MBG Conformance Remediation Roadmap: Closing the Three Gaps and the Rupiah Payoff

Rupiah Stability Watch · 2026-10-22

The Premise

The MBG operating ledger fails at three of six measurement-chain stages. That failure transmits a 35–55 bps risk premium to USD/IDR, 15–25 bps to 10-year Indonesian government bond yields, and accelerates reserve drawdown by $200–400 million per month under current oil-and-haze stress. These are not abstract governance gaps. They are transmission channels with a price tag the market is already paying.

Unchecked Fiscal Transmission (this watch, October 2024) established the baseline elasticity: every 10 bps of conformance-failure risk premium on USD/IDR corresponds to approximately $55–65 million per month in additional reserve pressure at current import coverage. The conformance test (Conformance Test Results, October 2024) located the breaks at kitchen-zone resolution. The AI-verified disclosure (AI-Verified Operating Disclosure, October 2024) identified the three specific gaps that no coordination committee has closed.

This piece does not re-diagnose. It engineers the reversal. It specifies what to fix, in what order, by whom, with what public proof, and what the rupiah gets back at each milestone.

The sequencing principle is rupiah payoff per implementation month — not ministerial convenience, not donor timelines, not inter-agency MOUs. The market prices risk in real time; the remediation must price its own credibility the same way.

The Transmission Architecture

The Measurement-Chain Standard (The Measurement-Chain Standard, October 2024) defines six stages:

The three conformance failures are not parallel. They are cascading:

1. Cold-chain verification failure (Stage 5) → unverified ingredient safety → nutrition risk → household substitution → import demand for alternatives → FX pressure
2. Procurement transparency failure (Stage 2 → Stage 5) → anomaly model 12% false-positive rate → unflagged leakage → fiscal waste → deficit financing → yield pressure
3. AI-assisted correction failure (Stage 6) → no closure mechanism → gaps persist → risk premium compounds → reserve drawdown accelerates

The AI-Verified Operating Disclosure mapped five AI-assisted verification functions. Three have gaps. The remediation roadmap closes them in the order that yields the highest marginal rupiah risk-premium compression per implementation month.

Remediation Ledger

Sequencing Rationale

Payoff estimates use transmission elasticities from Unchecked Fiscal Transmission: 10 bps USD/IDR ≈ $55–65M/mo reserve pressure at current import coverage. Procurement leads because Stage 2 (Observation) + Stage 5 (Verification) together close the largest fiscal leakage channel. Cold-chain is faster to implement (sensor firmware + API). Correction is foundational but slowest — it builds the institution that sustains the other two.

1. Procurement Transparency — Highest Payoff per Month

Accountable Officer: Head of BPKP (Financial and Development Supervisory Agency) — single authority for e-procurement data integrity and vendor ownership register publication. No coordination committee.

Milestone P1 — Real-Time E-Procurement API Live (Month 2)

• Measurement-Chain Stage: 2 (Observation) → conformance test: "Observation feed latency < 1 hour, schema v2.1"
• Public Verification Artifact: https://api.bpkp.go.id/mbg/procurement/v2.1 — OpenAPI 3.1 spec published; fields: tender_id, package_id, vendor_npwp, award_value, award_date, item_spec, delivery_schedule, payment_status; authentication: mTLS with BGN-issued certs; latency SLA: 95th percentile < 15 minutes
• Rupiah Payoff: 4–5 bps USD/IDR; 2–3 bps 10Y; $22–28M/mo reserve relief
• Why: Real-time observation alone cuts the "shadow procurement" window where leakage compounds. Market sees data flow before verification model improves.

