Data Integration Guide - Credit Card Applications

 

Credit Card Application Data for Analysis

Application data fields, as well as customer profile data fields associated with the application, are an important driver of DataVisor’s ability to identify credit card application fraud. This document serves as a general integration guideline for identifying application fraud as it pertains specifically to credit cards. 

Credit Card Application Data

Fields that are helpful for credit card application use cases include (but are not limited to):

Category	Data Field	Description	Required
Account-Level Information	Account Number	Unique account number	X
	CC Application ID	A unique ID that identifies the specific CC application	X
Credit Information	FICO Score	The FICO Score of the applicant, as pulled from a credit report	X
	Credit Report Pull Date	The date when the report was last pulled by the financial institution	X
	Credit Report Provider	The credit report agency from which the report was pulled (e.g. Equifax, Experian)
	Other Credit Report Data	Any other fields that are parsed from the credit report (as applicable)
	Total Credit Limit	Total amount of credit for the customer across all cards and lending institutions	X
	Soft Inquiry Count	Number of soft inquiries of the customer’s credit report in the last 12 months. Includes personal credit reviews or periodic checks
	Hard Inquiry Count	Number of hard inquiries of the customer’s credit report in the last 12 months. Includes reports being sought for new applications
	Bankruptcy Information	If any prior bankruptcies by the applicant, a total count and dates for each one, along with code representing type / chapter of bankruptcy
	Collections Information	if any prior payments have been turned over to collections, a total count and dates for each one. Generally available from credit reports
	Payment History	Applicant’s payment history	X
	Recently Opened Products	Count of recently opened credit cards (within last 12 months)
	Credit to Debt Ratio	Credit to Debt Ratio
Application-Level Information	Credit Card Channel	Initial Source from where the applicant applied for the CC	X
	Promotional Offer Type	If a promotional offer was available with opening the card, the type (e.g. cash-back, miles, etc.)
	Promotional Offer Details	If a promotional offer was available with opening the card, the drill-down details (e.g. how many miles, % cashback, etc.)
	Referral URL	If application was online, the URL from where they applied
	Application Time	Timestamp of the CC application	X
	Requested Credit Limit	If card application allows for customer to specify desired credit limit, provide this requested number in US dollars
	FSI Geo-Coordinates (in-person)	For in-person applications for a card, the address and branch ID of the location at which the card was sought	X
	Device Signals (digital)	For online applications for a card, the IP and Device ID corresponding to the device from which the application was made	X
	Cosigner / Joint Applicant?	Is there a joint applicant or co-signer on this credit-card?	X
	Cosigner / Joint Applicant Information	If applicable, all personal info about the joint applicant or co-signer (name, email, address, phone, SSN, etc)
	OFAC Sanction Status	Whether the applicant appears on any OFAC sanction lists
	Citizenship Status	Is the applicant a US Citizen or Permanent Resident? This can also optionally be provided in profile data
Label	Payment Delinquency	An indicator of whether there have been delinquent payments in the last X months (X is typically 3-12)	X
	Account Status	An indicator of if the account has been closed due to suspected fraud	X
	Account Closing Reason Code	If available, the reason code for why an account was closed
	Upgrade Eligibility	For good users, an indicator of when they are eligible for upgrades to higher-value cards or other promotions

An example of credit card application data:

CC App ID	Account ID	Channel	Transaction Time	Previous CC Applications	FICO Score	...
1203029281904	53277897	In Person	2022-04-16 12:35:06	0	675	...
9303283057123	40098005	Digital	2022-04-16 12:35:06	4	800	...

Customer Profile Data

Customer profile data is also important for fraud detection, when merging with transaction level data.DataVisor uses various types of at-rest and in-transit encryption methods to make sure the PII data shared in the customer profile data is handled in a secured way. If PII data sharing is not available, we recommend using SHA-256 hashing to hash the data before sharing the customer profile data with DataVisor.

