Forecast incident learning debt economics

Forecast how much corrective-action debt will remain open and what recurrent incident and operating loss it may create using hierarchical closure, recurrence and severity models that must beat global baselines on the latest whole period.

Forecasting & survival Forecast CTOs Tech CEOs IT managers Team leads Investors

What it's for

Gives CTOs and investors a financial answer to a neglected reliability question: how much post-incident learning debt is likely to survive, recur and cost—not merely how many tickets are still open.

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
current_actions array of objects (7 fields) Evidence Yes
historical_action_periods array of objects (15 fields) Evidence Yes
horizon_days number ≥ 1 Your calibration Yes
max_detail_rows integer ≥ 1, ≤ 500 Numerical control Optional
maximum_closure_ece number ≥ 0, ≤ 1 Your calibration Optional
minimum_local_actions integer ≥ 1 Your calibration Optional
minimum_log_loss_improvement number Your calibration Optional
minimum_recurrence_mae_improvement number Your calibration Optional
minimum_training_periods integer ≥ 1 Your calibration Optional
scenarios array of objects (6 fields) Evidence Optional
seed integer ≥ 0 Numerical control Optional
simulations integer ≥ 200, ≤ 100000 Numerical control Optional
tail_probability number > 0.5, ≤ 1 Your calibration Optional

Each historical_action_periods record

Field Type Required
closed_exposure_days number (≥ 0) Yes
closed_recurrence_count integer (≥ 0) Yes
closure_log_days_sq_sum number (≥ 0) Yes
closure_log_days_sum number Yes
context_id string (non-empty) Yes
eligible_action_count integer (≥ 1) Yes
evidence_verified boolean Yes
id string (non-empty) Yes
open_exposure_days number (≥ 0) Yes
open_recurrence_count integer (≥ 0) Yes
outcome_mature boolean Yes
period_index integer (≥ 0) Yes
recurrence_loss_log_sq_sum number (≥ 0) Yes
recurrence_loss_log_sum number Yes
verified_closed_count integer (≥ 0) Yes
Example input 
{
  "current_actions": [
    {
      "age_days": 25,
      "completion_cost": 2000,
      "context_id": "risky",
      "evidence_verified": true,
      "id": "checkout-timeout-action",
      "open_drag_cost_per_day": 100,
      "overdue": true
    }
  ],
  "historical_action_periods": [
    {
      "closed_exposure_days": 200,
      "closed_recurrence_count": 0,
      "closure_log_days_sq_sum": 46.62522709164422,
      "closure_log_days_sum": 28.969882423813804,
      "context_id": "stable",
      "eligible_action_count": 20,
      "evidence_verified": true,
      "id": "stable-0",
      "open_exposure_days": 200,
      "open_recurrence_count": 1,
      "outcome_mature": true,
      "period_index": 0,
      "recurrence_loss_log_sq_sum": 47.71708299430558,
      "recurrence_loss_log_sum": 6.907755278982137,
      "verified_closed_count": 18
    },
    {
      "closed_exposure_days": 200,
      "closed_recurrence_count": 2,
      "closure_log_days_sq_sum": 17.948823709625927,
      "closure_log_days_sum": 5.991464547107982,
      "context_id": "risky",
      "eligible_action_count": 20,
      "evidence_verified": true,
      "id": "risky-0",
      "open_exposure_days": 200,
      "open_recurrence_count": 12,
      "outcome_mature": true,
      "period_index": 0,
      "recurrence_loss_log_sq_sum": 1015.5961180386997,

Truncated for display — the full payload is 120 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.

Example output 
{
  "action_forecasts": [
    {
      "action_id": "checkout-timeout-action",
      "context_id": "risky",
      "expected_cost": 10344.9294,
      "expected_recurrences": 1.494,
      "locally_supported": true,
      "probability_outstanding": 0.778
    }
  ],
  "configuration": {
    "horizon_days": 30,
    "scenario_count": 1,
    "seed": 53,
    "simulations": 500,
    "tail_probability": 0.95
  },
  "decision": "forecast_available",
  "detail_truncated": false,
  "evidence_diagnostics": {
    "immature_history_rows": 0,
    "unsupported_current_contexts": [],
    "unsupported_holdout_contexts": [],
    "unverified_current_actions": 0,
    "unverified_history_rows": 0
  },
  "interpretation": "Observed recurrence rates before and after verified closure are predictive states, not causal effects of closure. The forecast is aggregate decision support and does not assess individuals or authorize action.",
  "method": "hierarchical_beta_gamma_poisson_lognormal_incident_learning_debt_simulation",
  "summary": {
    "conditional_value_at_risk": 31290.3926,
    "current_action_count": 1,
    "expected_outstanding_actions": 0.778,
    "expected_recurrences": 1.494,
    "expected_total_cost": 10344.9294,
    "value_at_risk": 26772.8251
  },
  "validation": {
    "closure_ece": 0.0923,
    "closure_log_loss_improvement": 0.3361,
    "context_closure_log_loss": 0.3571,
    "context_recurrence_mae": 0.8478,
    "global_closure_log_loss": 0.6931,
    "global_recurrence_mae": 6.5,

