Analyzing Daily Bank Transactions: Maximizing Insights
summary
Claims that daily bank transactions spot problems months before annual accounts, backed by a product called heartbit that parses raw statements, maps flows to balance-sheet/income statement proxies, yields rule-based indicators (salarytimeliness, related-party rounds, bank switches) and feeds ML for improved default prediction. Tested in-house on a portfolio (~€1.18M financed, ~150 clients) after a missed default that cost banks €1.5M and €440k to the lender, uses sources from manual XML to bank integration and open banking, and is in daily use and being deployed to Poland.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 0:43Daily transaction analysis in SME lending
- 2:30Default case study
- 4:35Why financial reporting fails SMEs
- 6:20Bank statement analysis solution
- 7:48Heartbit and its four layers
- 9:13Data collection challenges
- 10:39Transaction classification and bank access
- 12:08Mapping data to financial indicators
- 13:12Key financial and behavioral indicators
- 14:35Machine learning for defaults and patterns
- 15:17Heartbit in lending and fraud detection
- 16:20Heartbit distribution models
- 17:47Future data expansion plans
- 18:52Q&A: Bank changes and tracking
- 20:43Q&A: Integration with Serbian Ministry
- 22:09Q&A: Team skills needed
- 24:58Closing remarks
related talks
28mBeyond the Balance Sheet: Transactional Data as a Window to Company Health
13mTransforming Financial Services with Real-Time Analytics
21mData Symphony: Orchestrating Insights in Fintech
36mLeveraging real time data & actionable insights for the financial sector
41mAI For All: Enhancing Digital Financial Services using Data Science
37mApproaching AI in Banking
01h 19mOn financial modelling, credit scoring, and risk assessment
36mLLM-Driven Financial Ledgers: Real-Time Accounting Systems at Scale