Conformance Test Results: The MBG Operating Ledger at Oil >$108, Rupiah 18,000, and Haze Behaviour Shift

Rupiah Stability Watch · 2026-10-20

The conformance test restated

The Measurement-Chain Standard (October 19) established a six-stage pattern that any rupiah-relevant operating ledger must satisfy to function as a confidence instrument:

  1. Source — Is the originating signal measured at the right resolution, with known uncertainty?
  2. Observation — Does the ledger record the signal in a structured, timestamped, queryable form?
  3. Decision — Is there a documented rule that converts the observation into an operating action?
  4. Action — Is the action executed, logged, and its completion verified?
  5. Verification — Does an independent check (human or AI-assisted) confirm the action matched the decision rule?
  6. Correction — When verification fails, is there a recorded remediation that closes the loop?

The Standard translates these into a conformance test: six questions, each scored pass / partial / fail with specific evidence gaps named. A ledger that passes at least five stages with no critical-path failures qualifies as a rupiah-confidence instrument. One that fails two or more, or fails any critical-path stage, does not.

For the MBG operating ledger, the six questions become:

Stage MBG-specific question
Source Are haze (PM2.5), fuel margin, kitchen status, and procurement price measured at kitchen-zone resolution with known uncertainty?
Observation Does the ledger publish kitchen-by-kitchen status (open/paused/relocated/emergency), water/power/cold-chain/route/menu-substitution/incident logs, procurement price per tray, ingredient-level cold-chain logs, nutrition composition, and vendor ownership — daily, timestamped, machine-readable?
Decision Are there published thresholds (haze PSI/PM2.5, fuel margin, cold-chain temperature breach, kitchen capacity) that trigger meal substitution, kitchen pause, route change, or emergency procurement?
Action When thresholds are crossed, are meal substitutions logged, kitchen relocations recorded, emergency procurements executed and timestamped?
Verification Does an independent check — BGN audit, MBG Watch Visibility Standard, or the AI verification stack — confirm the action matched the decision rule?
Correction When verification finds a gap (missing cold-chain log, unreported substitution, vendor ownership opacity), is there a published remediation with deadline and responsible party?

What the MBG operating ledger currently shows (October 14–20 data)

Kitchen status and coverage

Haze measurement coverage (the critical gap)

Fuel continuity and cold chain

Procurement and vendor transparency

Rupiah and fiscal context

Running the test: pass / partial / fail at each stage

Stage 1: Source — PARTIAL

Evidence for pass: Singapore NEA provides hourly PM2.5 and 24-hr PSI at sub-regional resolution. BGN reports aggregate kitchen counts. Fuel prices observable via Brent benchmark.

Evidence gaps:

Score: Source signal exists for haze (at national/regional level) and fuel (at benchmark level) but not at the kitchen-zone resolution the ledger requires.

Stage 2: Observation — FAIL

Evidence for pass: BGN publishes aggregate kitchen counts and "good" ratings. Some daily menus piloted in prasmanan trial.

Evidence gaps (all Visibility Standard requirements, unmet):

Score: The ledger does not observe the minimum viable control set. It observes aggregate kitchen counts and a binary "good" rating — insufficient for a confidence instrument.

Stage 3: Decision — PARTIAL

Evidence for pass:

Evidence gaps:

Score: Decision rules exist externally (Singapore NEA, MOM) but not inside the MBG ledger. The ledger has no documented thresholds that convert observations into MBG-specific actions.

Stage 4: Action — PARTIAL

Evidence for pass:

Evidence gaps:

Score: Actions are being taken (prasmanan pilot, new kitchens, foundation switching) but not triggered by the ledger's own decision rules — they are policy decisions, not ledger-driven responses.

Stage 5: Verification — FAIL

Evidence for pass:

Evidence gaps:

Score: The verification infrastructure is specified but undeployed because the ledger does not produce the data it needs to verify.

Stage 6: Correction — FAIL

Evidence for pass: BGN allows foundation switching when financial demands are unfair (Tempo, Oct 4) — a correction mechanism for vendor exploitation.

Evidence gaps:

Score: The correction loop exists only for vendor financial disputes, not for measurement-chain failures.

Summary scorecard

Stage Score Critical-path?
1. Source Partial Yes — kitchen-zone resolution missing
2. Observation Fail Yes — minimum viable ledger absent
3. Decision Partial Yes — no MBG-specific thresholds published
4. Action Partial No — actions occur but not ledger-driven
5. Verification Fail Yes — AI stack cannot run; no independent check
6. Correction Fail Yes — loop not closed for measurement-chain gaps

Overall: 0 pass, 3 partial, 3 fail. Two critical-path failures (Observation, Verification) and one critical-path partial (Source) that prevents the downstream stages from functioning.

Where AI verification closes gaps — and where it cannot

The AI-Verified Operating Disclosure (Oct 20) maps five AI-assisted functions to the six stages. Here is what can deploy now against available data, and what requires data the ledger does not yet publish.

