Improving the Quality of Public Credit Registry Data Wook Jae Lee (Korea Credit Bureau) Sung Jun Chun (Korea Credit Bureau)
C o n t e n t s I. Introduction II. III. Debtor Identification Improvement Data Process Improvement IV. Data Quality Management Improvement V. Conclusions and Recommendations VI. Implications and Impacts 2
Ⅰ. Introduction-Project overview Indonesia KSP regarding Credit Infrastructure Year 2010 Way to Establish Credit Infrastructure for Enhancement of Consumer Credit Market in Indonesia Policy consultation to enhance Indonesian credit reporting structure by adopting 2-Tier system of Public Registry and Private Credit Bureau Year 2011 Enhancing Credit Market Infrastructure with Emphasis on Public Credit Registry System Technical consultation to enhance current PCR system (Bank Indoneisa SID) 3
Ⅰ. Introduction-Project overview Indonesia KSP regarding Credit Infrastructure Year 2012 Improving the Quality of Public Credit Registry Data Technical consultation on debtor verification and data quality Background Currently, there is no unique identification number in Indonesia. The Indonesian government is currently conducting a NIK project to assign unique ID number. Current debtor identifying logic cannot exactly identify a debtor (still needs human confirmation). These conditions causes several difficulties in sharing and distributing credit data in the technical aspect( low matching rate) and undermine quality of data. 4
Ⅰ. Introduction-Project overview Indonesia KSP regarding Credit Infrastructure Year 2012 Improving the Quality of Public Credit Registry Data Technical consultation on debtor verification and data quality Objectives Define a new logic to automatically identify a debtor and define the method to measure the power of the logic. Define a short term strategy to enhance the quality of SID data. Define a long term strategy to enhance the quality of BIK data. 5
Ⅰ. Introduction-Current Situations DIN Process (AS-IS) Every financial institution should get a DIN(Debtor Identification Number), an ID of a consumer, assigned by BI before reporting when a new customer is acquired. When there is an existing ID, it is used, and when there is not, a new DIN is created. Major Issues Low DIN matching rate AS-IS (Current Situations and Issues) Automatic matching rate is low (For the remaining, financial institutions select one from multiple DINs, or new DIN is created.) TO-BE (Objectives) Improve DIN automatic matching rate DIN duplication (One person has multiple DINs) Out of the total number of 50 million DINs currently held, it is considered that about 10 million are duplicated. (It is the result of reckless creation of new DINs.) Control newly created DINs Complicated DIN process When exact matching is not possible, request is made to financial institutions for verification. Financial institutions don t respond to them faithfully. Ultimately eliminated DIN verification process of financial institutions - Possible when the new ID system of the nation is completed Major considerations: quality of reported data, code system of address or zip code, performance of BI s IT system, work processes with financial institutions, etc. 6
Ⅱ. Debtor Identification Improvement Summary of Debtor Identification Improvement Plans Category Plans Remarks Data Improvement Integration of duplicated DINs (unification) Standardization of values for major fields related to DIN - IDCARDNO, TAXPAYERIDNO, DATEOFBIRTH, PLACEOFBITRH Utilize NIK and date of birth Representative DIN Garbage Data converted into NULL Matching Logic Improvement Statistical Methods DIN Process Improvement Use text matching technologies (using similarity matching functions) - DEBTORFULLNAME, MATHERSMAIDENNAME, DEBTORSADDRESS Take advantage of additional identification attributes (gender, address, place of birth, etc.) Select an optimal algorithm (apply decision tree) Improve DIN confirmation process - Use three ways only which include automatic mapping, confirmation of one candidate DIN, new DIN creation Offer guidance on enhancing DIN process 7
