Fraud detection for microfinance
0 likes767 views
The document addresses fraud detection within microfinance institutions in the Philippines, highlighting the prevalence of unbanked adults and the significant financial losses due to various common fraud schemes. It outlines a methodology for data analysis and insights derived from datasets, emphasizing the need for improved data collection, validation, and monitoring processes to mitigate fraud risks. Recommendations for enhancing fraud prevention measures are provided, including establishing better data management practices and stringent processes in application sourcing, loan disbursement, and payment collections.
Fraud detection for microfinance
- 1. FRAUD DETECTION FOR MICROFINANCE Team Women5Dev
- 2. 71 percent of Filipino adults are unbanked The unbanked represent 51.2 million adults in 2019 47.7 percent of Filipino households are low-income
- 3. MICROFINANCE INSTITUTION provides small business owners who don’t have access to traditional financing from major institutions access to capital
- 4. Provide access to financial services Aim to elevate the lives of its clients Push for women empowerment MICROFINANCE INSTITUTION
- 5. Someone always ruins it for everybody
- 6. A deception deliberately practiced in order to secure unfair or unlawful gain or causing loss to another party FRAUD
- 8. How can we prevent fraud? Business Problem
- 9. Methodology
- 12. 690.39 MB 306.93 MB 1.27 GB 10.03 GB 6.71 GB 40 KB 88.43 MB 8.11 KB 2.9 MB 309 Assets Gathering relevant datasets with the help of our sponsor who has the domain lens
- 13. Discovering relationships across different datasets and learning about their systems.
- 14. New tables ready for Analysis ... Table1 Table2 Table5 Table6 Creating main tables for analysis Cleaning and Transforming Data
- 15. Exploratory Data Analysis and Data Visualization Using the newly created tables/analytical base tables, insights were derived using the following tools
- 16. Insights
- 17. MILLION PHP in losses due to fraud 9.8
- 18. 68.18% of the total amount have yet to be recovered
- 20. Majority of fraud cases fall under Undeclared Collection, Loan Ride, and Falsification of Documents
- 21. Most staff involved in fraud includes Loan Officers and Branch Managers
- 22. Walk-in payments were remitted during weekends when centers and branches are closed
- 23. Walk-in Payments are remitted in some months only
- 24. Unusual number of members from different regions use the same contact numbers
- 30. Fraud is relatively more correlated with internal factors such as collection mode and loan officer involved LOANOFFICER LOAN OFFICER LOANID CONTACT NUMBER LOAN PRINCIPAL TOTAL INST NO NO OF INST COLLMODE PAYBALANCE FRAUD LOANOFFICER LOANID CONTACTNUMBER LOANPRINCIPAL TOTALINSTNO NOOFINST COLLMODE PAYBALANCE FRAUD
- 32. Recommendations
- 33. Improve Data Collection Data Quality Paper Database Analytics Real insights can only come from good data.
- 34. Recommendations Integrate near real-time dashboard with the new system ● Sustain efforts to utilize monitoring tools for decision making
- 35. Recommendations Application Sourcing Credit Evaluation Process Loan Disbursement Process Establish good data capture, validation and preparation per process
- 36. Recommendations Payments Collection Savings Withdrawal Establish good data capture, validation and preparation per process
- 37. Recommendations ● Scan copies of physical application form for cross referencing with any data validation ● Ensure consistent data types upon entry through text, numeric or list fields and restricting character inputs (e.g. punctuation marks, spaces) Application Sourcing
- 38. Recommendations ● Identify duplicate or related entries such as names, contact number and addresses of clients to monitor family relations or relatives in the program ● Record sources or channels of application to relate to other clients, chairwoman or loan officers Application Sourcing
- 39. Recommendations ● Call contact numbers provided by client to verify loan application ● Conduct field validation and adding geotags for client’s business pictures ● Flag incomplete requirements uploaded in the system to check validity of application Credit Evaluation Process
- 40. Monitor employee transactions and collections ● Create a watchlist of Loan Officers or Branch Manager based on volume of transactions Recommendations ● Apply stringent standard for loan disbursement (e.g. signatory, document needed such as ID’s) Loan Disbursement Process
- 41. Recommendations Build data dictionary and data schema ● Have common reference on the datasets across the organization. This may be relevant for fraud analysis to allow easier inter-department collaboration for fraud investigation
- 43. CHALLENGES SOLUTIONS Domain expertise Handling large datasets Questions and Clarifications Disaggregation and constraints Data issues and definitions Consultations with sponsor Timeframe Agile method
- 44. Jessa Gavila Jessa.Gavila@ftwfoundation.org Roch Derilo Rochelle.Derilo@ftwfoundation.org Gizelle Nacor Gizelle.Nacor@ftwfoundation.org Patcha Pangatungan Patricia.Pangatungan@ftwfoundation.org Women5Dev
- 45. FRAUD DETECTION ANALYSIS Team Women5Dev