Fraud & Financial Crime

Zero-ETL graph analytics for fraud and financial crime.

PuppyGraph brings Zero-ETL graph analytics to the transaction and identity data you already have. Generate relationship-aware features for existing fraud models and follow multi-hop paths in investigation workflows without loading a separate graph database.

transfers toWallet AWallet BWallet CRecipientShared destinationPuppyGraphExisting transaction & identity dataAccounts · Transfers · Wallets · Devicestransfers toWallet AWallet BWallet CRecipientShared destinationPuppyGraphExisting transaction & identity dataAccounts · Transfers · Wallets · Devices

01 / Graph analytics for fraud & FinCrime

Follow the connections behind financial crime.

Connect accounts, transactions, wallets, and devices to generate network features for existing fraud models and follow connected paths in investigations.

Examples use simplified graph models. Adapt labels and properties to your schema.

01

Payment and mule-account networks

Follow transfers beyond the first recipient to find other accounts sending funds through the same intermediaries, and explore common destinations and connected groups.

transferstransfersAccountSenderAccountIntermediaryAccountRecipientOther sendertransferstransfersAccountSenderAccountIntermediaryAccountRecipientOther sender
View Cypher example

Set $accountId to the account under review. TRANSFERRED_TO maps recorded transfers between accounts.

MATCH path = (start:Account)-[:TRANSFERRED_TO]->(via:Account)
             -[:TRANSFERRED_TO]->(recipient:Account)
MATCH (other:Account)-[:TRANSFERRED_TO]->(via)
WHERE start.id = $accountId AND other <> start
RETURN DISTINCT other.id AS otherSender,
       recipient.id AS recipient, path
LIMIT 25

Returns: Other senders using the same intermediary, with the starting account’s two-transfer route to a recipient. Shared routes alone do not establish fraud.

02

Fund tracing and crypto investigations

Explore wallet counterparties and transfer routes to trace connections to an investigation target, applying relationship properties to focus the traversal.

transfertransferWalletStarting pointWalletIntermediateWalletTargetCounterpartytransfertransferWalletStarting pointWalletIntermediateWalletTargetCounterparty
View Cypher example

Set the starting and target wallet IDs and $minAmountUsd. This example assumes amountUsd is a prepared, comparable value on every transfer.

MATCH path = (start:Wallet)-[:TRANSFERRED_TO*1..3]->(target:Wallet)
WHERE start.id = $walletId AND target.id = $targetId
  AND all(transfer IN relationships(path)
          WHERE transfer.amountUsd >= $minAmountUsd)
RETURN path
LIMIT 25

Returns: Up to 25 routes of one to three transfers to the investigation target, keeping only routes where every transfer meets the amount threshold.

03

Shared identities and coordinated accounts

Find accounts that share devices, phone numbers, or other identifiers with an account under review, using relationships defined by your prepared data and identity-matching logic.

AccountConnectedAccountConnectedAccountConnectedAccountConnectedShared deviceAccountConnectedAccountConnectedAccountConnectedAccountConnectedShared device
View Cypher example

Set $accountId to the account under review. USED_DEVICE is defined by your prepared identity and device records.

MATCH (start:Account)-[:USED_DEVICE]->(device:Device)
      <-[:USED_DEVICE]-(other:Account)
WHERE start.id = $accountId AND other <> start
RETURN DISTINCT other.id AS relatedAccount,
       device.id AS sharedDevice
LIMIT 25

Returns: Other accounts and the device IDs they share with the starting account. The result provides investigation context, not a fraud verdict.

02 / From financial data to graph analytics

Run graph analytics directly on transaction data.

The data behind financial crime analysis already lives in existing tables. PuppyGraph maps those tables into a graph model and queries them directly.

Traditional approach

Copy data. Keep two systems in sync.

ETL + duplicate storage + ongoing sync

  1. Curated transaction tables
    IcebergClickHouseDatabricksSnowflake
  2. Build and run a graph-loading pipeline
    ExportTransformLoad
    Export, transform, load, then synchronize changes
  3. Operate a separate graph databaseStore and maintain another copy of the data
    Keep the graph copy in sync
  4. Fraud & FinCrime analytics
  • More pipeline work. Propagate updates, backfills, and schema changes into the graph copy.
  • Another system to operate. Manage graph storage and the synchronization jobs alongside your data platform.

With PuppyGraph

Query existing data. Skip the graph copy.

Zero-ETL · Query data in place

  1. Curated transaction tables
    IcebergClickHouseDatabricksSnowflake
  2. PuppyGraph queries source tablesMap tables to vertices and edges in a graph schema
  3. Fraud & FinCrime analytics
  • No graph-loading pipeline. Query your existing transaction tables through the graph model.
  • No separate graph data store. Keep your source data in the platform you already operate.

Coinbase

“PuppyGraph is a very interesting graph query engine. It doesn’t require us to load or ETL any of the data into a specialized or proprietary database storage layer for graphs. PuppyGraph queries all our transactional and crypto data directly and performs all kinds of graph-based exploration and aggregation. It’s so powerful, and our users really enjoy this type of flexibility.”
Eric SunSr. Manager of Data Platform at CoinbaseAt Data+AI Summit 2024Read the Coinbase story →

03 / Customer-inspired analysis

Find wallets that share a recipient.

Adapted from a real customer query in crypto compliance, this example finds wallets that send funds to the same recipient, using illustrative wallets and results.

Illustrative data

transfers toWallet AStarting walletWallet BRelated walletWallet CRelated walletWallet RShared recipienttransfers toWallet AStarting walletWallet BRelated walletWallet CRelated walletWallet RShared recipient
A shared recipient is a connection to investigate, not evidence of wrongdoing.

From question to result

Who sends funds to the same recipient as A?

  1. Follow A’s outgoing transfer to recipient R.
  2. Find other wallets with incoming transfers to R.
  3. Return those wallets and the shared recipient, excluding A.
Illustrative query result
Related walletShared recipient
BR
CR

For investigation: review B and C alongside A, with their transfers to R as context.

For your model: use the number of other wallets sharing this recipient as a feature. In this example, the count is 2.

Build your financial crime graph

Bring graph context to fraud and financial crime workflows.

Download PuppyGraph to query your financial data, book a demo for your use case, or join our Slack community to ask questions.

$ docker run -p 8081:8081 \
  -p 8182:8182 -p 7687:7687 \
  puppygraph/puppygraph:latestPulling puppygraph/puppygraph:latestStarting PuppyGraphWeb console listening on 0.0.0.0:8081Gremlin endpoint listening on 0.0.0.0:8182PuppyGraph is running# no pipelines · no second database · no data movement