Concept
Outcome Attribution
Decomposing outcomes into their contributing factors to understand why decisions succeeded or failed.
Research Area
Intelligence
Status
Active Research
Version
0.1
Last Updated
July 2026
Related Discipline
Causal Inference, Performance Attribution
Implementation
Orqis
Outcome Attribution goes beyond measuring whether a decision was successful. It decomposes the outcome into its contributing factors — which aspects of the decision process contributed positively, which detracted, and which were irrelevant. This decomposition is what makes learning systems effective.
Precise Definition
Outcome attribution is the systematic decomposition of decision outcomes into parameter-level contributions, using bucketed decision fingerprints and regime-conditioned analysis to understand causation rather than merely observing correlation.
Key Properties
- Parameter-level granularity — attribution operates at the level of individual decision parameters (indicator choice, timeframe, stop-loss width), not aggregate strategy performance
- Regime-conditioned — the same parameter may contribute positively in one regime and negatively in another; attribution captures this environmental dependence
- Bucketed fingerprints — continuous parameters are discretized into meaningful buckets (e.g., RSI period 10-15, 16-20) to accumulate evidence across structurally similar decisions
- Evidence accumulation — attribution confidence grows with observation count, using statistical smoothing to avoid premature conclusions from small samples
Why It Matters
Without attribution, learning systems can only observe that a decision succeeded or failed. With attribution, they can understand why — which parameters contributed, which environmental conditions mattered, and which aspects of the decision process need improvement. Attribution transforms outcomes into actionable intelligence.
Boundary
What this concept is not
This is not performance attribution in the traditional finance sense, which decomposes portfolio returns into factor exposures. Outcome attribution decomposes the decision process itself — which parameters and structural choices contributed to the outcome.
Examples
Exit mechanism attribution: decomposing each closed position by which exit mechanism fired (stop loss, take profit, trailing stop, time decay, signal exit) and measuring the quality of that exit relative to the position's maximum favorable excursion
Parameter bucket scoring: aggregating outcomes across all strategies that used RSI with period 12-16 in trending regimes, producing a win rate and average return per bucket that guides future generation
Preferred/avoid parameter recommendations: the attribution aggregate computation produces per-parameter-bucket recommendations (preferred, neutral, avoid) that are injected into the learning summary for AI generation prompt guidance
Conceptual Neighbors
Requires
- Evidence Systems
- Execution data
Enables
- Learning Systems
- Generation improvement
Produces
- Parameter-level guidance
Opposes
- Aggregate-only evaluation
Current Research Questions
- ?How should attribution handle correlated decision parameters?
- ?Can attribution distinguish between skill and environmental tailwinds?
- ?What granularity of attribution maximizes learning system effectiveness?
- ?How should attribution data be aggregated across decision families?
Related Frameworks
Related Research
Related Papers
This concept is actively implemented in Orqis, Warren Labs' first decision infrastructure platform.