Banxware

Underwriting judgment rests on models, scoring cash flow from bank account data and platform sales feeds rather than a bureau file. The orchestration layer has value to a bank however the decision is reached, and part of the modeling sits on the roadmap funded by the 2025 round.

Capital reaching a merchant within 24 hours on little paperwork means the credit decision is automated, inside rules the lender configures at design time. At decision time there is no published referral threshold, manual review path or treatment for merchants with thin platform data.

Precise risk assessment is asserted with no accuracy figure, discrimination statistic or portfolio result. How models are validated and monitored as merchant behavior shifts is undescribed.

More than 50 platforms are live and six are named, from a major payments acquirer to a large food delivery group, and a white labeled lending product for a large German bank runs in production. A major European banking group invested in two rounds and funds the loan book, though no origination total or portfolio performance figure is published.

Platforms that compete with one another all feed merchant data into underwriting, so repayment outcomes seen on one marketplace may sharpen scoring on a rival's. Whether models train per platform or in common, and whether a platform can withhold performance data, is never said.

Bank transaction history combined with sales and order data from a merchant's trading platform gives a fuller picture of a small business than it has ever handed a lender. What the merchant is told about that sharing, and how long the data persists after a decline or repayment, is unpublished.

Attestation, certification and trust center are all absent from the public record. A new platform or lender evaluating the company starts its security review from scratch.

Under the forward flow arrangement the bank partner takes the full loan book, so the regulated lender holds the credit risk and licensing obligations, a reasoned position for a technology supplier. No supervisor, statute or lending regime is named in any market served.

Scoring a merchant on one platform's sales data measures creditworthiness inside that ecosystem, so a seller hurt by a ranking or fee change is downgraded as a borrower for reasons outside its control. Fair lending testing, decline explanations and performance across merchant categories are unaddressed.

Financing without a personal guarantee keeps a failing borrower's loss commercial rather than personal, a protection rarely stated so plainly. A merchant declined on its own sales data has no described way to learn which signals counted or to correct them.

The banking group behind the loan book, its German bank and two banking as a service platforms are all named, so an institution can see who sits on each side of a transaction. No provider is named for the scoring models, and no subprocessor list or hosting arrangement appears.

The full lending chain from onboarding to collections runs through one orchestration layer, and a bank can take the whole stack or only the modules it lacks. Interfaces pull sales and transaction data straight from the merchant's trading environment, so a platform integrates once to offer credit.

Hosting provider, region and residency are undocumented, beyond the floor that German operation sets under European data protection law. A bank outsourcing credit decisioning must tell its supervisor where processing occurs, and nothing published answers that.

Loans run from a thousand euros to five million, but those are end product terms, not what a platform or lender pays for the infrastructure. Whether the charge is a share of origination, a per decision fee or a platform license is unstated.

Platforms embedding credit for merchants, banks reaching small businesses they cannot acquire, and brokers placing bank backed loans all use one infrastructure. Reach is Germany and the Netherlands, limited to small business lending among merchants on integrated platforms.