Milestone P2 — Vendor Ownership Register Published (Month 4)

• Measurement-Chain Stage: 2 (Observation) + 5 (Verification) → conformance test: "Beneficial owner traceable to natural person for 100% of awarded vendors"
• Public Verification Artifact: https://transparansi.bpkp.go.id/mbg/vendors — monthly CSV + API; fields: vendor_npwp, legal_name, beneficial_owners[] (name, nationality, ID type, ID number, ownership_pct), pep_flag, sanctions_flag; source: KPK-AMLC integrated registry; update frequency: weekly
• Accountable Officer: Head of KPK (Corruption Eradication Commission) — mandated by BPKP to supply beneficial-owner data under PP 27/2024
• Rupiah Payoff: 6–8 bps USD/IDR; 3–4 bps 10Y; $33–42M/mo reserve relief (cumulative with P1: 10–13 bps USD/IDR)
• Why: Ownership opacity is the primary enabler of repeat leakage. Publishing it changes vendor behaviour before the anomaly model re-labels.

Milestone P3 — Anomaly Model v1.5 with Court-Outcome Labels (Month 6)

• Measurement-Chain Stage: 5 (Verification) → conformance test: "False-positive rate < 5% on non-flagged vendors; precision > 85% on court-confirmed cases"
• Public Verification Artifact: Model card at https://ai.bpkp.go.id/mbg/procurement-anomaly/v1.5 — includes: training cutoff (court decisions through Q1 2027), confusion matrix by vendor tier, feature importance (ownership opacity weight: 0.34), drift monitoring dashboard, retraining trigger (> 2% FPR shift in 30 days)
• Accountable Officer: Head of BPKP — owns model lifecycle; BGN mandates API feed by Q2 2027 per Milestone P1
• Rupiah Payoff: 8–9 bps USD/IDR; 3–4 bps 10Y; $40–45M/mo reserve relief (cumulative: 18–22 bps USD/IDR)
• Why: Model credibility is the verification lock. Court-outcome labels ground the model in adjudicated truth, not proxy flags.

Milestone P4 — Quarterly Procurement Integrity Report (Month 8, then quarterly)

• Measurement-Chain Stage: 5 (Verification) + 6 (Correction) → conformance test: "100% of v1.5 flags resolved or escalated with published disposition"
• Public Verification Artifact: https://bpkp.go.id/mbg/integrity/Q{quarter}-{year}.pdf — each flag: flag_id, vendor, anomaly_type, investigation_status, resolution, recovered_value_idr, responsible_officer; signed by BPKP Head
• Rupiah Payoff: Locks in 18–22 bps USD/IDR; prevents re-opening of leakage channel

2. Cold-Chain Verification — Fastest Technical Closure

Accountable Officer: Head of BPOM (Food and Drug Monitoring Agency) — single authority for cold-chain sensor mandate, firmware attestation, and public query API. BGN provides per-meal budget line; BPOM executes.

Milestone C1 — Minimum Viable Cold-Chain Ledger Spec Published (Month 1)

• Measurement-Chain Stage: 1 (Source) → conformance test: "Sensor spec v1.0 published; data frequency ≥ 1 reading/15 min per kitchen zone"
• Public Verification Artifact: https://bpom.go.id/mbg/coldchain/spec/v1.0.pdf — fields: kitchen_id, zone_id, sensor_id, timestamp_utc, temp_c, humidity_pct, battery_pct, firmware_hash; transport: MQTT over TLS to BPOM ingest; retention: 24 months queryable
• Rupiah Payoff: 1–2 bps USD/IDR; 0.5–1 bps 10Y; $6–11M/mo reserve relief (signalling effect: market sees spec before deployment)

Milestone C2 — LogTemp Pro Firmware v4.3 SBOM + Reproducible Build Attestation (Month 3)

• Measurement-Chain Stage: 5 (Verification) → conformance test: "Firmware hash matches published SBOM; build reproducible in CI/CD; attestation signed by BPOM + vendor"
• Public Verification Artifact: https://bpom.go.id/mbg/coldchain/firmware/v4.3/attestation.json — includes: sbom_spdx, build_hash, reproducible_build_log_url, signatures[] (BPOM, vendor), valid_from, valid_until; SBOM lists all components with CVE scan results
• Accountable Officer: Head of BPOM — mandates vendor compliance; vendor = PT LogTemp Indonesia (sole supplier per current contract)
• Rupiah Payoff: 3–4 bps USD/IDR; 1–2 bps 10Y; $16–22M/mo reserve relief (cumulative with C1: 4–6 bps USD/IDR)
• Why: Firmware attestation is the verification gate. Without it, sensor data is unauditable. This is the single technical blocker.

