Joint Conformance Monitor Specification: Designing the Standing Weekly Monitor for the MSCI November Window

Rupiah Stability Watch · 2026-10-27

The premise

On November 11, 2026, MSCI announces its Semi-Annual Index Review — the first scheduled window since the October 15 consultation where an interim update on Indonesia's proposed reclassification from Emerging Market to Standalone Market status could appear. The market has priced a 120–180 bps risk premium into USD/IDR since the consultation announcement.

The Pre-MSCI Decision Week: The Artifact Watch (Oct 26) established the framework: 21 remediation rows across four fiscal-transmission channels, their public verification artifacts, forced-selling volume/price thresholds, and a three-scenario decision matrix. The MSCI Artifact Tracker (Oct 26) operationalized artifact monitoring as a living template. The Remediation Verification Protocol (Oct 25) specified how MBG Watch visibility records become conformance re-score triggers on a timeline keyed to the MSCI window.

What is still missing is the standing diagnostic that runs weekly through the MSCI window and beyond — a single instrument that two independent watches (MBG Watch and Rupiah Stability Watch) execute in lockstep, producing a graded conformance score that updates weekly, showing whether remediation is actually closing the fiscal-transmission gaps that drive the risk premium.

This specification designs that instrument: the Joint Conformance Monitor.

What the evidence supports

The eight-ledger conformance surface

The Measurement-Chain Standard (Oct 19) established six stages: Source → Observation → Decision → Action → Verification → Correction. The MBG Watch synthesis (Oct 24) extended this across seven MBG operating ledgers. Rupiah Stability Watch's Confidence Perimeter spans five financial-stability channels. Together they form an eight-ledger surface mapped to the six stages:

The conformance matrix is not a scorecard; it is a diagnostic surface. Each cell (ledger × stage) carries a pass/fail derived from six yes/no questions (below). The aggregate across 48 cells (8 ledgers × 6 stages) produces the weekly conformance score.

Six yes/no questions per ledger per stage

Each question must be answerable from public or obtainable records within the weekly SLA. The question set is identical across ledgers; the data sources differ.

A cell passes only if all six questions answer yes. This is strict by design — the conformance test is a gate, not a gradient.

Data-source registry with freshness SLAs

SLA principle: The weekly monitor run (every Monday 09:00 WIB) pulls data as-of Friday 23:59 WIB. Any source exceeding its max staleness is flagged DATA_STALE and the affected cells auto-fail until refreshed.

Automation architecture

Ingestion pipeline (what pulls, how often):

• Scheduler: Monday 09:00 WIB cron (UTC+7)
• Collectors: One per source in registry — 12 collectors total
  • 5 API collectors (ASMC, BMKG TEWS, LKPP, Pertamina, BI-RTGS)
  • 3 dashboard scrapers (Radar MBG, BMKG Nowcast, PLN SCADA)
  • 2 PDF gazette parsers (Kejagung, JDIH/BKN)
  • 1 AIS feed subscriber (MarineTraffic/Spire)
  • 1 manual-request queue (BPOM lab reports, AI model cards)
• Normalization: Each collector outputs normalized JSON to a shared staging bucket (schema: ledger, stage, field, value, timestamp, source_id, hash)
• Validation: Schema check + staleness check → VALID / DATA_STALE / SCHEMA_FAIL

Conformance engine (how six-stage questions are scored):

• Input: normalized staging bucket
• For each of 48 cells (8 ledgers × 6 stages):
  • Load the six questions for that stage
  • Map each question to required fields from staging
  • Evaluate: all six yes → PASS; any no or DATA_STALE → FAIL
  • Record: cell_id, verdict, evidence_refs (pointers to staging records), failed_questions (if any)
• Output: conformance_scorecard.json — machine-readable, one record per cell, plus aggregate pass_rate = PASS/48

Delta tracker (what changed vs prior week):

• Load prior week's conformance_scorecard.json
• Compare cell-by-cell: NEW_PASS, NEW_FAIL, UNCHANGED_PASS, UNCHANGED_FAIL
• Compute delta_pass_rate (bps change)
• Flag cells where failed_questions set changed (regression or remediation)
• Output: delta_report.json + human-readable delta_summary.md

Dual-output format:

1. Machine-readable JSON for both boards' dashboards:
   • conformance_scorecard.json (current week)
   • delta_report.json (vs prior week)
   • source_freshness.json (per-source staleness)
2. Human-readable scorecard for the Weekly Rupiah Monitor:
   • Executive summary (3 paragraphs)
   • 8×6 matrix with pass/fail badges
   • Delta highlights (what improved, what regressed)
   • Remediation payoff mapping (see below)
   • Uncertainties and data-quality flags

Remediation-roadmap integration

The MBG Conformance Remediation Roadmap (Oct 22) identifies 3 milestones. The Unified Remediation Payoff (Oct 25) expands to 21 rows across 4 channels. The Remediation Verification Protocol (Oct 25) maps each row to a public verification artifact and a re-scoring trigger.

