Forecast software supply chain loss
Forecast aggregate software supply-chain loss with tenant-calibrated empirical-Bayes Gamma-Poisson incident frequencies, partially pooled lognormal loss marks, common frequency/severity/business scenarios, and multiplicative unique-application disruption paths; unsupported risk classes make the output diagnostic-only.
What it's for
Gives boards a locally calibrated expected and tail-loss distribution for software supply-chain exposure instead of multiplying CVE counts by generic industry loss tables.
What you give it
Inputs split into evidence read from your connected systems, calibration your team owns, and numerical controls that affect precision but never the result's meaning.
| Field | Type | Role | Required |
|---|---|---|---|
| applications | array of objects (3 fields) ≥ 1 item | Evidence | Yes |
| components | array of objects (5 fields) ≥ 1 item | Evidence | Yes |
| exposure_edges | array of objects (5 fields) ≥ 1 item | Evidence | Yes |
| historical_loss_events | array of objects (4 fields) ≥ 1 item | Evidence | Yes |
| historical_periods | array of objects (5 fields) ≥ 1 item | Evidence | Yes |
| horizon_periods | integer ≥ 1, ≤ 120 | Your calibration | Optional |
| max_detail_rows | integer ≥ 1, ≤ 500 | Numerical control | Optional |
| minimum_class_events | integer ≥ 1 | Your calibration | Optional |
| minimum_class_periods | integer ≥ 1 | Your calibration | Optional |
| prior_exposure_units | number > 0 | Your calibration | Optional |
| scenarios | array of objects (6 fields) ≥ 2 items | Evidence | Yes |
| seed | integer ≥ 0, ≤ 4294967295 | Numerical control | Optional |
| severity_prior_strength | number ≥ 0 | Your calibration | Optional |
| simulations | integer ≥ 200, ≤ 100000 | Numerical control | Optional |
| tail_probability | number > 0, ≤ 0.5 | Your calibration | Optional |
Each scenarios
record
| Field | Type | Required |
|---|---|---|
| business_loss_multiplier | number (≥ 0, ≤ 1000) | Yes |
| evidence_verified | boolean | Yes |
| frequency_multiplier | number (≥ 0, ≤ 1000) | Yes |
| id | string (non-empty) | Yes |
| probability | number (≥ 0, ≤ 1) | Yes |
| severity_multiplier | number (≥ 0, ≤ 1000) | Yes |
{
"applications": [
{
"evidence_verified": true,
"id": "checkout",
"value_at_risk": 200
},
{
"evidence_verified": true,
"id": "admin",
"value_at_risk": 80
}
],
"components": [
{
"event_exposure_units": 1,
"evidence_verified": true,
"frequency_multiplier": 1.2,
"id": "parser",
"risk_class": "third-party-runtime"
},
{
"event_exposure_units": 1,
"evidence_verified": true,
"frequency_multiplier": 1,
"id": "runtime",
"risk_class": "third-party-runtime"
}
],
"exposure_edges": [
{
"application_id": "checkout",
"component_id": "parser",
"evidence_verified": true,
"id": "checkout-parser",
"loss_fraction_if_incident": 0.8
},
{
"application_id": "checkout",
"component_id": "runtime",
"evidence_verified": true,
"id": "checkout-runtime",
"loss_fraction_if_incident": 0.6
},Truncated for display — the full payload is 159 lines.
What you get back
This is the actual output of running the example above — computed by the same function the platform calls, not an illustration.
{
"application_diagnostics": [
{
"application_id": "checkout",
"expected_business_disruption_loss": 4.56,
"simulated_loss_probability": 0.03,
"value_at_risk": 200
},
{
"application_id": "admin",
"expected_business_disruption_loss": 0.4,
"simulated_loss_probability": 0.01,
"value_at_risk": 80
}
],
"assumptions": [
"Historical periods are complete exposure denominators with cause-consistent loss-event counts; loss marks are finance-reconciled direct incident costs. Right-truncated, selectively recorded or post-outcome reclassified periods invalidate frequency and severity calibration.",
"Risk-class rates share a Gamma-Poisson empirical-Bayes prior and positive losses share a lognormal prior. Shared scenarios preserve frequency, severity and business-impact dependence; the model does not infer adversary intent, exploitability or causality from repository activity.",
"An application incurs its unique disruption value through a multiplicative union of incident paths and is never added once per dependency. Direct response losses remain separate. Unsupported classes or unverified evidence make every numeric result diagnostic-only.",
"This is aggregate decision support, not a breach prediction, vulnerability disclosure, insurance estimate, individual/vendor ranking or authority to patch, isolate, investigate, notify or spend."
],
"component_diagnostics": [
{
"class_support_sufficient": true,
"component_id": "parser",
"exposed_application_count": 1,
"risk_class": "third-party-runtime",
"simulated_expected_incident_count": 0.02,
"simulated_incident_probability": 0.02
},
{
"class_support_sufficient": true,
"component_id": "runtime",
"exposed_application_count": 2,
"risk_class": "third-party-runtime",
"simulated_expected_incident_count": 0.01,
"simulated_incident_probability": 0.01
}
],
"configuration": {
"application_loss_rule": "multiplicative_survival_union_per_unique_application",
"frequency_model": "empirical_bayes_gamma_poisson",
"minimum_class_events": 2,
"minimum_class_periods": 10,Truncated for display — the full payload is 93 lines.
How it works
Forecasting & survival — Estimate when something completes or fails, with censoring and unresolved work handled honestly rather than dropped.
