No IRN on the photocopy? The math still has to close.

IRN stands for Invoice Reference Number. When a supplier generates an e-invoice through the Invoice Registration Portal (IRP), the system assigns an IRN. It is a hash of the supplier's GSTIN, invoice number, and financial year. The IRN validates the invoice against the GSTN database directly. If the IRN matches, the invoice is genuine. If it does not, something is wrong.

For clean digital PDFs, IRN is strong. Clear's capture path uses it. The GSTN validates it. The credit is provable. When you have an e-invoice PDF with a valid IRN, you can claim ITC with confidence.

But most invoices are not clean digital PDFs

A lot of factory AP is still paper: photocopies, phone photos, skewed scans. Whether a given supplier is under the e-invoice mandate depends on current CBIC turnover thresholds, not on a round number we invented. Even when the supplier is mandated, what lands on the clerk’s desk is often a printout. The IRN may be in a QR that a poor scan cannot read. Then you are back to pixels and field checks, not a live GSTN query.

This is the photocopy tail. It is the largest category of supplier invoices in Indian manufacturing. It is also the hardest to validate with traditional tools, because the usual validation path (IRN lookup) does not work on it.

What happens when IRN lookup fails

When IRN lookup is not possible, some capture tools fall back to a model confidence score. That score is not a GSTN confirmation and it is not arithmetic. Treating 92% confidence as “8% of the rupee amount is at risk” is also wrong: confidence is not a calibrated probability of rupee error. It is a model’s self-score. Do not budget ITC on it.

Proof is either an IRN that GSTN accepts, or fields that survive checksum, HSN, and closure. Everything else is a guess you may still choose to review.

The math gate works without IRN

EntryLedger does not need IRN to validate an invoice. It runs a different path: GSTIN checksum, HSN validation, and arithmetic closure. These checks do not depend on the IRN. They depend on the numbers themselves.

The GSTIN checksum catches typos and transpositions. The HSN check catches missing or wrong codes. The arithmetic check catches totals that do not add up. Together, they prove that the numbers on the invoice are internally consistent. If the math closes, the numbers are right. If it does not, the invoice goes to review.

The photocopy tail is where the real AP pain lives. It is also where the real cost hides. A wrong number that passes a confidence check but fails arithmetic will cost you at filing time. The math gate catches it at entry time, before it reaches your books.

The e-invoice mandate and what it means for AP

The government has expanded the e-invoice mandate to cover more businesses over time. Businesses above certain turnover thresholds must generate e-invoices. This means more of your suppliers will eventually generate e-invoices with IRN. But the mandate has exceptions and phase-ins. Small suppliers are not yet covered. And even when they are, the photocopy that arrives at your factory does not carry the IRN from the digital original.

The practical reality is that for the foreseeable future, a significant portion of supplier invoices will be photocopies without IRN. The math gate is the validation path for these. It does not replace IRN. It complements it. When IRN is available, use it. When it is not, the math still holds.

The two validation paths

There are two ways to validate an invoice: IRN lookup and arithmetic closure. They serve different inputs and catch different errors.

IRN lookup works on clean digital PDFs. The IRN is generated by the IRP and embedded in the e-invoice. The capture tool reads the IRN, sends it to the GSTN, and gets back a confirmation. If the confirmation matches, the invoice is genuine. If it does not, something is wrong. This is a strong validation path. It proves the invoice was filed by the supplier.

Arithmetic closure works on any input, including photocopies. It does not need IRN. It checks whether the numbers on the invoice are internally consistent. Do the lines add up to taxable? Does taxable times rate equal tax? Do the taxes add up to total? If the math closes, the numbers are right. If it does not, the invoice goes to review. This is a different kind of proof. It does not prove the invoice was filed by the supplier. It proves the numbers are correct.

The two paths are complementary. IRN proves provenance. Arithmetic proves accuracy. For e-invoice PDFs, use both. For photocopies, use arithmetic. Together they cover the full range of supplier invoices.

What this means for your AP workflow

If your AP department processes both e-invoice PDFs and photocopies, you need both validation paths. The e-invoice PDFs go through IRN lookup. The photocopies go through arithmetic closure. The two paths converge at the export step: only validated invoices reach Tally.

EntryLedger handles both paths. It detects whether the input is an e-invoice PDF with IRN or a photocopy without IRN. It routes to the appropriate validation path. The result is a single export file that contains only validated invoices, regardless of the input type.

This is the practical reality of AP in Indian manufacturing. You cannot assume all invoices are e-invoices. You cannot assume all invoices are photocopies. You need a system that handles both. The math gate handles the photocopy tail. IRN lookup handles the e-invoice head. Together they cover the full picture.

Who must generate e-invoices

The mandate has moved down the turnover ladder over several years. Check the current CBIC notification, not a blog from 2021. Large suppliers you buy steel from are probably in. The neighbourhood fabricator who still prints on an 80-column printer may not be. Your AP pile is a mix. Policy will keep expanding. Paper When the mail says e-invoice but the attachment is just a PDF, here is the two-minute IRN check I run before posting. will lag policy by years.

Even when the supplier is mandated, what lands on your desk may still be a printout of the e-invoice PDF, photocopied. The IRN might be in a QR code that a 150 dpi scan cannot read. Then you are back to pixels and math. Do not design the process as if every page is a live GSTN query. Design it as if half the pages are pictures of paper. That is the factory.

EntryLedger will read IRN when it is there. It will not pretend IRN is there when it is not. The gate still runs. That is the point of this page.

QR-first extraction is a cheap next step when we have real tenant paper. Until then, do not stall AP waiting for a perfect e-invoice world. Process the photocopy. Prove the math. File. Chase IRN coverage as a supplier-onboarding project, not as a blocker for today’s pile.