Forecasting Flashcards
Forecast and examples of it being used for
Prediction of future events used for planning purposes etc product, labour, demand
Planning
Making management decisions on how to deploy resources to respond to demand forecast
Forecasts are based on (4)
multiple types of data, mathematical models, expert opinion, historical data
Forecast is used for (2)
process etc bottlenecks and supply chain management
Time/Demand Series
Repeated observations of demand for a product/service in their order of appearance
Horizontal
Fluctuation of data around a constant mean
Trend
Systematic increase or decrease in the mean over time
Seasonal
Repeatable pattern of increase or decrease in demand, depending on time, day, week, month, season
Cyclical and what is it caused by
Less predictable gradual increase/decrease in demand over longer periods of time (years, decades)
- life cycle of product or economic recession/inflation
Random
Un-forecastable variation in demand (lots of variability)
Outliers
fluctuations in data that do not reflect or resemble overall pattern
Manage Demand (5)
Complementary Service, Promotional Pricing, Prescheduled Appointments, Revenue Managing, Backlogs/Backorders/Stockouts
Complementary Service
same resources, different demand cycles (Assiniboine Park)
Promotional Pricing
increase demand, shift to new period (clear excess stock and attract buyers)
Prescheduled Appointments
level demand based on capacity (balance how much you can accept)
Revenue Management
adjust prices in real life time based on demand
Backlogs
accumulate orders for future delivery, decrease service level and risk of losing customers
backorders
orders that cannot be filled when demanded but filled later
Stockouts
customer goes else where as order cannot be fulfilled
Key forecasting decisions (3)
What are inputs
What are you predicting
What technique should you use
Forecast Inputs (6)
History of Past Demand, Notes Explaining Past Demand, Past Forecasts, consumer research, planned promotions, Inputs from Partners
CPFR
Collaborative Planning, Forecasting & Replenishment
CPFR what does it require and do?
collaboration with suppliers, independent forecasts generated & compared, adjusted until consensus (everyone has same prediction)
What are you trying to predict (2)
what is aggregation?
What is best way to predict revenue?
1.individual/family products 2.Units of measurements
2. cluster of similar products/services so company can make better forecasts
3. find units forecast then multiply by price
What techniques used (3)
Judgment, Causal, Time-Series Analysis
Judgement (opinions and subjectives -> quantitative) (4)
sales force estimates
executive opinion
market research
Delphi -> consensus of group but group remains anonymous
Qualitative Benefits
subjective, variety of information, does not require numerical data
Qualitative Downside
results biased or conflicting
Quantitative Benefits
objective, volume of information, do not rely on individuals
Quantitative Downside
data not available, models too simplistic
Time Series Method
predictions based on historical data, dependent variable. Past can predict future
Naive Method
appropriate for and what pattern
sensitive to?
Forecast for next period equals demand for most recent observed
short-term forecasts and horizontal trend
sensitive to random variation
Simple Moving Average and etc
smooths out
Forecast for next period equals average demand for n most recent periods etc 2-period moving means average of 2 previous weeks
random variation
Weighted Moving Average
weights given but most recent has
forecast for next period equals average demand for n most recent periods and each observation of demand has its own weight.
most weights
Exponential Smoothing
3 data points required? and more weight to?
weighted moving average assigning differing levels of weight to recent demand compared to older historical data
last period forecast, demand and smoothing parameter
Exponential smoothing factor to previous demand, (1-a) applied to forecast
Forecast Error
Observed demand - forecast
Cumulative Sum if Forecast Errors and evaluates
Assesses total errors in forecasts over time, presence and detection of bias
If forecast is consistently lower than demand then
If forecast is consistently higher than demand then
CFE is highly positive
CFE is highly negative
Mean Squared Error
on average how close forecast is to demand, magnify large errors
Mean Absolute Deviation
magnitude of error, does not reveal directional bias
Mean Absolute Percentage Error
study magnitude of error relative to demand
