Forecast governed release competing risks

What it's for

Upgrades the headline 'release risk based on historical patterns' from universal hand-set thresholds to a tenant-local, validation-gated probability of distinct production outcomes with honest uncertainty.

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.

{
  "as_of_ms": 10000,
  "current_changes": [
    {
      "change_class_id": "class-3",
      "decision_at_ms": 9000,
      "evidence_verified": true,
      "feature_values": {
        "clean_signal": 0,
        "hotfix_signal": 0,
        "incident_signal": 0,
        "rollback_signal": 1
      },
      "id": "current-payments-release",
      "repository_id": "payments"
    }
  ],
  "eligible_history": [
    {
      "change_class_id": "class-0",
      "decision_at_ms": 1000,
      "deployed": true,
      "deployment_propensity": 0.8,
      "evidence_verified": true,
      "feature_values": {
        "clean_signal": 1,
        "hotfix_signal": 0,
        "incident_signal": 0,
        "rollback_signal": 0
      },
      "id": "release-example-0",
      "legacy_risk_probability": 0.5,
      "propensity_source_id": "registered-release-policy",
      "propensity_verified": true,
      "repository_id": "payments"
    },
    {
      "change_class_id": "class-1",
      "decision_at_ms": 1100,
      "deployed": true,
      "deployment_propensity": 0.8,
      "evidence_verified": true,
      "feature_values": {
        "clean_signal": 1,

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.

{
  "accepted": true,
  "assumptions": [
    "Deployment propensities are point-in-time probabilities for the actual release decision and all joint causes of deployment selection and production outcome are represented by the submitted aggregate features.",
    "Release periods are contiguous at-risk intervals with one mutually exclusive first rollback, hotfix or incident event; censoring and the maturity window are independent conditional on the modeled history."
  ],
  "counts": {
    "current_changes": 1,
    "deployed_releases": 48,
    "eligible_historical_changes": 60,
    "holdout_releases": 10,
    "training_events_by_cause": {
      "hotfix": 6,
      "incident": 6,
      "rollback": 6
    },
    "training_releases": 38,
    "undeployed_changes_retained": 12
  },
  "decision": "release_competing_risk_model_accepted",
  "failed_gates": [],
  "forecasts": [
    {
      "any_failure_probability": 0.9428,
      "change_id": "current-payments-release",
      "competing_outcomes": {
        "hotfix": {
          "p05": 0,
          "p95": 0.8953,
          "probability": 0.0032
        },
        "incident": {
          "p05": 0,
          "p95": 0.9595,
          "probability": 0.0051
        },
        "none": {
          "p05": 0.0003,
          "p95": 0.5229,
          "probability": 0.0572
        },
        "rollback": {
          "p05": 0.0148,
          "p95": 0.9845,

How it works

Sequential Bayesian & bandits — Learn while deciding — update beliefs as evidence arrives and choose where the next unit of effort is worth spending.

1. 1 Retain every eligible historical change and its frozen deployment propensity, then build contiguous mature at-risk periods only for exactly linked deployed releases.
2. 2 Hold out the latest whole releases, fit inverse-propensity-weighted cause-specific softmax hazards on earlier releases and compare cumulative competing incidence against a base model and any supplied legacy risk score.
3. 3 Abstain on selection-support, cause-support, maturity, calibration or baseline failure; otherwise refit and return posterior cumulative-incidence intervals explicitly conditional on deployment.

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.

Calibrating it to your org

Calibration workflow

1. 1 Define the management decision, target outcome, aggregate unit, privacy boundary, cadence, and prediction/intervention horizon for this organization.
2. 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. 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. 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. 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. 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 companylocal rollback hotfix and incident" }
  → finds "forecast_governed_release_competing_risks"

gitrevio_capability_describe
  { "capability_id": "forecast_governed_release_competing_risks" }
  → returns the input schema and agent guidance shown on this page

gitrevio_capability_run
  { "capability_id": "forecast_governed_release_competing_risks", "arguments": { ... } }
  → returns the result shown above

Works in Claude Desktop, Claude Code, Cursor, Cline, Continue.dev, Goose and Aider. See the MCP server.

Related tools

See every tool in Delivery forecasting & commitments →