Blast Radius Scoring

The Impact Explorer assigns every architectural node — Service, Database, Message Channel, API Endpoint — a Downstream Gravity Score that quantifies the risk of that node failing or being modified. This score is then mapped to an Impact Tier (T0 through T4) for human-readable risk classification.

---

Impact Tiers

Every node in the architecture graph is classified into one of five tiers based on its gravity score:

---

How the Score Works

The gravity score answers a single question: "If this node dies, how much of the architecture breaks?"

Core Principle: Downstream Gravity

The score is based exclusively on downstream impact — the nodes that would be directly or transitively broken. It does not factor in upstream dependencies (those make you fragile, not dangerous).

What Makes It "Weighted"

Unlike a simple connection count, the gravity score weighs each downstream dependency by its connectivity. A downstream node that is itself a critical hub (with many connections) contributes more weight than an isolated leaf. This ensures that breaking a resource used by a critical hub service is scored higher than breaking one used by a standalone worker.

Worked Examples

Example 1: Standalone email worker

• Topology: email-worker → (LISTENS_TO) → notifications
• The worker has zero downstream impact — if it breaks, nothing downstream is affected.
• Gravity Score: 0 → T4 Contained

Example 2: Service writing to a shared database

• Topology: order-service → (WRITES) → orders_db ← (READS) ← reporting-service, analytics-service, ...8 others
• The orders_db table is used by 8 reader services. If order-service corrupts the data or the table schema changes, 8 downstream consumers are impacted.
• The downstream nodes each have moderate connectivity, and their degree is factored into the weight.
• Gravity Score: ~16 → T2 High

Example 3: Event bus producer

• Topology: payment-service → (PUBLISHES_TO) → payment.completed ← (LISTENS_TO) ← 12 consumer services
• If payment-service stops publishing events, 12 downstream consumers receive no data.
• Gravity Score: ~22 → T2 High

Example 4: Enterprise-scale data monolith

• Topology: identity-service → (WRITES) → core_users ← (READS) ← 30 reader services
• A massive shared database. An outage in the writer or a schema migration cascades to 30 downstream teams.
• Gravity Score: ~72 → T1 Critical

---

Interpreting the Banner

When you select a node in the Impact Explorer, the Target Banner shows:

---

When a High Tier Is Acceptable

Not every T0/T1 score requires immediate action:

• Authentication services — A central identity-service will naturally score T1 or higher. The mitigation is operational (redundancy, circuit breakers), not architectural.
• Event buses — A core message broker concentrates connections by design. Monitor the SPOF score in the SPOFs dashboard for concentration risk.
• API gateways — A shared API gateway serving many consumers will score high. Ensure it has horizontal scaling and rate limiting.

When to Act

Prioritize action when:

1. T0 or T1 with no redundancy — The node is critical but has no failover, circuit breaker, or graceful degradation strategy.
2. Cross-team T2+ — Multiple teams depend on a single resource with no formal data contract or API versioning.
3. Tight schema coupling — Services read directly from a shared database table without an abstraction layer (API or view).
4. No ownership — The high-scoring node has no assigned team owner (see Governance).

---

Relationship to SPOF Analysis

The gravity score and the SPOF Score measure related but distinct things:

Use both together: the SPOF dashboard identifies which resources are dangerously concentrated, and the Impact Explorer shows what happens when one of those resources fails.

---

Further Reading

• Architecture Dashboard — The unified dashboard containing the Impact Explorer
• SPOFs & Data Gravity — Concentration risk analysis
• Impact Evaluation — Pre-commit blast radius prediction for code changes
• MCP Server — analyze_blast_radius — Programmatic blast radius queries