Inter-chain analytics

(Section: Ecosystem and Network)

1) What is inter-chain analytics and why it is needed

• Unified accounting of value and activity: volumes, liquidity, commissions, retention.
• Observability of bridges and P2P connections: finalization, lags, reorg/challenge events.
• Traffic and conversion attribution: cheyn→cheyn, kanal→produkt.
• Risk and compliance: AML, sanctions, behavioral fraud, entity identification.
• Decision making: OKR/budgets, limits, update and liquidity regulations.

2) Data sources and events (canonical list)

1. Chains/registers: blocks, transactions, event logs, states of smart contracts.
2. Bridges: applications, receipts, evidence (light/optimistic/ZK), finalization statuses.
3. Payment providers/CCS: passing checks, limits, payment statuses.
4. Product events: onboarding, deposits/bets/conclusions, gaming and behavioral metrics.
5. P2P transport: Pub/Sub receipts, RPC success, latency.
6. Reference books: networks, assets, decimals, chainId, contract addresses, SDK versions.

3) Data architecture (streams and storages)

Ingest (streaming): connectors to nodes/indexes, webhooks bridges, CDC from operating databases.
Raw (Bronze/Raw): immutable batches labeled 'observed _ at' with source metadata.
Clearing/normalization (Silver): dedup, semantic enrichment, timezone alignment, asset mapping.
Kernel models (Gold/Core): unified facts' transfers', 'bridges', 'onchain _ events', 'kyc _ status', 'payouts'.
Marts: finance (GTV/TVL/Take Rate), product (retention/funnels), risk (scoring), operating system (SLO).
Cache/Serve: OLAP/HTAP for dashboards and APIs, and a separate search for/tx addresses.

Transport: Kafka/Pulsar (exactly-once semantics over idempotency), object storage for raw materials, parquet/column formats for analytics.

4) Finalization, reorgs and idempotence

Event states are 'observed' → 'confirmed (k)' → 'finalized' → 'invalidated (reorg)'.
K-confirmations rule - Configured by network/asset type.
Optimistic/Challenge windows: Support for "contested" status for bridges.
Idempotency: 'idempotency _ key = chainId' block 'tx' logIndex 'topic' (or payload hash).
Replay: scheduled backfill and index change recovery.

5) Entity resolution model

Address → Actor: addresses, keys, wallets ↔ account/organization/provider.
Cross-chain graph: address relationships by one owner (heuristics, signatures, onboarding data).
Confidence levels: hard-link (KYC, on-chain signature), soft-link (behavioral correlations).
Aliasing: Store stable identifiers (PIDs) instead of PII in analytics.

6) Unified event diagram (simplified)

yaml event:
id: string # global UUID observed_at: timestamp # when they saw chain_id: string # 'eth-mainnet', 'solana-mainnet',...
block_height: long tx_hash: string log_index: int event_type: string    # transfer    bridge. lock    bridge. mint    kyc. pass    payout. done...
status: string      # observed    confirmed    finalized    invalid actor_src: string # address/peer-id/source organization actor_dst: string # address/peer-id/destination organization asset: string # canonical symbol (e. g., USDC), + decimals amount: decimal usd_value: decimal # rate normalization at the observed_at bridge_ref: string # link with the application/receipt of the metadata bridge: object # network/contract/version/gac/fee, etc.
idempotency_key: string

7) Normalization of assets and prices

Canonical asset directory: symbol, decimals, chain mapping, contract addresses.
FX normalization: historical rates and asset prices at the 'observed _ at' timestamp.
Multi-active bundles: Group "wrapped" and native assets.

8) Key Metrics and Showcases

8. 1 Finance and liquidity

GTV (Gross Transaction Volume) over networks/assets/bridges.
TVL and Net Flow over bridges and pools.
Take Rate/Volume Fee; Cost-to-Serve to transfer.
Payout SLA Hit Rate, Finality p50/p95, Pending Backlog.

8. 2 Product and user

Cross-chain MAU/DAU (dedup по PID),

Retention D1/D7/D30 taking into account multi-chain activity,

Funnel: input network → bridge → target product → action.
QoT (traffic quality): traffic validation after anti-fraud.

8. 3 Risk and compliance

Fraud/Dispute Rate, High-Risk Score%, Sanctions Hit%.
Anomaly rate by translation patterns, velocity-check, clustering.
KYB/KYC Pass% and timings.

8. 4 Operating system and SLO

Bridge Success-Rate, p95 Finality, Relay Availability,

Reorg/Challenge events, Error budget burn.

9) SQL/Pseudo Query Examples

GTV by circuit pairs

sql
SELECT src. chain_id AS src_chain,
dst. chain_id AS dst_chain,
date_trunc('day', e. observed_at) AS d,
SUM(e. usd_value) AS gtv_usd
FROM events e
JOIN bridges b ON e. bridge_ref = b. id
JOIN networks src ON b. src_chain_id = src. id
JOIN networks dst ON b. dst_chain_id = dst. id
WHERE e. status = 'finalized' AND e. event_type IN ('bridge. lock','bridge. mint','transfer')
GROUP BY 1,2,3;

Cross-chain retention D7

sql
WITH first_touch AS (
SELECT pid, MIN(observed_at) AS t0
FROM product_events
WHERE event IN ('signup','first_deposit')
GROUP BY pid
),
week_activity AS (
SELECT DISTINCT pid
FROM product_events pe
JOIN first_touch ft USING(pid)
WHERE pe. observed_at BETWEEN ft.t0 + INTERVAL '1 day'
AND ft.t0 + INTERVAL '7 day'
)
SELECT 100. 0 COUNT() / (SELECT COUNT() FROM first_touch) AS d7_retention_pct
FROM week_activity;