Truncated for display — the full payload is 51 lines.

How it works

Forecasting & survival — Estimate when something completes or fails, with censoring and unresolved work handled honestly rather than dropped.

  1. 1 Build zero-inclusive context-period aggregates with verified closures, open/closed exposure, recurrence counts and log-moment sufficient statistics; reserve the latest mature whole period before fitting.
  2. 2 Partially pool Beta closure probabilities, Gamma-Poisson open/closed recurrence rates and log-normal closure/loss moments by local context, then require context predictions to beat global closure-log-loss and recurrence-MAE baselines with acceptable calibration.
  3. 3 Replay every current action under one coherent closure, frequency, severity and drag scenario per simulation; report outstanding-action, recurrence, expected-cost, VaR and CVaR distributions with unsupported contexts and abstention visible.

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.
  • Periods are zero-inclusive and mature, context/state definitions are stable, open and verified-closed exposure clocks are complete, recurrence families are consistently linked, financial loss is incremental and commensurable, and scenarios preserve common operating shocks.
  • A predictive interval or risk estimate is not a deadline promise, causal explanation, or individual-performance judgment.
  • Open-versus-closed recurrence is predictive, not causal proof that closure prevented an event. Unsupported or losing models remain diagnostic; no forecast scores people or authorizes corrective work.

Minimum evidence

  • historical_action_periods: required and organization-defined
  • current_actions: required and organization-defined
  • horizon_days: required and organization-defined

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

  • stable company-local action-context period spine joined point-in-time to postmortem action versions, verified control closure, mature recurrence-family incidents, finance-reconciled incremental loss and one coherent future scenario set
  • action/context/state and recurrence-family taxonomy, period and exposure clocks, closure verification, outcome maturity, loss/drag perimeter, holdout policy, minimum local support, prior and baseline gates, common scenarios, horizon and tail appetite

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 how much correctiveaction debt will" }
  → finds "forecast_incident_learning_debt_economics"

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

gitrevio_capability_run
  { "capability_id": "forecast_incident_learning_debt_economics", "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

Audit incident learning evidence integrity

Audit the complete point-in-time incident-to-postmortem-to-corrective-action lineage, separating missing or contradictory evidence from genuine overdue learning debt without attributing individual fault.

Statistical audit & measurement

Optimize incident learning portfolio

Choose immediate remediation or a predeclared experiment-contingent action for each failure mode using Bayesian value of information, prospective test accuracy and causal remediation effects, common scenarios, shared-loss accounting, Pareto search and hard cost, capacity and tail-risk gates.

Causal inference & experiment design

Audit CI pipeline evidence integrity

Audit the complete point-in-time change-to-pipeline-to-job-to-rerun cohort, exposing missing CI, orphan records, future leakage, inconsistent required-job outcomes, incomplete provider evidence and same-configuration fail-then-pass flake proxies without scoring people.

Statistical audit & measurement

Audit operational alert decision integrity

Audit every point-in-time operational alert evaluation by recomputing fire/suppress decisions and verifying effective policy, cooldown, evidence freshness, context, controls, severity routing, acknowledgement, action and mature outcome lineage.

Constrained optimization

Audit root cause traceback evidence integrity

Audit whether an anomaly traceback is complete, point-in-time, multiplicity-controlled and honestly labeled as temporal or causal, including every upstream candidate, path lag, edge identification basis and later root-recovery validation.

Causal inference & experiment design

Estimate software reliability growth

Estimate long-run software reliability growth with a power-law nonhomogeneous Poisson process, bootstrap trend evidence, and future incident exposure.

Forecasting & survival

See every tool in Quality, incidents & reliability →

Ready to See Your Engineering work clearly?

Request a free demo