Optimize incident learning portfolio

What it's for

Moves reliability investment beyond backlog ranking: Gitrevio can decide when to remediate now and when a controlled experiment is worth buying because its result would actually change the action under risk and budget constraints.

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.

{
  "budget": 10000,
  "capacity_units": 10,
  "experiments": [
    {
      "capacity_units": 0.5,
      "controls_verified": true,
      "depends_on": [],
      "design_reliability": 0.9,
      "disruption_loss_by_scenario": [
        20,
        80
      ],
      "evidence_verified": true,
      "excludes": [],
      "experiment_cost": 200,
      "id": "isolated-failure-injection",
      "risk_mode_id": "dependency-timeout",
      "sensitivity": 0.9,
      "specificity": 0.9
    }
  ],
  "minimum_information_gain": 0.01,
  "remediation_options": [
    {
      "capacity_units": 2,
      "causal_effect_verified": true,
      "controls_verified": true,
      "depends_on": [],
      "design_reliability": 0.9,
      "effectiveness_by_scenario": [
        0.9,
        0.8
      ],
      "evidence_verified": true,
      "excludes": [],
      "id": "adaptive-timeout-control",
      "implementation_cost": 3000,
      "risk_mode_id": "dependency-timeout"
    }
  ],
  "risk_modes": [
    {
      "conditional_loss_by_scenario": [

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.

{
  "configuration": {
    "budget": 10000,
    "capacity_units": 10,
    "maximum_conditional_value_at_risk": null,
    "maximum_expected_loss": null,
    "minimum_information_gain": 0.01,
    "risk_aversion": 0.25,
    "seed": 53,
    "simulations": 800,
    "tail_probability": 0.95
  },
  "decision": "portfolio_selected",
  "evidence_diagnostics": {
    "effect_shrinkage": "remediation effects shrink toward zero and experiment accuracy toward chance by design reliability",
    "unverified_risk_mode_ids": []
  },
  "excluded_options": [],
  "interpretation": "Information gain is valuable only through a predeclared contingent decision; sensitivity, specificity and remediation effects require prospective evidence. Selection is not authorization to run a production experiment, alter incident policy, or judge an individual.",
  "method": "two_stage_bayesian_value_of_information_common_scenario_portfolio_search",
  "pareto_frontier": [
    {
      "conditional_value_at_risk": 62421,
      "expected_risk_loss": 3908.55,
      "expected_total_cost": 4821.05,
      "information_gain_nats": 0.1299,
      "objective": 20426.3,
      "option_ids": [
        "adaptive-timeout-control",
        "isolated-failure-injection"
      ],
      "reserved_capacity": 2.5,
      "reserved_cost": 3200,
      "value_at_risk": 12720
    }
  ],
  "search": {
    "candidate_combination_count": 3,
    "evaluated_portfolio_count": 3,
    "feasible_portfolio_count": 1,
    "global_optimum_certified": true,
    "method": "exact_enumeration"
  },
  "selected_policies": [

How it works

Causal inference & experiment design — Separate what a change caused from what merely moved alongside it.

1. 1 Admit only controlled experiments with verified prospective sensitivity/specificity evidence and remediations with verified causal effects; represent each failure mode by a governed prior and coherent conditional-loss scenarios.
2. 2 Shrink experiment sensitivity/specificity toward chance and remediation effects toward zero by their governed design reliability; then compute Bayesian posterior risk after positive and negative signals and predeclare whether the linked remediation would execute in each branch, so information has value only through that contingent decision.
3. 3 Evaluate one baseline, immediate or experiment-contingent policy per mode under common latent states and scenarios, count shared loss groups once, and exact-enumerate or deterministically beam-search the feasible cost/capacity/expected-loss/CVaR/information Pareto set.

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": "choose immediate remediation or a predeclared" }
  → finds "optimize_incident_learning_portfolio"

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

gitrevio_capability_run
  { "capability_id": "optimize_incident_learning_portfolio", "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 Quality, incidents & reliability →