Data reconciliation is software that checks that a bank's records match the records of the other banks and firms it deals with, and that the numbers in its reports are right. The other side of a trade is called the counterparty. Reconciliation means comparing two sets of records and finding every difference. The page on reconciliation explains the general idea.
Every trade and payment is recorded twice: once by the bank, and once by the counterparty. Usually the two records match. Sometimes they do not. One side may have typed a wrong amount, a wrong date or a wrong account number. A difference between the two records is called a break. A break left unchecked can mean money paid twice, a payment missed, or a wrong number in a report to the regulator.
Here is a typical case. A bank's treasury team makes about 500 bond and foreign exchange trades a day with other banks. After each trade, post-trade processing confirms it with the counterparty. The reconciliation software then compares the bank's record with the counterparty's confirmation, field by field. One trade shows ₹5 crore in the bank's records and ₹5.05 crore in the counterparty's. The software flags the break. An employee in the operations team investigates and corrects it before the trade settles.
The software also checks the bank's own systems against each other. For example, it checks that the total of all loans in the core banking system matches the figure in the bank's accounts. These checks matter most for regulatory reporting. If the data going into a report is wrong, the report is wrong.
How is it different from payment reconciliation in a shop or a company? A shop checks that the money it received matches its orders. A bank's data reconciliation checks millions of records across many systems and counterparties every day, and records each check as evidence for the regulator.