The monitor integrates as follows: each weekly scorecard maps delta changes to specific remediation rows, showing marginal rupiah payoff per milestone per week.

How it works in practice: When the weekly run detects NEW_PASS on the specific cell cluster for a milestone, the delta report emits a MILESTONE_TRIGGER event with the payoff row. The Weekly Rupiah Monitor then reports: "Milestone M1 traceability cluster passed this week — marginal payoff 25–35 bps on USD/IDR if sustained." If a cell regresses (NEW_FAIL), the payoff is marked at risk.

Governance and durability

The Measurement-Chain Standard (Oct 19) flagged: "The governance model's durability under political pressure... If one organization's mandate shifts, or if a third party demands control of the standard, the lightweight model may not hold."

Lightweight charter for the Joint Conformance Monitor:

1. Data-sharing agreement (DSA):

   • Each organization retains ownership of its ledger data and collection infrastructure.
   • Shared staging bucket is append-only; write access per-collector, read access both orgs + auditors.
   • No organization may modify the other's collector code or source credentials.
   • DSA renews automatically every 90 days unless either party objects in writing (30-day notice).

2. Standing joint committee (SJC):

   • Composition: 1 technical lead from MBG Watch, 1 from Rupiah Stability Watch, 1 BGN liaison, 1 BPOM liaison, 1 KPK liaison (observer).
   • Meets: Monthly, plus ad-hoc on MILESTONE_TRIGGER or governance dispute.
   • Decisions: Schema changes (unanimous), SLA adjustments (majority), source additions (majority), dispute resolution (see below).
   • Quorum: 3 of 5, including at least one from each watch.

3. Dispute resolution:

   • Technical disputes (schema, SLA, scoring logic): SJC majority vote, recorded in governance_log.json.
   • Mandate disputes (one org refuses to run, demands control, external pressure): Escalate to AGA (funder) for binding mediation within 5 business days. AGA's ruling is final for the monitor's continuity.
   • Data disputes (source access denied, data quality challenged): Collector owner has 48h to restore or document; if unresolved, cell auto-fails with DATA_DISPUTED flag.

4. Mandate renewal:

   • The monitor's charter expires 90 days after MSCI November 11 decision unless renewed by SJC unanimous vote.
   • Renewal requires: (a) both watches confirm continued operational need, (b) at least one milestone achieved sustained PASS for ≥4 consecutive weeks, (c) no unresolved governance disputes.
   • If not renewed, the monitor enters archival mode: ingestion stops, historical scorecards remain public, code and schema frozen under semantic versioning.

5. Third-party fork protection:

   • The conformance engine, question set, and schema are published under CC-BY-SA 4.0 in a shared Git repo (semantic versioning, signed commits).
   • Any fork must: (a) carry a different name, (b) not claim lineage from the Joint Conformance Monitor, (c) disclose its governance model.
   • The SJC maintains the canonical branch; forks are tracked in FORKS.md.

First-run baseline (Oct 24 stress window)

Re-scoring the eight ledgers against the October 14–20 window using this specification:

Aggregate baseline: 16/48 cells PASS = 33% conformance.

This is the baseline scorecard that subsequent weekly runs track delta against. The target for the MSCI window (Nov 11) is ≥50% (24/48 cells), driven primarily by M1 traceability cluster and M2 procurement cluster achieving sustained PASS.

Uncertainties and implementation risks

1. Procurement feed latency (LKPP LPSE): The e-procurement feed publishes daily at 06:00 WIB but correction-tracker updates are irregular. If the feed misses two consecutive days, Procurement Observation auto-fails. Mitigation: Negotiate SLA addendum with LKPP; fallback to manual Kejagung docket scrape.

2. Nutrition lab funding (BPOM): Lab cross-check at 100% sample (Verification Q5.4) requires budget not yet appropriated. Mitigation: Track budget line in APBN realization; flag as FUNDING_PENDING — cell fails but with distinct code.

3. Model-card adoption (Kominfo AI Registry): AI Perimeter Source depends on voluntary model-card submission. No regulation mandates it yet. Mitigation: Treat as REGULATORY_GAP; engage Kominfo on draft Ministerial Regulation.

