Skip to content

Silo Imprint Audit

Audit — instantiates Organization–Artifact Topology Alignment

Hunts the artifact for boundaries that fossilize an obsolete or accidental team silo rather than the domain, logging each imprint and the cross-boundary coupling that gives it away.

Not every ugly seam in an artifact is a mistake — some track a genuine domain split that should be there. Silo Imprint Audit is the mechanism that tells the two apart. It goes boundary by boundary asking a single question: does this split follow the domain, or the ghost of an organization? Its defining move is to convict a boundary only on two kinds of evidence together — heavy coordination load straddling the seam (the coupling that a "clean" boundary should not have) and a social or historical rationale for the split that has nothing to do with the domain. Where the overlay shows a seam is mismatched, the audit does the archaeology: it confirms the imprint, sizes it, and enters it in a standing register of sociotechnical debt.

Example

A logistics company's shipment system has two services — rate calculation and route optimization — joined by a chatty, brittle interface; nearly two-fifths of cross-service incidents land on it. The audit profiles the seam and finds exactly the signature of an imprint: the two services share pricing data, call each other dozens of times per request, and change together in almost every release — coupling far too high for a real boundary. Then it checks provenance. The split dates to a period when two separate teams owned the pieces; the teams were merged years ago, but the boundary they cast stayed frozen in the code. The audit logs it: high-coupling seam, no domain rationale, traces to a defunct org split — candidate to merge. Crucially, it also clears a neighbouring seam that looks just as ugly — the boundary between domestic and cross-border shipping — because that one tracks a real regulatory difference, not an org accident. The register ends up shorter and far more actionable than a naïve "high coupling = bad" scan would produce.

How it works

The distinctive engine is the two-evidence imprint test, applied per seam:

  • Profile the coordination load. For each boundary, measure the cross-seam coupling — call volume, shared data, co-change frequency, incident co-occurrence — to find seams that cost far more to coordinate than a clean interface should.
  • Check the provenance. Ask why the boundary exists. A split that maps to a past reorg, an acquisition, or "who happened to be free that quarter" is a candidate imprint; one that maps to a domain, a regulation, or a real interface is not.
  • Convict only on both. Log a boundary as an imprint when it shows high coordination load and an org-shaped rationale — and record it with a severity so the worst offenders surface first.

Tuning parameters

  • Coupling signal — which measures count as evidence (runtime calls, shared schemas, co-change history, incident co-occurrence). Co-change is often the sharpest tell of a false boundary.
  • Imprint threshold — how much load, over how clean a boundary should be, before a seam is flagged. Set it loose and the register fills with load-bearing seams that are fine.
  • Provenance depth — how far back to trace a boundary's origin. Deeper tracing catches older ghosts but costs research and often runs into lost history.
  • Domain-vs-org adjudication — who rules whether a rationale is "domain" or "org," and against what reference (ideally an independent domain boundary map).
  • Register severity scale — how imprints are ranked so remediation attention goes to the costliest, not merely the oldest.

When it helps, and when it misleads

Its strength is discrimination: it separates ugly-but-right from ugly-because-org, which a coupling scan alone cannot do, and it leaves behind a durable register that survives the audit and can be worked down over time. It is the mechanism that turns the mirroring hypothesis from a slogan into a specific, evidenced list.[1]

It misleads when either half of the test is weak. Not all high coupling is an imprint — some domains are genuinely entangled, and flagging those invites a refactor that makes things worse. Provenance is frequently lost or rewritten, so a boundary's real origin can be unknowable, tempting the auditor to assume org-imprint when the honest answer is "unclear." And the register is easy to weaponize — assembled to justify a merger or reorg that leadership already wants. The discipline is to require both coupling evidence and a non-domain rationale before convicting, to mark unproven seams as "unclear" rather than guilty, and to treat the register as a queue of hypotheses, not a demolition order.

How it implements the components

The audit realizes the detection-and-logging face of the archetype — finding imprints and tracking them, not mapping the whole topology:

  • sociotechnical_mismatch_register — its primary product: the standing, severity-ranked log of confirmed org-imprinted boundaries and the evidence for each.
  • coupling_and_coordination_load_profile — a per-seam coordination-load reading computed as the evidence for imprint (the audit's diagnostic use of coupling, seam by seam).

It does not build the co-registered picture that first flags a seam — that is Organization–Artifact Topology Overlay — nor does it trace a boundary's full genealogy across successive reorganizations; the durable lineage record is Post-Reorganization Architecture Impact Review and the ownership decision log is Architecture Decision Record with Ownership.

References

[1] The mirroring hypothesis holds that an artifact's dependency structure tends to mirror the communication and organizational ties of its builders. A persistent, high-coupling boundary with no domain rationale is precisely where the mirror has outlived the organization that cast it. ↩