Shanghai Xinmin Information | Invoice verification and OCR recognition services

Enterprise overview

Shanghai Xinmin Information Technology Co., Ltd. is an enterprise providing information system construction and technical services, serving the digitalization needs of multiple corporate clients.

Business pain points

• Expenditures are concentrated in several categories: technical services, channel promotion, content production, office, and travel. There are many suppliers, short settlement cycles, and small per-transaction amounts with dense transaction volumes. Invoice face information currently relies on manual character-by-character entry. VAT special invoices, general invoices, itineraries, train tickets, and fixed-amount invoices each have their own data extraction positions, and when encountering differences in invoice types, the entry format must also be switched. Entry errors propagate along certification, deduction, and vouchers, and post-correction costs are high
• Duplicate reimbursement relies on memory and ledger comparison, and duplicates across periods and departments are hard to detect. Under spot-check methods, abnormal invoices are often exposed only at the certification stage, by which time the business has already completed payment
• Voided invoices and red-flushed invoices mixed into accounts directly affect input VAT deduction and cost recognition. After China Golden Tax Phase IV, tax comparison tends toward full quantification. Once invoice authenticity has problems, it directly affects deduction, and spot-check verification can no longer cover regulatory requirements

Solutions

• Introduce recognition capabilities covering all invoice types, automatically extract and structurally output invoice information, directly write it back to business systems, and eliminate manual entry one by one
• Automatically connect to tax authority verification to check invoice number, amount, invoicing date and issuer status, and perform cross-period duplicate reimbursement comparison
• Items with low recognition confidence are separately pushed for manual review, while the rest of the full process is unattended
• Ultimately upgrades verification coverage from sampling to full volume, intercepting anomalies before booking, with verification records retained for future reference
• And opening recognition and verification capabilities to other business systems for invocation, avoiding duplicate construction of similar capabilities across systems