- 1 Build complete class-period exposure denominators and finance-reconciled positive loss marks without dropping zero-event periods or reclassifying history after outcomes.
- 2 Partially pool class incident rates through a Gamma-Poisson tenant prior and log-loss severity through a tenant-wide lognormal prior, retaining period/event support gates for every current class.
- 3 Simulate shared scenario frequency, severity and business-impact multipliers; combine overlapping component incidents once inside each unique application and keep direct incident cost separate before reporting VaR/CVaR and support diagnostics.
Before you trust it
Every tool in the catalog ships with the conditions under which its answer is meaningful — and the conditions under which it should abstain instead of guessing.
Assumptions & guardrails
- Training examples precede their outcomes, censoring and unresolved work are represented, and deployment populations remain comparable to validation cohorts.
- Exposure denominators and incident/loss definitions are stable and complete, loss marks are positive and finance reconciled, current class mapping is frozen, and scenarios preserve relevant dependence.
- A predictive interval or risk estimate is not a deadline promise, causal explanation, or individual-performance judgment.
- This is not a breach prediction, exploitability judgment, insurance estimate or vulnerability disclosure; unsupported classes or unverified evidence are abstentions, not permission to borrow another tenant's rates.
Minimum evidence
- historical_periods: at least 1 rows/items
- historical_loss_events: at least 1 rows/items
- applications: at least 1 rows/items
- components: at least 1 rows/items
- exposure_edges: at least 1 rows/items
- scenarios: at least 2 rows/items
How to validate it
Use chronological train/calibration/test windows, compare proper scores and decision value with a simple baseline, and recalibrate only from outcomes resolved after prediction time.
Calibrating it to your org
Same for everyone
The mathematical kernel, validation rules, method version, and JSON output semantics are organization-independent; no tenant-trained coefficients or company benchmark is embedded in the function.
Specific to you
- tenant-scoped historical security/continuity incident cohort joined to deployed component exposure and finance loss, plus a frozen current runtime application-component graph and common scenario set, preserving period zeros and point-in-time class definitions
- incident/loss/exposure/class definitions and epoch, completeness and censoring, direct-versus-business loss perimeter, application-value uniqueness, edge fractions, scenario dependence/probabilities, empirical-Bayes prior/support, horizon/tail, simulation precision, evidence thresholds and accountable security/platform/finance/risk owners
Calibration workflow
- 1 Define the management decision, target outcome, aggregate unit, privacy boundary, cadence, and prediction/intervention horizon for this organization.
- 2 Build a tenant-scoped historical cohort using only information available before each prediction or decision; preserve zero periods, censoring, assignment probabilities, and unresolved outcomes when the method requires them.
- 3 Estimate statistical parameters on training history, but obtain costs, utilities, risk tolerance, practical-effect thresholds, capacity, and policy constraints from accountable decision owners.
- 4 Validate on later time windows or held-out aggregate units at the deployment grain, against a simple baseline and the function-specific validation strategy.
- 5 Deploy only if the returned decision clears evidence, overlap, calibration, robustness, and guardrail checks; warning, unsupported, schema-gap, and fallback decisions are abstentions.
- 6 Monitor realized outcomes, data drift, coverage, and decision regret; recalibrate at a governed cadence or after a detected regime/definition change, never merely because a stakeholder dislikes the result.
Call it from your AI
You don't wire up 388 tools in your MCP client. The GitRevio MCP server exposes 18 tools, three of which let an agent search the catalog, read a tool's schema, and run it — so the assistant finds this one on its own.
gitrevio_capabilities_search
{ "q": "forecast aggregate software supplychain loss with" }
→ finds "forecast_software_supply_chain_loss"
gitrevio_capability_describe
{ "capability_id": "forecast_software_supply_chain_loss" }
→ returns the input schema and agent guidance shown on this page
gitrevio_capability_run
{ "capability_id": "forecast_software_supply_chain_loss", "arguments": { ... } }
→ returns the result shown aboveWorks in Claude Desktop, Claude Code, Cursor, Cline, Continue.dev, Goose and Aider. See the MCP server.
Related tools
Audit software supply chain integrity
Audit the deployed runtime software supply chain from application roots through resolved dependency edges: reconcile SBOM freshness, version resolution, source pinning, artifact attestation, support horizon, license policy, vulnerability disposition, evidence coverage and unique application value without treating repository text as provenance or exploitability evidence.
Optimize software supply chain remediation portfolio
Choose exactly one accept, patch, upgrade, replace, isolate or remove option per governed component while unique application disruption paths, direct incident loss, transition/operating/upfront cost, license compliance, scenario availability, cross-option feasibility, budget, capacity and CVaR are optimized together with exact or disclosed beam search.
Audit cyber control evidence integrity
Audit whether claimed defense in depth is supported by current independent control tests mapped to declared attack-path steps, while preserving duplicate mappings and counting each exposed business asset only once.
Audit data sovereignty residency evidence integrity
Audit every governed data asset's point-in-time storage, processing, replica, backup, log/cache and key locations plus cross-region transfers against an effective counsel-supplied residency policy, independent evidence, encryption controls and retention limits.
Audit workforce identity access evidence integrity
Audit the point-in-time chain from an opaque workforce subject through authorized accounts, independent identity evidence, approved least-privilege grants and MFA/device-backed access events.
Bayesian account risk triage
Prioritize human review of auditable account-security and policy-conflict evidence using Bayes factors and decision costs.