Building a defensible ITC evidence trail before you claim

The conditions for availing input tax credit are not complicated to recite. You need a tax invoice, receipt of goods or services, the credit reflected in GSTR-2B, tax actually paid to the government, and payment to the supplier within 180 days. The difficulty is never the rule. It is proving, two or three years later and under time pressure, what you knew and when you knew it.

Evidence has a shelf life

An evidence trail assembled at the time of the claim carries a quality that a reconstruction never does: it shows the state of the world when the decision was made. A screenshot of a vendor's registration status taken on the day you claimed credit demonstrates diligence. The same screenshot taken after receiving a notice demonstrates only that you can operate a browser.

This is the single most useful mental model for the whole exercise. You are not building a file to satisfy an auditor. You are building a record of a decision, at the moment it was taken, so that the decision can be explained rather than defended.

Four artefacts per period

Most of what gets asked for is covered by four things, kept per period rather than per invoice:

1. The 2B reconciliation output. Matched, unmatched-in-books, and unmatched-in-2B, with the treatment applied to each bucket. Keep the working file, not just the summary.
2. A vendor status snapshot. Registration status, effective dates, filing status for the relevant period, and the risk band you assigned — captured on the claim date.
3. The exception register. Every case where you claimed despite a signal, or withheld despite a match, with a one-line reason and the person who decided. This is short and it is the file that earns the most credibility.
4. Payment evidence for the 180-day condition. A simple ageing of unpaid supplier invoices flagged against the credit availed.

Nothing here is exotic. What separates teams that survive scrutiny comfortably from those that do not is that the first group produces these four things every month, and the second produces them once, in response to a notice, for a period eighteen months old.

Timestamps are the whole game

An undated snapshot proves nothing. Where possible, evidence should carry a system-generated timestamp rather than a manual one, and it should be stored somewhere that does not permit silent editing. A PDF in a shared drive with a modification date after the notice date is actively harmful — it invites the question of what changed.

If you generate vendor risk reports from a platform, keep the generated file rather than a pasted extract. If you take portal screenshots, capture the full window including the date. If you rely on emailed confirmations from suppliers, keep the email with headers rather than the forwarded text.

The exception register deserves its own discipline

The temptation is to record exceptions only when they turn out badly. Do the opposite: record every exception, including the ones that resolved cleanly. A register showing forty exceptions of which thirty-eight resolved demonstrates a functioning control. A register showing two entries, both problems, demonstrates that the control only operates when something goes wrong.

Each entry needs four fields and no more: vendor, period, signal observed, decision and reason. Anything longer will not be maintained.

Reverse-charge and the awkward cases

Two categories consistently produce trouble because they sit outside the normal flow. Reverse-charge liabilities are self-assessed and therefore never appear in 2B, so a 2B-driven control silently ignores them. Import IGST arrives through a different route entirely. Both need a separate line in the reconciliation rather than an assumption that the main process covers them.

Retention and retrieval

The statutory retention period is long, but the practical constraint is retrieval. Evidence you cannot find within a working day may as well not exist. A flat folder per financial year, with a subfolder per period and consistent file naming, beats any elaborate taxonomy. Test it once a year by asking someone who did not build it to produce the file for a randomly chosen period.

The compounding return

The reason to build this properly is not the notice you may receive. It is that the same artefacts answer supplier disputes, support internal audit, shorten statutory audit queries, and make vendor renegotiation conversations factual. The file pays for itself several times before anyone from the department ever asks for it.

Indicative GST compliance and ITC risk information based on GSTN-reported data. Not a government portal, credit rating, legal opinion, professional advice or guarantee of ITC availability.