Twelve GST reconciliation mistakes that show up in audit files

The errors that undermine an input credit reconciliation are rarely arithmetic. They are structural: comparing the wrong documents, pooling populations that must stay separate, searching too narrow a period, and collapsing categories that carry different conclusions. Most of them produce a reconciliation that appears to balance, which is why they survive review and surface later as a demand the client was not expecting.

Errors in what is being compared

1. 01Reconciling against GSTR-2A instead of GSTR-2B2A is dynamic and keeps changing as suppliers file and amend. A reconciliation against it gives a different answer each time it is run, so the working paper cannot be reperformed, which is the whole basis on which documentation is expected to stand. Use 2B, which is static once generated, and record the download date.
2. 02Pooling a multi-state client's registrationsEach registration has its own GSTR-2B and its own credit ledger. Pooling the registers and the statements produces a reconciliation that can balance overall while concealing credit taken in a state where it was not available. Reconcile each GSTIN separately and never net across them.
3. 03Comparing the register to the returns instead of to the booksThree things need to agree: the purchase register, the credit recorded in the books, and the credit claimed in GSTR-3B. Checking the register against the returns and skipping the ledger leaves the possibility that the accounts show something different from both.
4. 04Reconciling a register that was never agreed to the ledgerIf the register does not total to the purchases in the general ledger, the reconciliation is between a statement and a spreadsheet. Agree the population before matching anything.

Errors in the population

1. 01Leaving reverse-charge supplies in the register sideThey are not reported by the supplier in a way that puts them in 2B, so every one appears as an unmatched line. In a client with significant reverse-charge purchases this can be most of the difference report, and it buries the items that matter.
2. 02Leaving credit and debit notes in with invoicesThey reduce credit rather than create it. Matching them against invoices produces noise on both sides. Reconcile them as their own population.
3. 03Stopping at the March statementCredit relating to the year under audit routinely appears in the 2B of April, May or later, because the supplier filed late. Searching only within the year reports these as supplier non-filing, which is the single most consequential miscategorisation in this working paper: it turns a timing difference into a potential loss of credit.
4. 04Ignoring the effect of actions in the Invoice Management SystemRecords rejected by the client do not carry credit into 2B even though the supplier filed them, and records on which no action was taken flow through as deemed accepted. A reconciliation that treats 2B as purely a product of supplier behaviour will misread both.

Errors in the conclusion

1. 01Treating availability as eligibilityAn invoice appearing in 2B means the supplier reported it. It says nothing about whether credit on that supply is permitted. Credit blocked under section 17(5) appears in 2B like any other credit, and a reconciliation that matches it and moves on has confirmed exactly the wrong thing.
2. 02Netting differences across categoriesA timing difference in one direction and an ineligible credit in the other produce a small net figure and two unreported problems. Quantify each category separately and conclude on each separately.
3. 03Missing the conditions that 2B cannot showRule 37 requires reversal where the supplier has not been paid within 180 days of the invoice date, and that is invisible in any comparison of the register to 2B. It requires an ageing of unpaid purchase invoices, which is a different procedure entirely and is regularly omitted.
4. 04Accepting an unexplained residue as a balancing figureThe residue is the number the conclusion actually turns on. Describing it as immaterial without investigating it is an assertion, not a conclusion. Investigate individually and state what remains.

How each mistake shows up on review

A short self-check before the paper goes for review

• Each registration reconciled separately, with nothing netted across them.
• 2B statements identified by period and download date, and retained in the file.
• The six expected-difference categories removed and quantified, not silently deleted.
• The following months searched for late-filed invoices, and the search evidenced.
• Every unmatched line assigned to exactly one category.
• Eligibility tested against section 17(5) by reference to the nature of the supply.
• The 180-day condition tested from an ageing, not from the reconciliation.
• Credit at risk quantified by category, with the recoverable and non-recoverable portions separated.
• The result agreed to the credit claimed in the returns actually filed.
• A conclusion that states what was found, written by a named person.