Module marginal_owner_risk 

Marginal-owner risk: ownership concentration × code-health fusion.

For each file in the yellow or red health band, reports the maximum knowledge share held by any author who committed within the trailing window_days. A low top-active share on an already-unhealthy file signals that the people most likely to fix regressions there have shallow familiarity — the “marginal owner” who would have to step in has little context to work with.

Risk tier:

• high — red band AND top active share < 0.10
• elevated — (red AND share < 0.30) OR (yellow AND share < 0.10)
• rows that do not meet either threshold are excluded (green band or sufficiently concentrated ownership)

The ownership × code-quality interaction is correlational, not causal. See: Palomba et al., “On the Interplay between Structural Smells and Code Ownership” EASE 2023, arXiv 2304.11636.

Structs

MarginalOwnerRiskRow

A single file flagged with marginal-owner risk.

Functions

classify_risk

Classify a file’s risk tier from its health band and the maximum knowledge share held by any active author.

run_marginal_owner_risk

Run the marginal-owner-risk analysis over the already-ingested fact store.