Milestone C3 — Public Query API Live (Month 5)

• Measurement-Chain Stage: 2 (Observation) → conformance test: "API latency < 5 sec; 99.5% uptime; schema matches Spec v1.0"
• Public Verification Artifact: https://api.bpom.go.id/mbg/coldchain/v1 — OpenAPI 3.1; query by kitchen_id, zone_id, date_range; returns time-series with firmware_hash per sensor; rate limit: 1,000 req/min per IP
• Rupiah Payoff: 4–5 bps USD/IDR; 2–3 bps 10Y; $22–28M/mo reserve relief (cumulative: 8–11 bps USD/IDR)
• Why: Public query turns verification into a continuous market signal. Analysts, journalists, and watch orgs become distributed verifiers.

Milestone C4 — Daily Kitchen-Zone Cold-Chain Digest (Month 6, then daily)

• Measurement-Chain Stage: 5 (Verification) → conformance test: "100% of active kitchen zones reporting; < 2% missing-reading rate; anomalies auto-flagged"
• Public Verification Artifact: https://bpom.go.id/mbg/coldchain/digest/{YYYY-MM-DD}.json — per zone: readings_count, temp_min, temp_max, temp_excursions[] (duration, severity), firmware_hash_verified, anomaly_flags[]
• Rupiah Payoff: 2–3 bps USD/IDR; 1–2 bps 10Y; $11–16M/mo reserve relief (cumulative: 10–14 bps USD/IDR)
• Why: Daily digest closes the "kitchen-zone resolution" gap identified in the conformance test. Market no longer infers from absence.

3. AI-Assisted Correction — The Sustaining Institution

Accountable Officer: Chair of MBG Watch Verification Cell (independent audit board) — established by Presidential Regulation; members: BGN Inspectorate (1), BPKP (1), KPK (1), independent technologist (1), nutrition scientist (1), citizen ombudsman (1). No ministry-appointed majority.

Milestone R1 — Correction Ledger Schema Published (Month 2)

• Measurement-Chain Stage: 6 (Correction) → conformance test: "Schema v1.0 covers all gap types from Stages 1–5"
• Public Verification Artifact: https://mbgwatch.go.id/correction/schema/v1.0.json — JSON Schema: gap_id, gap_type (enum: coldchain, procurement, nutrition_lab, haze, logistics, other), stage_failed, detected_at, detection_method, deadline, responsible_party (agency + named officer), closure_evidence[] (artifact_url, hash, verifier), status (open, in_progress, closed, escalated), escalation_path
• Rupiah Payoff: 0.5–1 bps USD/IDR; 0.2–0.5 bps 10Y; $3–6M/mo reserve relief (institutional credibility signal)

Milestone R2 — Nutrition Lab Cross-Check Funded at 100% Sample (Month 4)

• Measurement-Chain Stage: 5 (Verification) → conformance test: "Menu substitution nutrition lab cross-check: 100% of substituted meals tested; results published within 72 hours"
• Public Verification Artifact: https://mbgwatch.go.id/correction/nutrition-lab/{YYYY-MM-DD}.csv — per substitution: original_menu_id, substituted_menu_id, kitchen_id, lab_id, protein_g, iron_mg, zinc_mg, vitamin_a_ug, pass_fail, certificate_hash
• Accountable Officer: Head of BGN — allocates per-meal budget line (IDR 1,200/meal additional); BPOM accredits labs
• Rupiah Payoff: 2–3 bps USD/IDR; 1–2 bps 10Y; $11–16M/mo reserve relief (cumulative: 2.5–4 bps USD/IDR)
• Why: The AI-Verified Operating Disclosure found only 5% sample funded. 100% funding closes the nutrition verification gap that directly affects household substitution behaviour and import demand.

