Case study · Financial Services & Lending

Loan files verified before an analyst opens them

A consumer lender whose analysts spent the first twenty minutes of every file confirming that the documents agreed with each other.

The client

Who they are.

A consumer lender competing on decision speed. Its analysts are paid to judge credit risk, but the file that reaches them is assembled by the applicant, from documents of very uneven quality.

Anonymised at the client’s request. Client names are never disclosed without written permission.

85kloan files per year
9documents per file
20 minsaved per file
The task

What had to be solved.

The first twenty minutes of every file went to reconciliation: does the name on the ID match the payslip, does the income on the payslip match the tax notice, is the IBAN the same everywhere. Doctored payslips were caught by instinct, or not at all, and files bounced back to applicants two or three times.

The goal

What success looked like.

  • Give the analyst a file whose documents already agree
  • Detect an altered income document rather than trusting it
  • Ask the applicant once, for the exact page that is wrong
The solution

What Ceertia does here.

One use case per product, declaring the documents required and the consistency questions to answer across them. Identity, income and account details are reconciled before the file is queued, and every mismatch cites both source zones — so the applicant is told precisely what to correct.

A real rule One rule, as it runs.

The declared income must be consistent between the payslips and the tax notice; flag any gap beyond the policy tolerance.

The procedure

How it runs, step by step.

1
Intake

The file arrives with its documents, in any format, from whoever sends it.

Identity documentProof of addressPayslipsTax noticeBank statementsLoan application
2
Recognition & classification

Every page is identified and mapped against the expected document list. Missing documents, duplicates and out-of-scope pages are flagged on receipt.

3
Verification

Each rule asks the file a question and answers it directly on the documents, with page, zone and source text attached to the verdict.

4
Arbitration

The team approves, requests a fix, or rejects. No file is approved or rejected automatically.

The result

What changed.

Analyst time moved from reconciliation to credit judgement. The number of return trips to the applicant fell, because the request that goes out is specific instead of "your file is incomplete".

Before
  • Analysts reconciled names, incomes and IBANs by hand before any credit thinking started.
  • Doctored payslips were caught by instinct, or not at all.
  • Files bounced back to applicants two or three times before being complete.
After
  • Identity, income and account consistency are checked before the file reaches an analyst.
  • Every mismatch cites both source zones, so the applicant can be told exactly what to fix.
  • Analyst time goes to credit decisions, not document reconciliation.

Anonymised case. Client names are never disclosed without written permission. Figures describe the file volumes and checks in this scenario.

Financial Services & Lending All case studies

See it run on your own files.

Bring one real file to a 30-minute demo. Watch the verification run, evidence on screen.