Infer competing root cause posterior

What it's for

Upgrades root-cause traceback from one deterministic path to a calibrated probability distribution across competing and co-occurring explanations, including an honest unknown state.

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": 130,
  "current_observations": [
    {
      "evidence_verified": true,
      "id": "current-queue",
      "measurement_reliability": 0.95,
      "observed_at_ms": 125,
      "present": true,
      "signal_id": "queue"
    },
    {
      "evidence_verified": true,
      "id": "current-defects",
      "measurement_reliability": 0.95,
      "observed_at_ms": 125,
      "present": false,
      "signal_id": "defects"
    }
  ],
  "diagnostic_observations": [
    {
      "case_id": "case-0",
      "evidence_verified": true,
      "id": "observation-case-0-queue",
      "observed_at_ms": 0,
      "present": true,
      "signal_id": "queue"
    },
    {
      "case_id": "case-0",
      "evidence_verified": true,
      "id": "observation-case-0-defects",
      "observed_at_ms": 0,
      "present": false,
      "signal_id": "defects"
    },
    {
      "case_id": "case-1",
      "evidence_verified": true,
      "id": "observation-case-1-queue",
      "observed_at_ms": 10,
      "present": false,
      "signal_id": "queue"

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": {
    "as_of_ms": 130,
    "holdout_fraction": 0.2,
    "minimum_log_loss_improvement": 0,
    "prior_pooling_strength": 5,
    "random_seed": 17,
    "signal_pooling_strength": 10,
    "target_scope_id": "company"
  },
  "decision": "use_ranked_hypotheses",
  "failed_gates": [],
  "guardrails": [
    "A ranked hypothesis is a calibrated diagnosis, not a causal finding. Escalate causal language only after intervention, quasi-experimental or mechanistic evidence passes a separate causal-attribution review.",
    "Resolved incidents must include every signal as present or absent. Missing telemetry cannot be encoded as a negative observation, and post-resolution signals are excluded to prevent label leakage.",
    "The unknown/other hypothesis prevents forced certainty. Do not suppress it, collapse compound mechanisms, or identify a person as the cause from aggregate behavioral telemetry."
  ],
  "hypothesis_ranking": [
    {
      "bayes_update_ratio": 1.452,
      "hypothesis_id": "capacity",
      "is_unknown": false,
      "posterior_interval_lower": 0.2137,
      "posterior_interval_upper": 0.7513,
      "posterior_median": 0.4775,
      "posterior_probability": 0.484,
      "prior_probability": 0.3333
    },
    {
      "bayes_update_ratio": 0.9215,
      "hypothesis_id": "unknown",
      "is_unknown": true,
      "posterior_interval_lower": 0.096,
      "posterior_interval_upper": 0.5667,
      "posterior_median": 0.3122,
      "posterior_probability": 0.3072,
      "prior_probability": 0.3333
    },
    {
      "bayes_update_ratio": 0.6265,
      "hypothesis_id": "quality",
      "is_unknown": false,
      "posterior_interval_lower": 0.0586,
      "posterior_interval_upper": 0.4447,

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 Build complete mature resolved incidents with one explicit present-or-absent row for every governed signal; reserve the latest local cases for a strict temporal holdout.
2. 2 Estimate company-local hypothesis priors and signal likelihoods with governed shrinkage toward peer evidence, then compare held-out multiclass log loss with a prior-only baseline.
3. 3 Condition Dirichlet-Beta posterior draws on current verified signals and their measurement reliability, preserving an explicit unknown/other state and uncertainty intervals.

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": "rank competing compound and unknown root" }
  → finds "infer_competing_root_cause_posterior"

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

gitrevio_capability_run
  { "capability_id": "infer_competing_root_cause_posterior", "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 →