FinBox

Bank statement parsing in BankConnect, the device based score in DeviceConnect and the risk score feeding decisions all depend on models. A loan origination system, co lending rails and embedded distribution are worth having regardless.

Instant underwriting and real time decisioning are the pitch, with Sentinel letting each lender author its own policies at design time. Referral thresholds, review queues for marginal applications and confidence on a score are not described.

Accuracy, discrimination statistics and validation results are unpublished, and approval gains are reported as business results rather than model measurements. How the statement parser performs across bank formats, or the device score across handsets, is left unexplained.

At IIFL Finance, BankConnect cut costs by 30 percent and raised small business loan approvals by 17 percent, with 41 percent conversions. More than 100 enterprise clients are claimed beside that named result.

Many lenders compete for overlapping borrowers in one market, and alternative data scoring improves with pooled repayment outcomes, so the incentive to learn across clients is strong. Whether one lender's device signals or performance data inform another lender's scores, and whether a lender can decline, is not said.

Data collected through the infrastructure products is covered by the privacy policy, which lists the device signals gathered to assess creditworthiness and names a grievance officer. Retention periods, subprocessors and lender to lender sharing are unstated, and the policy carries no update date.

ISO 27001:2013, a SOC 2 Type II report covering security, availability and confidentiality, and a CERT In empanelled auditor's safe to host certificate are named. Dates, auditors and report availability are not given.

The Unified Lending Interface from the Reserve Bank of India's innovation hub, the Account Aggregator framework and the Open Network for Digital Commerce are supervised rails that require qualification to join. The rules governing what runs across them are not named directly.

Call logs, contact aggregates and installed application metadata feed device based underwriting, signals that proxy for income, religion and health. Thin file borrowers least able to contest an outcome are the target, and fair lending testing, outcome analysis and adverse action handling are unpublished.

Borrowers scored on bank statements and phone signals get a real time decision without knowing which system made it or which signals counted. Beyond a deletion request and a privacy grievance officer, no correction route, reason codes or statement of who bears a wrong score's loss is published.

Public lending rails make the data chain visible, and the device inputs behind a decision are listed. Bureaus, identity sources and any model provider for scoring or parsing are unnamed.

An interface first, SDK based platform claims more than 2,000 integrations and 500 systems integrators, with bureau data and national lending rails connected. Live deployments at IIFL Open Neobank and Muthoot FinCorp ONE show it working at institutional scale.

A verified hosting environment is mentioned without a provider, region or residency commitment. India's localization rules for payment and financial data make that location a live question for any lender.

BankConnect, DeviceConnect and the origination system carry no published price or plan, and the homepage links no pricing page. Per statement and per decision fees circulating elsewhere come from outside reviews.

Banks, non bank finance companies such as Muthoot FinCorp and digital lenders such as Ring are served across origination, underwriting and portfolio monitoring, plus co lending. Loan types run from personal and working capital lending to buy now pay later, with India the market evidenced.