How TRBank used Kita to turn ₱8.77B of past loans into evidence for future lending
A Philippine bank turned four years of paper credit folders, from handwritten notes to photographs of originals, into structured records its teams can search, compare and reuse.
“PHP 8.77 billion of approved loan value is now fully queryable for the first time: the bank’s own ground-truth record of what it lends, to whom, on what terms and against what collateral.”
Philip LynchChairman of the Board, TRBank- Company
- TRBank, Inc. (A Rural Bank)
- Segment
- Tech-forward bank serving MSMEs
- Backed by
- Asialink Finance Corporation and Stem Financial
- Network
- 9 full branches, 11 Branch Lite Units and a Head Office
- Product
- Capture
- Scope
- 5,100 loan records approved 2022 to 2025
- Turnaround
- Full archive processed in less than a week
Challenge: four years of lending history, locked in paper
TRBank, Inc. (A Rural Bank) is a tech-forward bank serving MSMEs across the Philippines, lending through 9 full branches, 11 Branch Lite Units and a Head Office. Between 2022 and 2025 it approved thousands of loans, and each one was recorded in its own way, with no common format binding the archive together.
Every approved loan has a credit folder, the bank’s source of truth for that loan: the application, underwriting analysis, collateral appraisal, bank-statement review, scorecards and bureau report, as typed forms, handwritten notes, scanned documents and photographs of originals. One folder often runs to several hundred pages. The archive runs to thousands of folders.
Answering even basic questions about past lending meant pulling folders and reading them page by page. The data was there. It could not be searched, compared or reused at scale.
Why Kita: benchmarked on TRBank’s own records, then chosen
TRBank chose Kita after a pilot on its own lending documents, its own 194-field schema and its own scanned historical records showed that high-accuracy extraction was achievable. That gave the team the confidence to move from testing to the full archive.
Two more things mattered to TRBank: Kita’s focus on the Philippine market, and its familiarity with banking documents, local IDs, peso-denominated records and how banks like TRBank actually operate.
Tagalog handwriting, Philippine government IDs, peso amounts and local bureau formats, all handled out of the box.
194 fields across 19 sections, named exactly as TRBank defines them. Seven scorecard template families. 26,568 exception categories.
Faded scans, photographs of originals and partial crops ran through a validation-oriented workflow that flags ambiguity for review instead of guessing.
How it works: every credit folder becomes a 194-field record
Each loan record came back as a structured 194-field output across 19 sections: loan application metadata, borrower identity, residential and employment history, financial position, source of income, collateral appraisals, bureau checks, bank-statement summaries and scorecard outputs.
The two richest parts of a credit folder get dedicated handling. Every scorecard page is matched to the right TRBank template and its full scoring grid captured. Handwritten margin notes and Remarks-box entries are captured and classified into TRBank’s exception categories.
Every output then went through a separate, independent automated validation framework, across the full corpus rather than a sample.
- Problem
- Historical loan records were difficult to search, compare and reuse at scale.
- Transformation
- Each record was converted into 194 structured fields across 19 sections.
- Validation
- Outputs were reviewed through a separate automated validation framework across the full corpus.
- User impact
- Teams retrieve and analyze historical lending information without starting from physical folders or manual page-by-page review.
- Business impact
- The archive becomes a reusable base for portfolio analytics, policy review, operational reporting and future credit decisions.
Results: four years of lending, structured in under a week
The full archive was processed in less than a week, with the core production run taking no more than three days. Once source files, schema and validation rules are ready, converting a historical archive happens on a timeline that changes what is operationally possible.
The deliverable is four years of TRBank’s lending archive in flat-file form, ready for direct use: field names preserved from TRBank’s own specification, scorecard grids, remark categories and bank-statement snapshots included. TRBank’s teams can now answer questions about past lending without starting from a physical folder.
Structured values captured across 5,100 loan records.
Classified into one of 26,568 exception categories, so past judgment calls can be compared.
Scorecard instances extracted with their full scoring grids.
Monthly snapshots across 604+ issuing institutions.
Total loan value now queryable. Median ticket ₱692,500.
Folders with collateral appraisal data attached and structured.
What it unlocks: the archive becomes an underwriting asset
Once a lending archive is structured and validated at the field level, past credit decisions become easy to retrieve, compare and analyze. Portfolio analytics can rest on actual historical records. Policy review can move from anecdote to evidence. The bank’s own lending experience becomes a reusable input for future credit decisions.
Most legacy lenders sit on an archive like this: bound credit folders, paper applications, photocopies and scans from before digital underwriting, carried as overhead for storage, retrieval and compliance. Yet the lending book is the moat: the relationships, the local credit signal, the years of what actually repaid and what did not.
TRBank showed that historical credit files are not just operational records. Structured correctly, they become a competitive advantage.
Curious what this could look like for your bank?
Speak with our team.
Or write to hello@kita.ai