FREQUENTLY ASKED QUESTIONS

How does the GSTR-2B similarity match algorithm work?

[−]

The similarity match engine applies Levenshtein distance calculations within specific supplier GSTIN buckets. It evaluates the string edit distance between invoice numbers, allowing it to catch character typos, transposed digits, omitted serial prefixes/suffixes, and different year-marking variations. This resolves the majority of vendor billing mismatches without throwing false positives.

How does the CDNR Identity Inversion feature help?

[+]

In many ERP databases (like Tally, BUSY, Zoho, or SAP), a supplier registers an outward return as a Credit Note (CN), but the buyer enters it in their purchase register as a Debit Note (DN). This mismatch in document types causes standard matching systems to ignore the lines. Our engine implements an identity inversion step, automatically cross-checking Credit Notes against Debit Notes inside matching cycles to align these boundaries seamlessly.

Can the tool match books when invoice numbers are completely different?

[+]

Yes. While standard matching tools rely solely on invoice number matching, our AI Confidence engine scores matches across multiple vectors: GSTIN, Taxable Value, Tax Amount, and Date. If a supplier and buyer enter different invoice references (e.g. internal voucher number `VCH-089` vs invoice number `INV-9981`), but the GSTIN matches, the tax amount matches exactly, and dates are close, the engine applies a **Sole-Pair Context Boost** (+20%) to pair them automatically, flagging it for user approval.

What is the 3-Pass matching process?

[+]

The matching pipeline processes records in three sequential stages to optimize speed and accuracy:

1. Pass 1: Exact Match – Pairs invoices where GSTIN, Invoice Number, Date, and Amount match verbatim.
2. Pass 2: Cleaned Match – Cleans and compares numeric cores of invoice numbers (ignoring FY tags like `/24-25`, spaces, symbols, and leading zeroes).
3. Pass 3: Fuzzy / AI Match – Applies Levenshtein distance calculations and AI scoring rules to resolve remaining discrepancies.