Accuracy (forecast accuracy)
How close a forecast is to what actually happens. Important, but not enough on its own—accurate forecasts can still lead to poor decisions if they are unstable or not trusted.
AI-assisted forecasting
Using AI tools to help explore data, generate ideas, or simulate scenarios. AI supports thinking but does not replace human judgment.
AI-Enabled Reasoning
Using AI as a learning partner to test ideas, explore alternatives, and deepen understanding—not just to generate answers.
ARIMA
A forecasting approach that uses past patterns in the data to predict the future, focusing on how values depend on previous values.
Backtesting
Testing a forecasting method on past data to see how it would have performed. Helps build trust before using it in real decisions.
Behavior (forecast behavior)
How a forecast responds over time—whether it is stable, reactive, or erratic. Behavior matters more than a single accuracy score.
Causal drivers
External factors that influence outcomes, such as price, promotions, or weather.
Context-aware forecasting
Forecasting that incorporates external information, not just past values of the series.
Data conditions
The characteristics of your data, such as how much history you have, how noisy it is, and whether patterns are stable.
Decision context
The situation in which a forecast is used—what decision is being made, how urgent it is, and what is at risk.
Decision design
The process of structuring how forecasts are used to support decisions, including thresholds, responsibilities, and actions.
Decision systems
The broader system in which forecasts are used, including data, models, people, and processes.
Decomposition
Breaking a time series into parts such as trend, seasonality, and noise to better understand patterns.
Deep learning
Advanced models that learn patterns automatically from large datasets, often with limited interpretability.
Diagnostics
Tools used to check whether a forecasting model is behaving properly and can be trusted.
ETS models
A family of forecasting methods that combine error, trend, and seasonal components into a unified framework.
Explicit structure
A way of modeling where patterns like trend and seasonality are clearly separated and visible.
Feature-based forecasting
A forecasting approach that uses engineered inputs (features), such as lags or rolling averages, instead of explicit components.
Feature engineering
The process of creating useful inputs for models, such as lagged values or seasonal indicators.
Forecast behavior
See Behavior.
Forecast governance
The rules and processes used to monitor and manage forecasting systems.
Forecast reliability
How consistent and dependable a forecast is over time.
Forecast stability
Whether a forecast changes smoothly or reacts too strongly to small fluctuations.
Forecast trust
The degree to which decision-makers believe a forecast is reliable and safe to use.
Forecasting by design
The idea that forecasting is not just about predicting values, but about designing systems that support good decisions.
Generative AI
AI systems that can create new content or scenarios, often used for exploration rather than precise forecasting.
Governance
The structures and processes used to ensure forecasts are used responsibly and monitored over time.
Holt–Winters method
A forecasting method that updates trend and seasonal patterns over time to adapt to new data.
Implicit structure
A modeling approach where patterns are not separated visibly but are embedded in relationships within the data.
Interpretability
How easy it is to understand how a model works and explain its results.
Lagged values
Past observations used as inputs to predict future values.
LSTM (Long Short-Term Memory)
A type of deep learning model designed to learn patterns over time by remembering and forgetting information.
Machine learning (ML)
A set of methods that learn patterns from data automatically, often used when relationships are complex.
Model behavior
See Behavior.
Model stability
How consistently a model performs over time.
Moving average
A simple method that smooths data by averaging recent values to reveal underlying trends.
Noise
Random variation in data that does not reflect meaningful patterns.
NorthStar Retail Group
The main case example used throughout the book to illustrate forecasting concepts in a business setting.
Overfitting
When a model learns patterns that are too specific to past data and fails to generalize to new data.
Operational forecasting
Forecasting used for short-term, day-to-day decisions.
Persistence
The idea that past values influence future values.
Prediction vs. forecasting
Prediction focuses on estimating outcomes; forecasting focuses on supporting decisions over time under uncertainty.
Random forests
A machine learning method that combines multiple decision trees to improve predictions.
Regression models
Models that explain relationships between variables, often used in feature-based forecasting.
Residuals
The difference between actual values and forecasted values.
Residual diagnostics
Checking residuals to see whether a model is missing patterns or behaving poorly.
Robustness
How well a model continues to perform under changing conditions.
SARIMA
An extension of ARIMA that accounts for seasonal patterns.
Scenario analysis
Exploring different possible futures to support planning and decision-making.
Seasonality
Patterns that repeat at regular intervals, such as weekly or yearly cycles.
Signal vs. noise
Distinguishing meaningful patterns (signal) from random variation (noise).
Smoothing
Reducing short-term fluctuations to reveal underlying trends.
Stability
See Forecast stability.
State-space models
Models that represent systems using hidden (latent) components that evolve over time.
Structure (temporal structure)
The underlying patterns in data, such as trend, seasonality, and dependence.
TBATS
A forecasting method designed to handle complex seasonal patterns.
Temporal dependence
The relationship between current and past values in a time series.
Time series
Data collected over time in sequence.
Transformers
Deep learning models that use attention to learn relationships across time.
Trust (forecast trust)
See Forecast trust.
TSLM (Time Series Linear Model)
A regression-based approach that represents time-related patterns as variables.
Uncertainty
The fact that future outcomes are not known and forecasts are imperfect.
Under-specification
Using models that are too simple for the decision context.
Validation
Testing whether a model performs well on new, unseen data.
Volatility
Frequent and unpredictable changes in data.
Workflow (forecasting workflow)
The full process of forecasting, from data collection to decision-making.
