Decision Traces: Why Systematic Investing Needs to Explain Why, Not Just What
Most investment systems record outputs, not reasoning. Decision traces capture why a system acted at the moment of commitment—inputs used, constraints applied, actions taken—turning opaque models into legible, self-auditing decision engines.
What are Decision Traces?
A decision trace is a structured record of why a system took an action, not just what action it took.
Most analytical systems capture state:
- Scores
- Rankings
- Outputs
- Portfolios
- Backtests
Decision traces capture reasoning at the moment of commitment:
- What inputs were referenced
- Which constraints were evaluated
- What rules or policies were binding
- Which alternatives were rejected
- What action was taken
- Under what exceptions or overrides
- And how outcomes later unfolded
In short:
State tells us what happened.
Decision traces tell us why it happened.
Without decision traces, a system can be statistically sophisticated yet logically opaque.
What This Means for RevDog
RevDog was explicitly designed to separate:
- Conviction (stock-level attractiveness)
- Construction (portfolio-level risk control)
This separation is philosophically correct—but it creates a new requirement:
the system must explain how conviction was transformed into portfolio construction.
Decision traces provide that missing layer.
For RevDog, decision traces mean the system can explicitly answer questions such as:
- Why was a high-scoring stock excluded?
- Which constraint actually limited portfolio breadth?
- Was risk driven by selection or sizing?
- Which regimes systematically compress diversification?
- Which rules reduce drawdown, and which merely shift it?
This moves RevDog from being:
- a scoring and optimization engine
to being: - a self-auditing decision system
That distinction matters for professional capital.
Why This Matters for Professional Users
For PMS, family offices, and institutional allocators, the key question is not:
“What return did the model produce?”
It is:
“Can the system explain its behavior consistently across cycles?”
Decision traces support:
- Capital preservation narratives
- Governance and oversight
- Model iteration without regime overfitting
- Confidence during drawdowns
- Clear attribution without forward-looking claims
They also reduce dependence on human storytelling after outcomes are known.
A Subtle but Important Shift
Decision traces change how systems evolve.
Instead of:
- Designing a perfect world model upfront
- Hard-coding all possible relationships
- Freezing assumptions early
The system learns from its own behavior:
- Which entities matter
- Which relationships recur
- Which constraints dominate decisions
- Which exception paths are common
Over time, the structure of the decision-making process becomes clearer—not because it was prescribed, but because it was observed.
How This Is Implemented
Decision traces do not require rebuilding the entire system.
They start at commit surfaces:
- Portfolio inclusion / exclusion
- Weight capping
- Sector or risk vetoes
- Cash retention decisions
At each surface, the system records:
- Inputs referenced
- Constraints evaluated
- Action taken
- Reason codes (machine-readable)
Over time, these traces form a coherent map of how the system actually behaves under different conditions.
Bottom Line
Decision traces are not about explainability for marketing.
They are about discipline.
They ensure that:
- Every action taken by the system is inspectable
- Every outcome can be traced back to a concrete decision
- Every model improvement is grounded in observed behavior, not hindsight
For RevDog, decision traces are the natural next step in building systems that:
- Do fewer stupid things
- Fail in understandable ways
- And earn trust not through performance claims, but through structural clarity