Vendors show matching on tidy narrations, but Indian statements carry truncated UPI, NEFT, RTGS and IMPS narrations, missing UTRs and duplicate amounts that cause wrong matches.
Give each vendor the same eight real narration lines and record whether each is matched with a visible reference, matched within a stated tolerance, or left open with a reason.
Eight masked test lines: truncated narration, missing UTR, two equal amounts on one day, bank charge with GST, interest credit, failed UPI reversal, one payment across invoices, round-sum transfer.
A scored test sheet per vendor showing expected treatment, actual treatment and whether the audit trail recorded the reference and user.
A reconciliation tool earns trust on messy narrations, not tidy ones. Before the demo, pick eight real lines from last month’s statement, write down the correct treatment for each, and ask the vendor to process them live. Score each line on one question: is the result explained on screen, or silently forced?
The structure of each narration type is explained in the narration patterns guide. Teams staying manual can use the Excel narration parsing recipe.
What does each rail put in the narration?
- UPI: usually a 12-digit reference (RRN), a VPA such as name@bank and sometimes a short remark typed by the payer. The general UPI limit is ₹1 lakh per transaction, with higher NPCI limits for some merchant categories, so large customer receipts usually arrive by NEFT or RTGS instead.
- NEFT: a UTR, commonly 16 characters, and a remitter name that is often cut short. NEFT has run round the clock since 16 December 2019.
- RTGS: a UTR, commonly 22 characters, and the remitter name. RTGS is for ₹2 lakh and above and has run round the clock since 14 December 2020.
- IMPS: a 12-digit reference and sometimes the payer’s mobile number. The per-transaction limit is ₹5 lakh.
Formats vary by bank, so treat these as patterns to test, not rules a tool may assume.
Which eight lines should you pick?
- A narration truncated mid-name, with no invoice number.
- An NEFT credit with the UTR missing from the narration.
- Two customers paying the same amount on the same day.
- A bank charge plus 18 percent GST as two debits or one combined debit.
- A quarterly interest credit.
- A failed UPI payment and its reversal a day or two later.
- One payment that settles three invoices.
- A round-sum transfer, such as ₹1,00,000, with no reference.
What is a fair answer for each line?
There are only three acceptable outcomes: matched with the reference used shown on screen; matched within a stated date or amount tolerance that you configured; or left in the exception queue with a reason. Anything matched without a visible basis is a fail.
| Test line (masked) | Expected treatment | Acceptable tolerance | What the audit trail should record |
|---|---|---|---|
| Truncated UPI narration | Suggested match for review | None on amount | Reference used, reviewer and time |
| NEFT without UTR | Open with reason, or match on amount and date with review | Date within 2 working days | Basis of match and approver |
| Equal amounts, two customers | No automatic match on amount alone | None | Both candidates shown |
| Bank charge with GST | Matched to charge and GST ledgers | None | Rule applied |
| Interest credit | Matched to interest ledger | None | Rule applied |
| Failed UPI and reversal | Paired as a net-zero item | Within 5 days | Both line IDs linked |
| One payment, three invoices | One-to-many match | Exact total | Each invoice linked |
| Round-sum transfer | Open with reason | None | Reason and owner |
The tolerances above are examples. Use your own policy and ask the vendor to show where it is set.
Why does reference beat name in a worked case?
A UPI credit of ₹18,750 arrives with a narration holding only a VPA and a 12-digit reference. Two invoices are open: ₹18,750 for Customer A and ₹18,500 for Customer B, whose name looks similar to the VPA. A name-led match might pick Customer B and leave a ₹250 short payment that does not exist. An exact-amount match supported by the UPI reference recorded on Customer A’s receipt picks the right invoice. If no receipt carries the reference, the tool should suggest Customer A and ask for approval. The unreconciled credit decision tree covers what to do when neither applies.
How Terra Insight fits
TransactIG delivers bank-to-books reconciliation that matches statement lines to your books, with unmatched lines kept in an exception queue and audit-ready exports. Bring your eight lines and we will run them in a representative-sample fit benchmark. Details are on the bank reconciliation software page.