Ⅱ. Debtor Identification Improvement Remedy 1. DIN Confirm Process - Eliminate a financial company s confirm process for multiple DINs: Only one DIN with the highest score to be used 2. How to operate - To improve the process, server side application needs to be changed - As the role of NIK Format is expanding, it is expected that automatic mapping will increase, while the number of DIN confirmations made by financial companies will decrease. DIN Confirm Process Example First Matching (NIK Format) When it exceeds a certain score (for example 70): Automatic mapping for the highest scored DIN When it is below a certain score (for example 70): For the one DIN which has the highest score, verification is requested to the financial company When it exceeds a certain score (for example 80): Automatic mapping for the highest scored DIN Second Matching (Non NIK Format) Score range (for example from 30 to 80): For the one DIN which has the highest score, verification is requested to the financial company in question (If the company doesn t select the DIN, a new DIN is created When it is below a certain score (for example 30): A new DIN is created No Matching DIN A new DIN is created 8
Ⅱ. Debtor Identification Improvement Improvement Results New matching method to maximize automatic matching - will reduce bank's DIN confirmation by 66% - will reduce DIN duplication (unnecessary New DINs) by 25% Automatic Matching DIN Confirmation New DIN Creation 41% 61.6% 25% DOWN 37.5% 46% 400% UP 66% DOWN 13% 0.9% OLD LOGIC NEW LOGIC OLD LOGIC NEW LOGIC OLD LOGIC NEW LOGIC
Ⅲ. Data Process Improvement DIN Field Validation Enhancement Category Details Remarks Timing of validation Contents of validation Measures DIN Master CREATE / UPDATE DIN History CREATE DEBTOR Data CREATE / UPDATE DATEOFBIRTH/SEX : If it is NIK Format ID, date of birth and gender in IDCARDNO should have the same value as the ones for DIN. Check on mandatory fields : it is checked whether the value has been entered and it is a garbage value - DATEOFBIRTH, SEX, DEBTORFULLNAME, MATHERSMAIDENNAME, IDCARDNO, ZIPCODE, REGION, DEBTORSADDRESS, PLACEOFBIRTH - PLACEOFBIRTH to be encouraged to be reported in the city level TAXPAYERNO : When a value is entered, it is confirmed whether it is garbage When there is an error with value, reporting is on hold, and error feedback is sent to the reporter Induce revised correct value to be reported again At the time of reporting, validation is done on a real-time basis Encourage DIN value quality to be improved continuously To manage NIK properly, the field of whether it is NIK should be added and make members report the relevant data.
Ⅲ. Data Process Improvement Better Validation of DIN Set up of Error Feedback System Category Details Remarks What is regarded as error Everything which is considered error after DIN validation - Example : Data Type Error, Domain Error, NOT NULL Error, error between columns (IDCARDNO DOB and gender check) Error Feedback Process REQUEST ID_DATA = DIN010 MEMBER BANK INDONESIA ERROR FEEDBACK ID_DATA = DIN099 ID_DATA = DIN099 = ID_DATA = DIN010 + Error check (error response code + error check bit)
Ⅲ. Data Process Improvement Change of How to Manage DIN MASTER UPDATE Category Details Remarks At the time of AS-IS : FIRST MEMBER TO-BE : ALL MEMBER DEBTOR DATA UPDATE Entity At the time of DIN MASTER UPDATE, DIN HISTORY CREATE UPDATE, DIN UPDATE, as well Every DIN should have the below fields. CREATE/UPDATE Field Management - DIN MASTER : Institution/date involved in first creation, institution/date involved in last update - DIN HISTORY : Institution/date involved in first creation of MASTER, Institution/date involved in HISTORY reporting DIN History Tracking Possible AS-IS Existing DIN Case (When DIN field value is changed) TO-BE Existing DIN Case (When DIN field value is changed) FIRST MEMBER YES UPDATE DIN MASTER ALL MEMBER UPDATE DIN MASTER CREATE DIN HISTORY CREATE DIN HISTORY NO CREATE DIN HISTORY
Ⅳ. Data Quality Management Improvement Establishment of Effective Data Quality Management System Category Description Remarks Overview Accumulation Processing Services Systematize data quality management (quality measurement analysis improvement) Overall index - Information Index - Error Index, etc. Main Points Systematize data quality management and monitoring (DQMS) Objective evaluation and improvement based on Data Quality Management Index (DQMI) Expected Benefits Member Companies: Use high quality data services - Increase in data analysis and service usage PCR : Strengthen competitiveness and trust by enhancing data quality - Reduced customer disputes and raised credibility
