KPIs

Purpose: These KPIs make the rail measurable and auditable end-to-end. They map directly to receipts and on-chain events, so finance, ops, and auditors see the same numbers.

Foundational law (never a KPI, always a constraint): No EMT, no funds · Must-fund before shipping · One-Claim · Locked EDSD → Unlocked EDSD only on proof · 50% of each protocol fee burns in EDM.

A. KPI taxonomy (one glance)

Domain

KPI

Target / SLO

Why it matters

L2 & DA

Block time p50/p95

≤ 2s / 4s

User latency floor

EMT→Release latency p50/p95

≤ 5s / 15s

Business latency

Batch-to-L1 median

2–10 min

Anchor speed

Forced inclusion resolution

≤ 30 min

Anti-censorship

DA “anchor-guarded” time

0 per month

Degrade predictably

PoV / Admissibility

Gate-fail mix

≤1% serialization, ≤2% identity, ≤1% sensors

Hygiene vs. substance

Duplicate block rate (One-Claim)

> 99.9%

Uniqueness by design

Revocation SLA (detect→freeze / pack / resume p95)

≤15m / ≤4h / ≤72h

Narrow pauses

Marketplaces (Trade)

Disputes / 100 gates

≤ 3

Friction proxy

False-block rate (buyer review)

≤ 1%

UX trust

Top-up SLA breaches

≤ 1% of gates

Must-fund discipline

Throughput capacity

≥ 300 orders/mo (mature region)

Scaling bar

Marketplaces (Tokens)

Burn coverage

100% of settles/retires

Economic integrity

Retirement throughput

≥ 200–300/day (Phase 3+)

ESG capacity

Finance / Economics

Burn coverage (global)

100%

Every event burns 50% fee

PoR daily freshness

100%

Reserve trust

Escrow reconciliation (Trade)

100% (start − Σfees = end)

Accounting truth

Advance loss rate (if enabled)

0

DeFi safety

Compliance / Interop

Mirror SLA (freeze→replace p95)

≤ 48h

External consistency

Audit replay success

100%

Independent verification

Redaction compliance

100% redactable paths salted

Privacy provable

Governance

Param changes within bounds

100%

Safety rails intact

Timelock visibility

100% ≥72h

Predictable change

Vote participation

≥ 15% veEDM avg

Legitimacy

B. L2 & DA (sequencer & anchoring)

  • Definitions:

    • block_time: ts(block_n) − ts(block_n−1)

    • emt_to_release: ts(release) − ts(emt)

    • batch_lag: now − ts(last_L1_blob_with_event)

  • Targets:

    • Block time p50/p95: ≤ 2s/4s

    • EMT→Release p50/p95: ≤ 5s/15s

    • Batch-to-L1 median: 2–10 min; anchor-guarded mode at ≥60 min lag

    • Forced-inclusion resolution: ≤ 30 min

    • Availability: ≥ 99.9% / month

  • Alerts (yellow/red):

    • Block time p95: > 5s (Y) / > 8s (R)

    • Batch lag: > 15 min (Y) / > 60 min (R, enter anchor-guarded)

C. PoV / Admissibility

  • Gate-fail mix (last 30d, by lane):

    • Serialization: (E_CANONICAL_DRIFT, E_FORMAT_INVALID, E_SET_NOT_SORTED, E_FILE_MISSING) ≤ 1%

    • Identity: (E_SEAL_MISMATCH, E_LOT_LIST_MISMATCH, E_HASH_MISMATCH) ≤ 2%

    • Sensors: (E_TEMP_OUT_OF_RANGE, E_MERKLE_PROOF_INVALID) ≤ 1%

  • Duplicate block rate (One-Claim): duplicates_blocked / (duplicates_blocked + duplicates_finalized) → > 99.9%

  • Revocation SLA (p95): Detect→freeze ≤ 15 min; Dispute Pack ≤ 4 h; Resume (corrective EMT/mirror) ≤ 72 h.

D. Marketplaces

  • Trade (milestones):

    • Disputes / 100 gates: ≤ 3

    • False-block rate (buyer review windows used but dossier later passes unchanged): ≤ 1%

    • Top-up SLA breaches: ≤ 1% of eligible gates (must-fund)

    • Throughput capacity: ≥ 300 orders/month (per mature region)

  • Tokens (Energy/Carbon):

    • Burn coverage: 100% of settle/retire events include a FeeBurned with burn tx hash

    • Retirement throughput: ≥ 200–300/day sustained at Phase-3 scale

E. Finance & Economics

  • Burn coverage (global): #events_with_burn / #billable_events = 100%

  • PoR freshness: daily snapshot posted 100% days (hash on-chain; ladder visible)

  • Escrow reconciliation (Trade): per order, escrow_start − Σ(stage_fees) = escrow_end 100%