SLO Bridge Showcase

sql
SELECT date_trunc('hour', observed_at) AS h,
100. 0 SUM(CASE WHEN status='finalized' THEN 1 END)/COUNT() AS success_rate,
percentile_cont(0. 95) WITHIN GROUP (ORDER BY (finalized_at - observed_at)) AS p95_finality_min,
SUM(CASE WHEN challenge_event THEN 1 END) AS challenges
FROM bridge_events
WHERE observed_at >= now() - INTERVAL '7 days'
GROUP BY 1;

10) Attribution and multi-channel path

last-touch/position-based model with weights for network source, bridge and product.
UTM→On -chain: associate clicks/referrals with the onchain address when onboarding (with consent).
Associative models: Shapley/Markov for complex set→most→produkt paths.

11) Anti-fraud and behavioral signals

Graph features: common counterparties, circular transfers, quick turnover.
Velocity limits and anomalies: bursts, "splitting" of amounts, night clusters.
Bridge fraud schemes: resubmission, KYC bypass attempts, sandwich patterns with liquidity.
Models: gradient boosting/graph-embeddings; teach on incident marking.

12) Privacy and compliance (privacy-by-design)

PII minimization: PID instead of direct identifiers, tokenization.

Data residency: partitioning by region, encryption "at rest/on the road."

Right to delete: tombstone/redaction events with provability.
Access and auditing: role ACLs, read logs, signed reports for checks.

13) SLI/SLO for Analytical Pipelines

• Freshness (median lag from 'observed _ at' to appearance in Gold),
• Completeness (% of events without holes as expected by K-confirmations),
• Correctness (% of events validated by schemes/rules),
• Reorg handling success
• Serve latency (p95 requests to storefronts/dashboards).

• Freshness p95 ≤ 3 min (streaming), ≤ 15 min (batch).
• Completeness ≥ 99. 7%, Correctness ≥ 99. 9%.
• Reorg handling success ≥ 99. 9%.
• Serve p95 ≤ 500 ms (main showcases).

14) Observability and lineage

Data Lineage: from dashboard to raw event (column-level).
Quality signals: completeness, uniqueness, referential integrity, schema drift.
Alerts: "quiet failures" (no new data), distribution jumps, growth of'unknown' fields.

15) Dashboards (templates)

• Success-Rate, p95 Finality, Relay Availability, Challenge/Reorg, backlog, error budget burn.

• TVL, Net Flow per chain, cost-per-transfer, utilization, insurance fund.

• MAU/DAU (dedup), cross-chain retention, channel funnels, QoT.

• Fraud/Dispute Rate, sanctions hits, high-risk share, speed of proceedings.

16) Operating regulations and playbook

Incident: Freshness Lag> SLO

Check connectors/indexers, switch to backup, enable degradation mode (showcases show "last finalized"), eskalate to the source owner.

Incident: reorg/challenge surge

Enlarge K-confirmations/dispute window, enable "delayed finalization" for large amounts, notify bridge/operators.

Incident Currency/Asset Divergence

Freeze affected pairs, roll back the directory, recalculate USD normalization, publish a report.

Incident: Fraud/Dispute jump

Tighten the limits/scoring, turn on the manual high-risk review, finish training the model on a fresh pattern.

17) Configuration Example (Pseudo-YAML)

Finalization windows by networks

yaml finality:
eth-mainnet: 12  # блоков polygon: 256 solana: "optimistic: 32 slots"
optimistic-bridge: { challenge_minutes: 20 }
zk-bridge: { proof_time_sla: 180 }

Idempotency and deduplication rules

yaml dedup:
key_template: "${chain_id}    ${block_height}    ${tx_hash}    ${log_index}    ${event_type}"
ttl_hours: 48

Pipeline SLO

yaml pipelines:
ingest_stream:
freshness_p95_min: 3 completeness_min_pct: 99. 7 gold_build:
correctness_min_pct: 99. 9 reorg_success_min_pct: 99. 9

18) Implementation checklist

1. Capture sources, schemes, finalization windows and owners.
2. Enable idempotency and reorg-handling (states + replay).
3. Build a kernel of models (transfers/bridges/onchain_events/kyc/payouts).
4. Set up asset directories and FX normalization.
5. Define the pipeline SLI/SLO and dashboards.
6. Implement entity resolution and privacy-by-design.
7. Include anti-fraud scoring and incident regulations.
8. Run backfill and tests on historical reorg/challenge cases.
9. Regularly review schemas, metric weights, and sources.

19) Glossary

Finality - irreversibility of the state/event.
Reorg - reassembly of the chain, leading to the cancellation of part of the blocks.
Challenge period - a challenge window in optimistic models.
Entity resolution - mapping of addresses/accounts of a single entity.
GTV/TVL - transaction volume/blocked value.
Completeness/Freshness/Correctness - basic data quality metrics.

Bottom line: inter-chain analytics is not just a summary of metrics, but a manageable discipline: a single scheme of events, correct finalization, stable pipelines, privacy, anti-fraud and understandable showcases. By following this framework, the ecosystem gets a truly "end-to-end" view of value, risk and growth - from raw block to business solution.