Preparing curated data sets for an effective Data Analysis

Before finding the next fraud pattern/modus operandi affecting your organization, we need to make sure you have the right complete data for analysis. Having complete organized transaction data to detect suspicious activities, ensuring compliance with regulatory standards, and improving fraud prevention efforts.

Scope of Data

The first step in analysing data for your transaction monitoring system is to define the scope, which involves determining the types of transactions and data points relevant to the analysis. The scope should align with regulatory requirements and the financial institution’s risk management framework.

1. Time Period: Define the period under analysis, such as daily, weekly, or monthly transaction data. The chosen time frame should reflect typical patterns of behaviour but also account for potential anomalies.
2. Geography: Consider whether the transactions are domestic, international, or from specific high-risk regions. Transactions from countries with higher AML risks might require enhanced scrutiny.
3. Transaction Type: Include all relevant transaction types, such as wire transfers, credit card payments, cash deposits, withdrawals, or cryptocurrency transactions. Each type may carry different risks.
4. Customers: Determine if the analysis should focus on specific customer segments, such as politically exposed persons (PEPs), high-net-worth individuals, or customers in high-risk industries.
5. Other analysis relevant data: Are you looking for tap transactions, swipes, or bank transfers?

Key Data Columns

Some of the columns required to perform an analysis would be as the ones mentioned below.

1. Transaction ID: A unique identifier for each transaction.
2. Account Number: Identifies the account where the transaction originated or was deposited.
3. Customer ID: Links the transaction to a specific customer for further analysis of their transaction history.
4. Transaction Date and Time: The exact date and time the transaction occurred, important for time-based analysis.
5. Transaction Amount: The value of the transaction, which can indicate suspicious patterns (e.g., frequent high-value transactions).
6. Currency: The currency used in the transaction, especially important for international transactions.
7. Transaction Type: Details whether the transaction is a deposit, withdrawal, wire transfer, etc.
8. Origin and Destination: Locations or banks involved in the transaction, crucial for detecting cross-border or high-risk geographic movements.
9. Status: Whether the transaction is complete, pending, or rejected.
10. Flags: Previous alerts or unusual activity associated with the customer or transaction.
11. Geographical locations: If transactions are from a certain geography, or merchants listed in a certain country.
12. Merchant Type: Are you looking for specific merchant categories such as crypto service businesses, show stores, bookstores, etc.?

Analysis of Dates

Understanding dates and times on a transaction helps to identify time-based patterns of suspicious activity. Some groupings of dates include:

1. Transaction Frequency: Calculate how frequently customers perform transactions. Unusual frequency may indicate illegal activity, such as money laundering.
2. Time Series Analysis: Analyze transaction volumes over specific periods (e.g., daily, weekly, or monthly). Sharp increases or decreases in transaction frequency could signal unusual behaviour.
3. Anomaly Detection: Spot outliers by comparing current transaction patterns to historical data. For example, if a customer usually transacts once a month but suddenly makes 10 transactions in a day, this could be flagged as suspicious.
4. Holiday and Weekends Analysis: Certain criminal activities may spike during holidays or weekends when there is less oversight. Flagging these periods for closer review can help catch anomalies.
5. Unusual Times: Is a customer transacting during a certain time frame in a day over a long period and you get a completely different time for a high value, high risk transaction.

Stratified or sectional Sampling

When you cannot analyse all transactions for an analysis, stratified sampling involves dividing the dataset into distinct groups (strata) based on certain characteristics, such as customer risk level, transaction type, or geographical region. Then, a random sample is taken from each section for analysis, which will ensure that the sample represents different types of transactions, making the analysis more reliable and comprehensive.

1. Define strata or section: Group the data based on key characteristics such as:
   Risk levels of customers or geographic regions.
   High, medium, and low transaction amounts.
   Domestic vs. international transactions.
2. Select Sample Size: Choose a proportionate sample size from each section to reflect the overall dataset. This avoids bias in the sample and ensures that all segments are analysed fairly.
3. Apply Sampling: Use statistical methods or tools (e.g., Python, Excel) to apply stratified sampling. This allows for efficient analysis of large datasets without compromising accuracy.
4. Validate: After sampling, validate the results by comparing them with the overall dataset to ensure consistency.