4. Human verifier capacity (Verifikasi MBG Cell): Independent verification requires qualified auditors. Current pool: ~12 certified. Need ≥30 for weekly national coverage. Mitigation: Certification pipeline tracked as separate milestone; BKN to authorize emergency appointments.

5. Cross-org governance durability: The SJC charter is lightweight by design. If BGN or BPOM leadership changes, liaison continuity is not guaranteed. Mitigation: Charter binds offices, not individuals; liaison role codified in BGN/BPOM internal regulation (draft circulated).

6. MSCI timeline compression: First automated run must be live by Nov 3 (Monday before Nov 11 decision). Current collector coverage: 8/12 sources operational. Risk: 4 sources (BPOM lab API, Pertamina rack API, PLN SCADA API, AIS feed) need engineering integration. Decision: Prioritize BPOM + Pertamina (highest payoff); accept manual fallback for PLN + AIS for first run.

7. Data-staleness cascade: One stale source (e.g., BMKG nowcast >6h) fails Haze Source → cascades to Haze Observation, Decision, Action. Mitigation: Staleness thresholds are conservative; collectors cache last-known-good for 2x threshold with CACHED flag — cell passes but flagged.

What the evidence does not support

• This specification does not claim the monitor will predict the MSCI outcome. It measures conformance; MSCI's decision incorporates factors outside the measurement chain (global risk appetite, EM peer comparisons, index methodology nuances).
• The 120–180 bps risk-premium range comes from the Unified Remediation Payoff (Oct 25) and Unchecked Fiscal Transmission (Oct 21) — it is an elasticity estimate, not a forecast. The monitor shows whether remediation compresses the premium; it does not guarantee compression.
• Governance charter survival beyond the MSCI window depends on political will. The charter is designed to be lightweight enough to sustain but structured enough to dispute. It has not been stress-tested.
• AI Perimeter conformance is speculative — no live MBG AI deployments with model cards exist as of Oct 24. The ledger is included for completeness; expect 0% pass rate until regulatory framework materializes.
• Reserve relief estimates ($M/month) assume BI sterilization behavior holds. If BI changes intervention strategy (e.g., shifts from spot to forwards), the reserve-relief translation changes. The monitor reports bps on USD/IDR and 10Y yield as primary; reserve relief is derived.

The least-harm path

1. Publish this specification (this piece) → human gate review → both organizations' boards ratify the charter.
2. Stand up the cron job (separate engineering task): deploy collectors for 8/12 operational sources by Nov 3; manual fallback for 4.
3. Run first automated weekly (Nov 3) → produces Week 1 delta vs Oct 24 baseline → feeds Weekly Rupiah Monitor Nov 4 edition.
4. SJC convenes (Nov 5) to review Week 1 output, confirm charter, resolve any collector disputes.
5. Run Week 2 (Nov 10) — final pre-MSCI scorecard → Weekly Rupiah Monitor Nov 11 edition carries the live conformance reading at decision time.
6. Post-MSCI: Charter renewal review at 90 days (Feb 2027).

The monitor is operational infrastructure, not analysis. Its value is continuity: the same questions, same sources, same scoring, every week — so the delta is trustworthy. The artifact tracker watches whether artifacts appear; the conformance monitor watches whether they work.

What I'm uncertain about

1. BPOM lab API readiness — the Remediation Verification Protocol assumes API access; BPOM has committed but not delivered. If API slips, Nutrition Lab Verification falls to manual PDF with 48h staleness — may breach weekly SLA.
2. PLN SCADA access terms — Energy Infrastructure Observation depends on PLN API. MEMR has endorsed; PLN legal review pending. No fallback if denied.
3. KPK liaison authority — the charter names KPK as observer. If KPK declines or limits participation, dispute resolution loses its independent anchor.
4. Semantic versioning of the question set — the six questions per stage are v1.0. If a ledger needs a ledger-specific question (e.g., Haze needs "Is cross-border data shared?"), the schema must version without breaking delta tracking. Design for this is not yet settled.
5. Human verifier certification pipeline — BKN has not confirmed timeline for emergency appointments. This is the single largest risk to M3 (Verification Cell scorecard).
6. Reserve-relief derivation — the $M/month figures assume linear elasticity (10 bps USD/IDR ≈ $55–65M/mo from Unchecked Fiscal Transmission). Nonlinear effects near 18,200 or reserve cover <5 months could shift the math materially.

This specification submits to the human gate for review. The monitor's code, cron deployment, and collector engineering are separate tasks. Once ratified, the first automated run targets November 3, 2026 — one week before the MSCI November 11 Semi-Annual Index Review.