to implementVariations could be due to patterns of consumption and use, and the purchase behavior of consumers. For example, the perishability of products, the weather and holidays. These are regular variations.
Based on this information, operations could employ:
1. Chase strategy: Flexibility and responsiveness
2. Level strategy: Holding inventory - 5Explain the causes of peaks in demandPeaks in demand could be due to ‘regular’ variations or ‘one-off’ demands, such as events. These are challenging to predict and estimated based on an educated guess. This affects the levels of inventory (except for services) and the responsiveness of supply in case of excess of educated guesses.
- 6Explain aggregate demandAggregate demand uses aggregation strata. The loss of detail implies greater accuracy but does not solve anything. It is only useful when more or less the same resources are used. However, it is better to study the overall demand market than to predict the individual market shares.
- 7List the components of demand• Components of demand:
1. Average demand for a given period of time
2. Trend
3. Seasonal / cyclical elements
4. Random variation
5. Autocorrelation
Because of these influences the formula for naïve forecasts does not hold (Ft = Dt-1) - 8Explain what happens with increasing time-span for forecasting and how to deal with it• With increasing time-span for forecasting, there will be reduced accuracy and increased uncertainty across three zones; forecasting, scenarios and hope.
• Hence, businesses should prepare to react in the short-term and to competitors and anticipate driving forces mid- to long-term. - 9List and explain two ways of averaging demandThe simple moving average ‘removes’ random fluctuations, whereby the demand neither steadily increases or declines.
This is achieved by adding recent closing prices and then dividing by the number of time periods in the calculation average.
• The weighted moving average values more recent observations dependent on their weight for each period. - 10Define exponential smoothingExponential smoothing is the same as the moving average. The smoothing follows variations but is dependent on a smoothing constant (alpha). Exponential models may include trend, so the result may follow the actual demand more closely. However, these numbers are still based on estimating trend.
- 11Explain how exponential smoothing works• During exponential smoothing, each increment in the past is decreased by (1-alpha); so the most recent weighting would be alpha(1-alpha), the older weighting alpha(1-alpha)1, etc. This method also looks at the previous period, comparing the forecast with the actual demand.
• Take the forecast of the previous period and compare the actual demand with the forecast period to find the ‘error’. Apply a smoothing constant that is between 0.0 and 0.1). Then add the smoothed error previous period to the forecast of the previous period.
• To include the trend, calculate the forecast as smoothing and add the trend. - 12Evaluate the accuracy of exponential modelsExponential models are surprisingly accurate and relatively easy to formulate. The user will usually easily understand the model and little computation or computing requirements are required. A test for accuracy is thereafter easy to perform.
- 13Define and evaluate regression analysis• The aim of a linear regression is to determine the trend and minimize the distance between individual values and the trend. The demand is presumed to follow: Y = a + bx.
• A regression analysis is more accurate than methods based on averaging. It assumes that there is a trend that (here) is linear, by reducing variations around the trend. However, inclusion of seasonal trend is possible. - 14Downsides of forecastingForecasting is often not accurate, as it is based on historical data and may contain measurement errors and bias. Some methods display lagging. Errors can be estimated using the standard error of estimate or mean absolute deviation.
- 15Causes of errors in forecasting1. Measurement: Data is often inaccurate, although sometimes presented otherwise.
2. Calculations: There may be a bias, sometimes called systematic errors, towards certain variables and trends. Calculation errors can also be random, for example through statistical variations.
3. Different methods: Different outcomes
4. The effect of smoothing constants: A higher value is closer to the actual demand. The inclusion trend is necessary if increasing or decreasing.
• Errors can be calculated using the mean absolute deviation or the standard error of estimate - 16Gathering data for forecasts⎝ Data for forecasting can be gathered using sophisticated methods, both qualitatively and quantitatively. Although, a lot of information is publically available. Yet, the reliability for forecasting is often questionable since it is reliant on the behavior of customers and their response to products.
- 17Data for forecasting: Three factors of demand increase1. Market growth (increased sales, increased market size)
2. Stealing share (increased sales, same market size)
3. Forward buying (same sales, same market size)
In managing demand, the timing of changes in promotion and pricing are important. - 18Data for forecasting: The purpose of market research + three typesMarket research gathers information about markets or customers and is a component of business strategy. It includes social and opinion research and might use statistical data about trends, demographics, etc. Market research focuses on marketing processes.
• Panel consensus
• Case based reasoning
• Delphi panel - 19Data for forecasting: Panel consensusPanel consensus: Pooling forecast or economic studies; panel consensus increases reliability. It could also indicate open meetings, focus groups etc. to extract relevant information.
- 20Data for forecasting: Case-based reasoningAlso called historical analogy; reasoning is based on an existing case and parallels are drawn.
- 21Data for forecasting: Delphi panelA Delphi Panel is based on the consultation of experts. The method is the following:
1. Choose experts
2. Questionnaire or survey
3. Summarize results and redistribute with new questions
4. Summarize again, refining forecasts and conditions, and formulate new questions
5. Repeat until there is consensus; then elaborate. - 22Data for forecasting: What to look out for and tips•Forecasting methods can contain error, and with more detail comes less accuracy. Hence, call for aggregate demand or planning to improve accuracy.
• The simple way out is probably collecting data from publically available sources such as news sites and papers, statistical publications or the government. Other ways include using common sense and piecing that information together.
Información
- Descripción
- (Operations Management) Business and Management 2A Fichas sobre 7. Forecasting, creado por Nina Olsson el 23/07/2018.
- Creado por
- Nina Olsson
- Visitas
- 0
- Valoración