Ⅳ. Data Quality Management Improvement DQMS - Data Quality Management System Quality Object Identification Quality Rule Quality Measurement Quality Analysis Quality Improvement Quality Management - CTQ Definition - DQI Definition - DQI Evaluation Standard - DQI Score Mgt - DQI Weight Mgt - Quality Rule Management - Rule Mgt - Error Code Management - idqms Job Mgt - idqms BatchJob Processing Result - Batch Job List Processing Result Warning Monitoring Ruling Profiling Verifying Member Quality Improvement Member Service Communication Info Registration Result Data Quality Index Data Standard Data Structure Data Flow Impact Analysis Meta Management Word/Terms Code/Domain Analysis Terms Format Structure/Model Compliance - Data Flow Mgt - Application Flow Mgt Impact Analysis Process Management Dash Board Data Process Control
Ⅳ. Data Quality Management Improvement DQMS - Quality, Meta Integrated System Quality System Control Improvement Analysis Quality Object Identification Measurement Quality Rule Definition Provide data structure and analyze the impact of data change Provide standard job management for quality measurement and analysis Support quality analysis and activities for improvement Enterprise-Wide Meta Data Standardization Compliance Analysis Data Flow Management Meta System Word Terms (Table Column) Domain Analysis Terms Data Layout Information Common Code Physical DB DB Catalog Logical DB Model Mart (ERD) Application SQL Analysis Staging Credit EDW Impact Analysis
Ⅴ. Conclusions and Recommendations Measures for debtor identification improvement are necessary including existing data enhancement, matching logic improvement and application of a statistical method Existing data enhancement refers to treatment of duplicate DINs, and improvement of low quality data as a prerequisite for boosting the DIN matching rate. Matching logic improvement includes active use of meaningful fields, use of additional fields for DIN matching and introduction of a technical function through which text matching similarity is generated as a percentage. In terms of a statistical method, a matching logic will be generated and applied based on a statistical approach, using the case with the highest matching rate. Short-term strategies through process improvement are indispensable. This process improvement includes DIN confirmation process, better validation of DIN fields and strengthened validation of data loading. Current DIN confirmation process needs to be enhanced to prevent duplicate DINs from being issued. Better validation of DIN fields is necessary. Proper validation framework and system should be in place in the DIN field quality management system of the current SID. Strengthened validation of data loading is recommended.
Ⅴ. Conclusions and Recommendations Long-term strategies through establishment of the data management system are recommended. Since a credit reporting agency is doing business based on data, data is the most important key element in the system. Therefore, it is crucial to put the proper framework in place which encompasses data format, structure, quality, and value from the beginning. In the long-term strategies to improve data quality, measures for enhancement have been suggested by taking into account data quality management framework, data quality analysis methods, standardization, and data quality management system in a comprehensive manner, along with establishment of the proper data management system.
Ⅵ. Implications and Impacts Implications & Impacts Provide practical consultations about debtor identification and data quality Consultation results will be applied while Bank Indonesia conducts PCR system enhancement project in 2014. Contribute to the enhancement of credit reporting infrastructure in Indonesia Credit Infrastructure enhancement by adopting 2-Tier system of Public Registry and Private Credit Bureau,which is reflecting on KSP recommendations since 2010. ( enacted in Feb 2013) Support overseas expansion of Korean financial Institutions by sharing Korean credit reporting infrastructure It is expected that this program will support Korean financial institutions entry into the Indonesian market, which are experienced with Korean credit reporting infrastructure.