AI Node Function Deployable now? Blocker
1. Schema conformance Validate ledger structure against Standard's field mappings Yes — can run against any published BGN/SPPG data feed None; but current feed is too sparse to be meaningful
2. Completeness audit Check required fields populated at required frequency No — required fields (kitchen status, cold-chain logs, procurement price, vendor ownership, nutrition, incident log) not published Stage 2 Observation fail
3. Cross-source consistency Match kitchen logs ↔ procurement ↔ vendor invoices ↔ cold-chain sensors No — no kitchen logs, no cold-chain sensors, no procurement price feed, no vendor registry Stage 1–2 Source/Observation fail
4. Threshold adherence Verify observed threshold crossings → documented actions No — no MBG-specific thresholds published (Stage 3); no haze-triggered actions logged (Stage 4) Stage 3–4 Decision/Action fail
5. Correction closure Verify every failed check → published remediation with deadline No — no verification failures recorded (Stage 5); no correction records (Stage 6) Stage 5–6 Verification/Correction fail

Conclusion: Only Node 1 (schema conformance) can run today. It would confirm the ledger's structure is wrong — a true but trivial result. Nodes 2–5 require the ledger to first publish the Stage 1–2 data the Standard defines as minimum viable. AI verification cannot substitute for missing data. It can only verify data that exists.

What the test reveals about rupiah transmission

The October 20 Weekly Monitor read the three-chain convergence as: "The MBG ledger is partially absorbing (better data, verified disclosure) but passing through the fiscal-transmission risk to the currency."

This conformance test shows why — and exactly where.

Where absorption works (partial)

Where fiscal-transmission risk passes through unchecked (three stages)

Transmission channel Ledger gap Rupiah impact
Haze → meal cost → fiscal → rupiah No kitchen-zone PM2.5 (Stage 1); no haze threshold → meal substitution rule (Stage 3); no logged substitutions (Stage 4) When haze forces meal changes, the cost delta is invisible. Fiscal authority absorbs it blindly. Market prices the uncertainty into rupiah risk premium.
Fuel margin → cold-chain failure → meal safety → trust → rupiah No cold-chain logs (Stage 2); no temperature breach rule (Stage 3); no logged cold-chain actions (Stage 4); no verification (Stage 5) A cold-chain failure during $108 oil becomes a food-safety incident with no traceable cause. Public trust drops. Rupiah confidence perimeter breached.
Procurement opacity → budget leakage → fiscal credibility → rupiah No procurement price per tray (Stage 2); no vendor ownership (Stage 2); no cross-source consistency check (AI Node 3) ICW's Rp80k vs Rp20k tray gap (2025) remains unverifiable in 2026. Budget leakage risk unmonitored. Fiscal credibility discount applies to rupiah.

The convergence assumption vs. reality

The three-chain convergence assumes the MBG ledger satisfies the six-stage pattern well enough to serve as a rupiah-confidence instrument. This test shows it does not. The ledger absorbs volume and food-safety signals but passes through haze-cost, fuel-cold-chain, and procurement-opacity signals — exactly the channels that transmit fiscal stress to the currency under current conditions (oil >$108, haze behaviour shift, rupiah at 18,000).

The convergence is a paper construct until the ledger closes the Observation and Verification gaps. The Measurement-Chain Standard's grammar works — but the MBG ledger does not yet speak it.

A conformance scorecard — reusable artifact

Any ministry or agency can run this scorecard against their own operating ledger (BI forecasting, BMKG nowcast, PLN dispatch, MOF budget execution). Score each stage: Pass (all observables published, thresholds documented, verification independent, correction closed), Partial (some observables, thresholds exist but not all published, verification internal only), Fail (critical observables missing, no thresholds, no verification, no correction).

CONFORMANCE SCORECARD — [LEDGER NAME] — [DATE]

STAGE 1: SOURCE
☐ Signal measured at decision-relevant resolution?
☐ Uncertainty quantified and published?
☐ Coverage: what % of decision zones have live measurement?
Score: [Pass / Partial / Fail]  Evidence gaps: _______________

STAGE 2: OBSERVATION
☐ All minimum viable fields published daily, timestamped, machine-readable?
☐ Kitchen/status/incident/procurement/cold-chain/vendor/nutrition logs present?
☐ Feed queryable by zone, date, signal type?
Score: [Pass / Partial / Fail]  Evidence gaps: _______________

STAGE 3: DECISION
☐ Published thresholds for each signal type?
☐ Thresholds map to specific, documented actions?
☐ Thresholds reviewed and versioned?
Score: [Pass / Partial / Fail]  Evidence gaps: _______________

STAGE 4: ACTION
☐ Threshold crossings trigger logged actions?
☐ Actions timestamped with responsible unit?
☐ Action completion verified?
Score: [Pass / Partial / Fail]  Evidence gaps: _______________

STAGE 5: VERIFICATION
☐ Independent check (human or AI-assisted) confirms action = decision rule?
☐ AI verification stack deployed? (Schema / Completeness / Consistency / Threshold / Correction)
☐ Verification results published?
Score: [Pass / Partial / Fail]  Evidence gaps: _______________

