This glossary defines technical terms used across AlphaX Decision Sciences' products, research, and publications. The intent is to provide clear, consistent definitions for concepts spanning production forecasting, artificial intelligence, statistics, and decision science.
Where applicable, definitions align with established industry and scientific standards. AlphaX retains ownership of interpretation and application.
Oil & Gas Forecasting and Reserves
- Analog Well
- A historical well used as a reference for forecasting another well based on similarities in geology, completion design, location, or operating conditions.
- Arps Decline
- A parametric decline curve model used to extrapolate future production from historical rates. A mathematical fitting approach, not a physics-based model.
- Decline Curve Analysis (DCA)
- A forecasting method that fits mathematical decline functions to historical production data. Effective for long, stable production histories; unreliable for early-life or noisy wells.
- Estimated Ultimate Recovery (EUR)
- The total expected production from a well over its producing life.
- P10 / P50 / P90
-
Probabilistic forecast percentiles representing uncertainty ranges:
- P10: optimistic
- P50: median
- P90: conservative
- Production Regime
- A distinct phase of well behavior (e.g., transient flow, boundary-dominated flow) that influences forecast reliability.
- Reserves Classification (PDP / PUD)
- Standard categories describing the development and certainty status of hydrocarbon reserves.
Artificial Intelligence and Modeling
- Artificial Intelligence (AI)
- Algorithms that learn patterns from data to make predictions or classifications. In forecasting, AI is used to generalize behavior across large populations of wells.
- Assistive AI
- AI designed to support human decision-making by surfacing patterns, scenarios, or exceptions rather than replacing accountability.
- Bias (Model Bias)
- Systematic error that causes consistent over- or under-prediction, often introduced by data limitations or modeling assumptions.
- Ensemble Model
- A modeling approach that combines multiple independent models to improve robustness, reduce overfitting, and quantify uncertainty.
- Feature Engineering
- The process of selecting, transforming, or constructing input variables used by a machine learning model.
- Generalization
- A model's ability to perform well on new, unseen data rather than memorizing training examples.
- Overfitting
- When a model captures noise instead of signal, performing well on training data but poorly on new data.
- Probabilistic Forecasting
- Forecasting that produces distributions of outcomes rather than single-point estimates, explicitly modeling uncertainty.
Decision Science and Workflow
- Decision Intelligence
- The application of analytics, AI, and domain logic to improve the quality, speed, and consistency of decisions.
- Exception-Based Review
- A workflow in which automated analysis flags only material deviations, focusing expert attention where it adds the most value.
- Forecast Convergence
- Agreement across multiple independent models or methods, increasing confidence in forecast outcomes.
- Human-in-the-Loop
- A system design that intentionally incorporates human expertise at validation, review, or decision points.
- Repeatability
- The ability to produce the same result using the same inputs and methodology.
- Reproducibility
- The ability for different users or teams to arrive at the same result using the same data and process.
- Workflow Collapse
- The reduction of multi-step, multi-role analytical workflows into fewer, faster, integrated decision steps through automation and AI.
References
Certain foundational terms align with definitions and concepts used by established standards bodies and scientific literature, including:
- Society of Petroleum Engineers (SPE-PRMS)
- U.S. SEC Regulation S-X, Subpart 1200
- Hastie, Tibshirani & Friedman, The Elements of Statistical Learning
- National Institute of Standards and Technology (NIST) guidance on repeatability and reproducibility