Fields that are helpful for customer profile data include:

Category	Data Field	Description	Required
Customer Profile Data	Customer ID / SSN	Unique identifier of the customer who performs the transaction	X
	Registration time	The time the user first registered with the financial institution	X
	Registration Channel	The method through which the customer first registered for an account with the FSI
	Registration Address	For in-person registrations, the address of the location where the customer first registered for an account
	Registration IP	For online registrations, the IP from the device where the customer first registered for an account
	Product types	A list of products that the customer has previously subscribed to from the FSI	X
	Customer tenure	Tenure of the customer
	Email	Email address of the customer	X
	Phone	Phone number of the customer, or phone prefix	X
	Full Name	The full name of the customer	X
	Address	The address of the customer	X
	Country	The country of the customer	X
	City	The city of the customer	X
	State	The state of the customer	X
	Zip Code	The zip code of the customer	X
	Date of Birth	Year of birth of the customer
	Annual Income	Annual income of the customer
	Monthly Rent	Monthly rent paid by the customer
	Employment Tenure	How long the customer has been working for their current employer
	Employment Status	Customer’s current employment status
	Employer	Employer name of the customer
	Applicant History	Any information about previous CC applications from the customer. Includes the following:  Number of previous Credit Card Applications Number of Approved and Declined CC Applications Credit Limit History Number of Credit Limit Increase Inquiries Delinquent Payments (x = 30, 60, 90 days) Rewards Average Monthly Balance Carryover Credit to Debt Ratio Bank-specific credit history	X

An example of customer profile data:

Customer ID	Name	Customer Tenure	Product Type	Email	Registration Time
53277897	Alice Jane	14 months	Premier Credit Card	[e-mail]	2014-01-01 12:03:03
40098005	John Doe	60 months	Premier Credit Card	[e-mail]	2017-01-02 12:00:00

Data Structure Report

Before sending the full data over, please send over a data structure report to your Technical Account Manager. This process allows DataVisor to ensure a faster turnaround time and a higher quality of results during the fraud assessment. This data structure report should contain the following for each data set:

• Schema
• Preview of 10 rows
• Earliest date and latest date (If there is an associated timestamp)
• Number of rows

Data Anonymization

DataVisor’s algorithm is able to work with anonymized data fields if required. For sensitive PII information, such as name and address, anonymized information can be processed by our algorithm as long as the data structure remains.

Choosing a Connection Type: Batched or Real-time data transfer

For a batch data transfer, the client sends data in bulk to DataVisor. The frequency of batch data transfer is determined by client’s use case and business requirements. Batch transfers are faster to implement, as they don’t require an integration with DataVisor’s APIs. Account application data should be given as is at the end of the batch time period, and included in the batch. 

Real-time data connection

For a real-time data connection, the client integrates DataVisor’s APIs to send a real-time stream of data. After a ~2 week observation and tuning period, DataVisor returns real-time results, which can be used as an additional signal in the client’s fraud detection infrastructure. All account application data should be uploaded at the beginning of the fraud assessment. 

Formatting and transferring data

Formatting Data

DataVisor can accept data in the following formats:

1. JSON (preferred) 
2. Tab Delimited
3. CSV

For other formats, like TXT or report exports, please contact your Technical Account Manager.

Example JSON format: 

{"txn_id":1203029281904, "customer_id":53277897, "txn_type":"Online", "event_time":"1423160520015", "time_zone":"UTC+0200", "amt":88.4, "txn_recipient":"Amazon Inc"}

Data Labels

Data labels are metadata that indicate which accounts are confirmed to be fraudulent. Although not required for DataVisor’s unsupervised technology, labels can be helpful to gauge results. If you have labels available, your DataVisor Technical Account Manager can help you determine whether it would be worthwhile providing them for the fraud assessment.

Transferring Data

DataVisor’s fraud detection solution can be deployed on-premise or in a cloud environment. For on-premise deployment, please contact us for customized data transfer details.

For cloud deployment, DataVisor will create an Amazon S3 bucket exclusively for you to ensure your data’s privacy. A customized python script will be provided to securely upload data to this bucket.