Signals for review, not verdicts for rejection
TransactIQ reads an Indian bank statement the way a forensic accountant reads it — checking the file, the numbers, the dates, and the counterparty shape for patterns that should not occur in a genuine account. Every finding is evidence the credit officer can look at directly. The decision stays with a human.
Document-level signals
What the file itself tells you — before you read a single transaction.
PDF authenticity check
Every uploaded statement is inspected for the metadata fingerprint left behind by Indian banking software. Files that came straight out of an issuing bank's system are cleared. Files that have passed through a consumer PDF editor, an online converter, or a desktop merge tool are flagged for reviewer attention.
Creation-vs-modification date consistency
When a PDF reports a creation date earlier than its last modification date, something has been written to the file after it first left the bank. TransactIQ surfaces the discrepancy so the credit officer knows a closer read is warranted — without attempting to conclude what was changed.
Balance chain verification
For each row, TransactIQ re-derives the running balance from the prior row together with the deposit and withdrawal on that row, then compares the result against the balance the statement itself prints. Any row where the two disagree is highlighted. The outcome — agreement or disagreement — is what reaches the report; the method behind the check stays internal.
Unexplained opening-balance jumps
If the opening balance is materially larger than the observed inflows and outflows can account for, the report calls out a possible inflated or carried-over starting position. This is a common shape in manually prepared statements where an operator writes in a higher opening figure to support a credit decision.
Transaction-level signals
What the transactions, as a population, reveal about how the statement came to exist.
Transactions on impossible dates
NEFT, RTGS, and cheque clearing rails are closed on Sundays, second and fourth Saturdays, and RBI-notified national bank holidays. Transactions printed on those dates through those channels cannot actually have settled. TransactIQ flags each such row. UPI, IMPS, and cash are 24×7 rails and are excluded — they move legitimately on closed-bank days.
Round-number clustering
Statements with an unusual concentration of perfectly round amounts are flagged. The threshold self-adjusts when most of the round amounts are legitimate ATM withdrawals — a common source of round numbers that should not generate a false alarm.
Digit-pattern forensics
Transaction amounts in genuine statements distribute across leading digits in a well-studied way. Fabricated or manually generated numbers distribute differently. TransactIQ runs a digit-pattern test on every statement and reports when the distribution is inconsistent with natural financial data — a classic forensic-accounting technique that catches fabrication visual inspection tends to miss.
Sequence pattern analysis
Genuine transactions are independent events. Fabricated sequences, by contrast, often drift into rhythms — amounts that track a pattern or feed a running total. TransactIQ identifies statements whose amount sequences show structure that real spending does not produce.
Counterparty spread check
Real bank statements show a small number of dominant counterparties — employer, landlord, the main utility — and a long tail of one-off payees. Fabricated statements frequently spread counterparties too evenly, because the person assembling them reaches for variety. A spread that looks unnatural is reported.
Duplicate transaction detection
Exact duplicates — same date, description, and amount — are surfaced in a dedicated sheet. Real statements occasionally contain genuine duplicates (two identical UPI payments to the same merchant on the same day); a reviewer can read context and decide. Copy-paste fabrication tends to produce duplicates at a rate that stands out.
Signal, not verdict
None of these indicators claims to prove a statement is fraudulent. Each one surfaces a pattern that should not occur in a clean file, points the reviewer at the evidence, and stays out of the decision. A flagged PDF metadata trail may reflect a customer who rotated a page before uploading. A flagged balance jump may reflect a legitimate high-value inflow the reviewer can see elsewhere. A flagged counterparty spread may reflect a genuinely unusual but legitimate account.
TransactIQ gives the underwriter, auditor, or resolution professional the evidence they would have looked for by hand — and the time to actually look. The final call on every flagged statement stays with the person who is accountable for the decision.
See the fraud signal set on a real statement
Evaluating lenders, auditors, and resolution professionals can request a walkthrough on a representative sample file under NDA.
Request a walkthrough