Absolute errors, 103
Absolute percentage errors (APEs), 108
Accuracy, 102–107
Actual inventory management systems, 18
Advertising data, 74–75
Anecdotal pareto principle, 81
ARIMA (Autoregressive Integrated Moving Average) models
autocorrelation and partial autocorrelation, 66–68
autoregression, 61–62
integration, 62–64
moving averages, 64–66
Audi A3, 40
Autocorrelation function (ACF), 66–68
Autoregression (AR), 61–62
autocorrelation and partial autocorrelation functions, 67
The Bank of England (BoE), 14–15
Bias, 102–107
Biased forecast, 103
Big data, 74
“black box” argument, 92
Bottom-up forecasting, 137–139
Box–Jenkins models. See ARIMA models
“Brute force” technique, 65
Bureau of Labor Statistics, 42, 44
Causal modelling, 71–78
Chi-squared test, 115
Coefficient alpha (α), 52
Cognitive biases, 19
Combination method, 95–98
Continuous normal distribution, 79
Correction method, 95–98
Corresponding error, 102
Count data distributions, 79
Count data forecasts, quality measures for, 116–118
Critical fractile, 16
Cross-functional communication, 6
Croston’s method, 83–87
Dampening, 57
Data availability, for time series analysis, 33–35
Decision making, 28–29
Decomposition, 41
methods, 42–44
Demand averaging, 5
Demand chasing, 90. See also Naïve forecasting
Demand forecasting, 27
meaning of, 35
Dickey–Fuller test, 36
Domain-specific knowledge, 91–93
Economic order quantity model, 85
Economies of scale in forecasting, 38
Error definition, 102
Error measures, 101
Exponential smoothing model, 23, 34, 77
change and noise, 51–54
extensions, 55–58
optimal smoothing parameters, 54–55
weights for past demand under, 54
Fill rates. See Type II service levels
Financial forecasting, 33
First-order differencing, process of, 64
Forecast errors
empirical distribution of, 25–28
standard deviation of, 25–28
symmetry of, 11–12
Forecast quality measures, 101–118
assessing prediction intervals, 114–116
bias and accuracy, 102–107
for count data forecasts, 116–118
percentage and scaled errors, 107–114
“Forecast Value Added”, 37
analysis, 95
Forecastability, and scale, 37–40
Forecasting
easy- and hard-to–forecast time series, 4–5
decision making, 28–29
preparation of, 3
methods, 9–11
prediction interval, 25–28
from probabilistic perspective, 13
reporting uncertainty, 11–15
service levels, 15–18
time series, 22. See also Time series forecasting
value of, 3–6
Forecasting competitions, 119–125
additional aspects, 124–125
data, 121–122
description, 119
planning, 119–121
procedure, 122–124
Forecasting hierarchies, 135–145
bottom-up forecasting, 137–139
middle-out forecasting, 141–142
multivariate time series, 135–137
optimal reconciliation forecasting, 142–144
other approaches, 144–145
top-down forecasting, 139–141
Forecasting model
domain-specific knowledge in, 91–93
group decision making, 97
judgemental
combination method, 95–98
correction method, 95–98
and leading indicators, 71–72
with multiple methods, 77
political and incentive aspects, 93–95
traditional, 82–83
Forecasting organization, 129–130
Free statistical software R, 7
GARCH (generalized autoregressive conditional heteroscedasticity), 69
GDP growth, 14–15
Goethe, 11
Google Correlate, 74
Google Trends, 74
Group decision making, 97
Hedging, 94
Hindsight bias, 90
Holt’s method, 55
Holt–Winters exponential smoothing, 57
Human judgment, in forecasting, 9
Illusionary trend perception, 21
In-stock probability. See Type I service level
Inherent forecastability, 4
Integration, 62–64
Interaction effect, 92
Intermittent demand series, 81
Croston’s method, 83–87
Intuition vs. Cognition, 89–91
Judgmental forecasts, 95
Leading indicators, 71–74
illustration, 72
and time series, 74–78
use of, 72–73
Level of time series, 41
additive/multiplicative, 46
“Level only” model, 24
Lumpy demand, 81
M-competitions, 51
Make-to-order system, 3–4
Managerial thinking, common mistake in, 36–37
Mean absolute deviation (MAD), 103
Mean absolute error (MAE), 25, 76, 103
Mean absolute scaled error (MASE), 112
Mean error (ME), 103
Mean squared error (MSE), 25, 105
Microsoft Excel, 7
Middle-out forecasting, 141–142
Modern ERP software, 35
Moving averages (MA), 64–66
autocorrelation and partial autocorrelation functions, 67
Multiplicative seasonality, 46
Multiplicative trend, 46
Multistep-ahead forecasts, 113
Multivariate time series, 135–137
Naïve forecasting, 5
News-vendor problem, 16
Normal distribution
continuous, 79
unbounded, 79
Normally distributed demands, 80
One-step-ahead forecast, 21, 113
Optimal reconciliation forecasting, 142–144
Organizational barriers, 130–133
Partial autocorrelation functions (PACF), 66–68
Pearson’s χ2 test, 115
Pegels’ classification, 58
Percentage errors, 107–114
Performance measurement, 101
Point forecast. See Forecast
Poisson-distributed demands, 80
Prediction interval, 25–28, 27f
assessment of, 114–116
Professional forecasters, 9
Regression equation, 75–76
Retail organizations, 10
Rolling origin forecast, 75
Rolling regression forecasts, 75–76, 77
Root mean squared error (RMSE), 106
Sales and operations planning (S&OP), 129–133
forecasting organization, 129–130
organizational barriers, 130–133
steps in, 130
Sandbagging, 94
Scaled errors, 107–114
additive/multiplicative, 46
ARIMA model, 62
methods to remove, 43
stability of, 44–45
variants of, 58
Seasonality parameters, 57
“Seasonally adjusted”, 42
Second guessing, 94
Service levels, 15–18
Shrinkage methods, 34
Single exponential smoothing model, 24
Smoothing parameter, 23
Software
Spinning, 94
Spreadsheet model, 9
software, 7
Squared errors, 106
Stationarity, 35–37
Statistical model, 9
Statistical software R, 58
Stock-keeping units (skus), 135
Syntetos-boylan approximation, 86
System neglect, 19
Theory-based formulas, 25–28
Time series forecasting, 72, 74–78
additive and multiplicative components, 46
cognitive biases in, 89
combination with, 74–78
components of, 41–42
decomposition methods, 42–44
data availability, 33–35
forecastability and scale, 37–40
on sales and advertising, 72
stationarity, 35–37
stability of components, 44–45
Top-down forecasting, 139–141
Traditional forecasting methods, 82–83
Trend
additive/multiplicative, 46
methods to remove, 43
stability of, 44–45
in time series, 41–42
Trends in time series
ARIMA model, 62
variants of, 58
Two-step-ahead forecast, 21
Type I service level, 16
Type II service level, 16
Unbiased forecast, 103
Unbounded normal distribution, 79
Uncertainty, reporting forecast, 11–15
Volkswagen Touran, 40
Weighted MAPE (wMAPE), 110
Wisdom of crowds literature, 96
Withholding, 94
X-13ARIMA-SEATS algorithm, 44