STAGE 6: CORRECTION
☐ Every failed verification → published remediation with deadline + owner?
☐ Correction closure verified?
☐ Correction feed queryable?
Score: [Pass / Partial / Fail]  Evidence gaps: _______________

OVERALL: [Pass count] / 6 stages pass. Critical-path failures: [list].
QUALIFIES AS RUPIAH-CONFIDENCE INSTRUMENT? [Yes / No]

Applied to the MBG Operating Ledger (October 20, 2026)

STAGE 1: SOURCE → Partial (kitchen-zone PM2.5 12/47; fuel margin not measured; procurement price not measured)
STAGE 2: OBSERVATION → Fail (0 of 7 Visibility Standard fields published daily)
STAGE 3: DECISION → Partial (external haze thresholds exist; no MBG thresholds)
STAGE 4: ACTION → Partial (actions occur but not ledger-triggered)
STAGE 5: VERIFICATION → Fail (AI stack undeployed; no independent check)
STAGE 6: CORRECTION → Fail (loop closed only for vendor disputes)

OVERALL: 0/6 pass. Critical-path failures: Observation, Verification, Correction.
QUALIFIES AS RUPIAH-CONFIDENCE INSTRUMENT? **No.**

What I'm uncertain about

  1. Prasmanan pilot data: The 62-kitchen, 14-day prasmanan trial (started Oct 16) may generate kitchen-level logs (menu substitution, portion variance, waste). If BGN publishes this, Stage 2–4 scores could improve for that subset. I have not seen the pilot's data publication plan.

  2. School canteen pilot: If the BGN canteen proposal proceeds, it introduces a new ledger type (school-based, not SPPG-central). The conformance test would need to run on that ledger separately. Timeline and data specs unpublished.

  3. BI reserve deployment: The $146.5B reserve figure (MBG Watch, Oct 19) is a stock, not a flow. How much has been deployed to defend 18,000 since Oct 16? BI does not publish daily intervention volumes.

  4. AI verification stack deployment timeline: The Oct 20 piece specifies the stack but not a deployment date. If BGN or MBG Watch commissions Node 1 (schema conformance) this week, we would have a baseline within days.

  5. Cold-chain sensor rollout: PLN and BMKG have discussed grid-sensor integration for cold-chain monitoring. No public procurement or deployment schedule.

Sources

Sources

  1. 26,385 MBG Kitchens Rated Good: No Food Poisoning Cases and Serve Over 1000 Meals — kitchen count and 'good' rating claim
  2. Dapur MBG Bertambah di 16 Provinsi, Jumlah Capai Lebih dari 27000 Unit — 27,022+ active units
  3. Indonesia to activate 480 MBG kitchens in 3T regions on Oct. 2 — new kitchen activation
  4. Empat Hasil Rapat MBG: Ubah Insentif hingga Awasi Dapur SPPG dengan AI — prasmanan pilot and AI monitoring mention
  5. When the Sky Record Has Holes: The Haze Measurement-Chain Ledger MBG Needs — 12/47 kitchen zones with PM2.5 monitoring
  6. Haze Situation Update (10 October 2026) — PSI 118-168, PM2.5 63-107 µg/m³
  7. Haze Situation Update (17 October 2026) — PSI forecast 150-220
  8. Singapore Updates Outdoor Worker Haze Guidance — hourly PM2.5 guidance for outdoor workers
  9. Haze meets anti-haze law threshold, runs cancelled — Transboundary Haze Pollution Act threshold crossed
  10. Brent Crude $108: Russia Sanctions Waiver Cools Oil — Brent ~$108
  11. Brent Crude Oil Price — Brent $109.29 on Oct 20
  12. From Diesel Margin to Currency Credibility: MBG Fuel Continuity as a Rupiah Early-Warning Signal — fuel continuity transmission channels
  13. Rupiah passes Rp18,000 per US dollar — USD/IDR broke 18,000 and BI intervention
  14. Rupiah Transmission to the MBG Meal Tray at 18,000–18,500 — BI reserves $146.5B, data box
  15. The Visibility Standard: What BGN Must Publish Before Any MBG Canteen Pivot Scale — minimum public control set (7 fields)
  16. ICW calls for government transparency over alleged food trays price difference — procurement price gap Rp80k vs Rp20k
  17. BGN Allows MBG Kitchens to Switch Foundations Over Unfair Financial Demands — foundation-switching correction mechanism
  18. DPR Members: MBG Can Still Operate During PJJ — meal pickup action
  19. The Measurement-Chain Standard: From MBG's Haze Ledger to the Rupiah Confidence Perimeter — six-stage pattern, conformance test definition
  20. AI-Verified Operating Disclosure: The MBG Measurement Chain Meets the Rupiah Confidence Perimeter — five-node AI verification stack
  21. Weekly Rupiah Monitor: October 20, 2026 — three-chain convergence reading