Skip to content

v22.0 Phase 0 Baseline Metric Definitions

This document is the canonical metric dictionary for v22.0 Phase 0.

Source strategy definitions:

  1. v22.0 Non-Duplicate Value Decision Plan
  2. v22.0 Phase 0 Implementation Plan

Baseline Window

Use a fixed 4-week baseline window immediately before pilot start:

  1. baseline_start: YYYY-MM-DD (inclusive)
  2. baseline_end: YYYY-MM-DD (inclusive)

All baseline calculations must:

  1. use the same window for all primary metrics,
  2. be reproducible from saved query specs,
  3. log query version and execution date.

Phase 0 Computability Status (2026-04-01)

Gate 0 minimum mode remains the default interpretation for historical baseline reporting, but the required Phase 0 instrumentation dependencies now exist in the pilot schema and internal API layer. Metrics M2, M4, M5, M6, and M7 are computable once the new source tables are migrated and populated for a pilot cycle.

Metric Computability Now Notes
M1 Computable Uses pilot_contact_attempt_events directly.
M2 Conditionally computable Uses pilot_connection_events plus contact-attempt/referral anchor timestamps after recompute.
M3 Computable Uses pilot_referral_events directly.
M4 Conditionally computable Uses pilot_service_scope plus service_operational_status_events recency by SLA tier.
M5 Conditionally computable Uses pilot_contact_attempt_events.entity_key_hash for repeat-failure attribution.
M6 Conditionally computable Uses pilot_data_decay_audits after audits are recorded for the pilot window.
M7 Conditionally computable Uses pilot_preference_fit_events after cohort events are recorded.

Metric Dictionary

Metric ID Name Formula Numerator Source Denominator Source Reporting Minimum Data Quality
M1 Failed Contact Rate failed_contact_events / total_contact_attempts Contact attempts with failure outcomes All contact attempts rate + trend >=95% outcome coding completeness
M2 Time to Successful Connection timestamp(successful_connection) - timestamp(initial_contact_attempt_or_referral) Successful connection events Initial contact-attempt/referral anchor events median, p50/p75/p90 >=90% valid timestamps
M3 Referral Completion Capture Rate referrals_with_terminal_state / total_referrals Referrals in terminal states Total referrals created rate + trend >=95% referral state completeness
M4 Freshness SLA Compliance services_meeting_status_sla / pilot_services_total Services updated within SLA by tier Pilot services in scope rate 100% service SLA tier assignment
M5 Repeat Failure Rate hashed_entities_with_2plus_failures / distinct_hashed_entities Distinct entity_key_hash values with 2+ failed attempts Distinct non-null entity_key_hash values rate >=95% stable entity hash coverage
M6 Data-Decay Fatal Error Rate records_with_access_blocking_errors / records_sampled Fatal errors from verification sample Records sampled in period rate + severity mix dual-source verification completed
M7 Preference-Fit Indicator cohort_tasks_preferably_completed_via_careconnect / cohort_total_tasks Tasks completed via CareConnect in target cohort All tracked cohort tasks rate cohort attribution completeness >=90%

Enumerations (Locked For Phase 0)

Failed contact outcomes (M1)

  1. disconnected_number
  2. no_response
  3. intake_unavailable
  4. invalid_routing
  5. other_failure

Terminal referral states (M3)

  1. connected
  2. failed
  3. client_withdrew
  4. no_response_timeout

Fatal data-decay categories (M6)

  1. wrong_or_disconnected_phone
  2. invalid_or_defunct_intake_path
  3. materially_incorrect_eligibility
  4. service_closed_or_unavailable_but_listed_available

Data Source Mapping (Planned Tables/Views)

Metric Primary Source Secondary Source
M1 pilot_contact_attempt_events pilot_metric_snapshots
M2 pilot_connection_events pilot_metric_snapshots
M3 pilot_referral_events pilot_metric_snapshots
M4 pilot_service_scope + service_operational_status_events pilot_metric_snapshots
M5 pilot_contact_attempt_events + entity_key_hash pilot_metric_snapshots
M6 pilot_data_decay_audits Manual verification logs
M7 pilot_preference_fit_events pilot_metric_snapshots

Quality Gates For Baseline Acceptance (Gate 0 Minimum Mode)

All must pass before Gate 0:

  1. Query reproducibility confirmed for executable metrics.
  2. M1 and M3 baseline values are reported (or explicitly null when no pilot events exist).
  3. M2/M4/M5/M6/M7 are either reported from populated instrumentation sources or explicitly marked N/A because the pilot window has no qualifying source events yet.
  4. Missingness and exclusion rates are documented per metric.
  5. Any known bias/confounders are listed in notes.

Required Output Artifact

Generate one baseline report with:

  1. metric value table (M1-M7),
  2. confidence caveats and known limitations,
  3. query version references,
  4. sign-off section for product + governance owners.

Current artifact:

  1. v22.0 Phase 0 Baseline Report (2026-03-09)