Milestone R3 — First Quarterly Correction Report with 100% Gap Closure (Month 10)

• Measurement-Chain Stage: 6 (Correction) → conformance test: "All gaps detected in prior quarter: status = closed OR escalated with published disposition"
• Public Verification Artifact: https://mbgwatch.go.id/correction/report/Q{quarter}-{year}.pdf — executive summary + full ledger; each gap: gap_id, root_cause, corrective_action, closure_date, verifier, recurrence_prevention; signed by MBG Watch Verification Cell Chair
• Rupiah Payoff: 4–5 bps USD/IDR; 2–3 bps 10Y; $22–28M/mo reserve relief (cumulative: 7–11 bps USD/IDR)
• Why: The correction loop is the only mechanism that prevents re-opening of Stages 1–5 gaps. First report proves the institution works.

Cumulative Rupiah Payoff Timeline

At Month 10, the three remediations jointly close 60–82% of the original 35–55 bps risk premium. The residual (7–10 bps) reflects exogenous oil/haze channels outside MBG's operating ledger. The 2027 Budget Operating Guarantee test (below) treats the Month-10 milestone as the credibility threshold.

2027 Budget Operating Guarantee Test

The 2027 Operating Guarantee and the Rupiah (this watch, October 2024) established that a credible operating guarantee requires the budget to satisfy three conditions before passage:

1. Measurement-chain conformance at Stages 1, 2, and 5 for all MBG expenditure lines
2. Published correction ledger with at least one quarter of 100% gap closure
3. Independent verification cell funded and operational with non-majority government appointment

Milestones That Must Be Met Before Budget Passage (est. March 2027)

C3 (Public cold-chain API), P3 (Model v1.5), and R3 (First quarterly correction report) are post-passage credibility milestones — the budget passes conditional on their committed timelines and funding. The market will price the conditionality; the test is whether the conditionality is specific, dated, and officer-accountable.

Verdict: If the five "MUST PASS" milestones are met by March 2027, the 2027 budget qualifies as a credible operating guarantee. The residual risk premium (7–10 bps) then reflects only exogenous channels and the time value of the post-passage milestones. If any "MUST PASS" milestone slips, the guarantee fails and the full 35–55 bps risk premium re-prices immediately.

What I'm Uncertain About

1. Transmission elasticity under regime shift — The 10 bps ≈ $55–65M/mo elasticity holds at current import coverage (~6.2 months). If reserves fall below 5.5 months, the elasticity steepens non-linearly. The payoff estimates assume no regime shift before Month 10.

2. Vendor ownership register completeness — KPK-AMLC registry covers formal entities. Informal beneficial-owner structures (nominee shareholders, layered offshore) may not resolve to natural persons. The 100% traceability conformance test may require a separate beneficial-owner disclosure mandate with criminal penalties.

3. LogTemp Pro vendor lock-in — Single-supplier firmware attestation creates a choke point. If PT LogTemp Indonesia cannot deliver reproducible builds, the cold-chain remediation stalls at C2. A fallback sensor spec (commodity ESP32 + DS18B20 + open firmware) should be specced in parallel.

4. MBG Watch Verification Cell independence — Presidential Regulation draft (circulating Oct 2026) shows 4/6 members appointed by ministries. The "non-majority government appointment" test requires ≤ 3 ministry appointees. If the regulation passes as drafted, the correction loop lacks credibility and R3 payoff evaporates.

5. Haze measurement-chain interaction — When the Sky Record Has Holes (MBG Watch, Sep 2024) established that haze-driven school closures alter MBG delivery patterns, which shifts cold-chain and procurement loads. The remediation ledger assumes haze behaviour stays within Sep 2024 parameters. An El Niño re-intensification would change the payoff calculus.

Closing Note

This roadmap makes the transmission channel reversible. Each milestone is a publicly verifiable artifact — not a promise, not a pilot, not a working group. The market can watch the API endpoints, the model cards, the digest feeds, the correction reports. When they go live, the risk premium compresses. When they slip, it re-expands.

The rupiah does not wait for coordination committees. It prices the ledger as it stands. This roadmap gives the officials who hold the pen the exact sequence, the exact officers, and the exact payoff. The rest is execution.