  • Advance loss rate (if Supplier Advance is enabled): 0; auto-sweep success ≈ 100% at release

F. Compliance & Interop

  • Mirror SLA (freeze→replace p95): ≤ 48 h

  • Audit replay success: 100% (auditor replays from L1 blobs + proof packs)

  • Redaction compliance: 100% of schema-declared redactable paths contain salted commitments

G. Governance

  • Param changes within bounds: 100% (ParameterStore refuses out-of-range)

  • Timelock visibility: 100% (≥ 72 h queue; Explorer diffs)

  • Vote participation: ≥ 15% veEDM average per quarter

  • Zero brake violations: (no proposals against No EMT/One-Claim/must-fund/50% burn/fee constants make it to execution)

H. How to compute (recipes on events)

  • Event fields (topics/args abbreviated; full ABIs in “Contract headers”):

    • EMTMinted(order, stage, subLot, claimId, povHash)

    • ReleasePosted(order, stage, amountEdsd, fee, burnHash)

    • TokenSettled(tokenId, buyer, fee, burnHash)

    • TokenRetired(tokenId, beneficiary, fee, burnHash)

    • FeeBurned(ref, fee, burnedEdm, burnTx)

    • GatePass(schemaId, claimId, povHash)

    • ClaimFinalized(claimId)

    • PoRSnapshot(snapshotHash, tBill, cash, asOf)

    • MirrorCreated/MirrorReplaced

    • ParamChanged, ProposalQueued/Executed

  • Examples:

    • EMT→Release latency: join EMTMinted to ReleasePosted on (order,stage,subLot); compute ts(ReleasePosted) − ts(EMTMinted); aggregate p50/p95

    • Burn coverage: count FeeBurned where ref ∈ { ReleasePosted, TokenSettled, TokenRetired } / count of those refs

    • Duplicate block rate: attempted duplicates = submitProof reverts with E_ONECLAIM_TAKEN (from indexer logs); finalized duplicates should be 0

    • PoR freshness: days with PoRSnapshot(asOf in day) / days in period

I. Alerting bands (green / yellow / red)

KPI

Green

Yellow

Red

Block time p95

≤ 4s

4–8s

> 8s

EMT→Release p95

≤ 15s

15–25s

> 25s

Batch lag

< 15 min

15–60 min

> 60 min (enter anchor-guarded)

Burn coverage

100%

99–100%

< 99% (blocker)

PoR freshness

100%

95–100%

< 95%

Disputes/100 gates

≤ 3

3–5

> 5

False-block rate

≤ 1%

1–2%

> 2%

Duplicate block rate

> 99.9%

99.5–99.9%

< 99.5%

Mirror SLA p95

≤ 48 h

48–72 h

> 72 h

Red triggers incident + post-mortem; Yellow requires owner action within 48 h.

J. Ownership (RACI)

KPI group

Owner

Consult

Inform

L2 & DA

SRE / Protocol

Security

Apps, Governance

PoV / One-Claim / Revocation

PoV & Contracts

SRE

Apps, Governance

Marketplaces

Apps (Trade/Tokens)

PoV, Finance

Partners

Finance & PoR

Finance Ops

Security

All

Compliance & Interop

Partner Ops

Legal

Apps, Auditors

Governance

DAO/Governor

Council

Community

K. Phase gates (quarterly acceptance checks)

  • Phase 2 exit: EMT→Release p95 ≤ 15s; 100% burn coverage; daily PoR 100%; disputes ≤ 3/100 gates; duplicate block > 99.9%

  • Phase 3 exit: ingest success ≥ 99.8%; data freshness p95 ≤ 5 min; mirror SLA ≤ 48 h; retirement throughput ≥ 200–300/day

  • Phase 4 exit: 2+ regions stable; credit loss 0; governance proposals (≥ 20) executed within bounds; no brake violations

  • Phase 5 mid: 2–3 new regions live; zero bridge incidents; DAO participation ≥ 15% veEDM; ≥ 25 grants funded

L. Dashboards & exports

  • Explorer (public): per-lane latencies, burns, PoR, mirror SLA, attestor SLAs, governance diffs

  • Status page (SRE): block time, batch lag, inbox depth, forced inclusion events

  • Finance CSVs: per-order escrow reconciliation, fee/burn lines, PoR ladder

  • ESG packs: proof-of-origin, mirrors, retirements, burn hashes, Scope-2 mappings

Drawing

Plain recap

These KPIs keep EDMA operationally honest and auditable: seconds-level business latency, minutes-level anchoring, 100% burn coverage, One-Claim uniqueness, narrow revocations, daily PoR, and bounded governance. If a number turns yellow/red, we don’t guess—we follow the playbook. The settlement rhythm never changes: facts → PASS → EMT → Locked→Unlocked EDSD → fee → 50% EDM burn → receipt—or nothing moves.

Last updated