Pave

Lenders supply the bank transactions, credit reports and repayment history, and the product is entirely the transformation of that material into more than 4,000 attributes and outcome trained scores. Remove the models and the lender holds only its own raw data, since the company adds none of its own.

The company names the misconception that it replaces lenders' models and corrects it, stating that its products lift those models and lenders keep full control. The interface returns attributes and scores with no decision output, so there is nothing to over delegate and the credit decision stays with the regulated lender.

Scores are trained on observed loan performance and built separately per credit product and per small business industry, since a seasonal restaurant and a trucking firm have different cashflow shapes. No discrimination statistic, backtest or stability monitoring across thousands of attributes is published, so the lender's own validation has nothing from the vendor to start with.

The platform runs 100 million credit risk evaluations a month, and reports approvals up about 80 percent alongside defaults down roughly 45 percent, the pairing that matters since either alone can come from moving a cut off. Partners are named, but no single lender is, so the base is evidenced by throughput rather than identity.

Scores trained on repayment outcomes for each credit product need performance data from many lenders, so a pooled training corpus is implied by design. Whose outcomes train them, whether contributing is a condition of use and whether a lender can keep its data out of competitors' scores are never stated.

Data passed for analysis is described as anonymized, and results are delivered through a cloud data platform's secure sharing, so they stay encrypted inside the customer's environment under its own access policies. No retention schedule, subprocessor list or deletion commitment accompanies that architecture.

Replacing bulk file transfer with platform native sharing, explained as a way to keep data encrypted inside customer boundaries, is a control described rather than asserted. No attestation, certification or trust center offers the independent assessment a lender's examiners will want.

Furnishing creditworthiness information for credit decisions is activity consumer reporting law defines, and the repeated line that the company neither provides nor aggregates data reads like the boundary keeping it outside. It is never articulated as a position, and reporting agency status, disputes and adverse action duties go unaddressed.

Cashflow underwriting can extend credit to people a bureau describes poorly, and the same transactions reveal where someone shops, worships and sends money, carrying proxy exposure a bureau score lacks. An attribute family named Willingness judges character from spending rather than capacity, and no fairness testing or per population performance is published.

Supplying inputs rather than decisions keeps the regulated lender accountable, and attributes grouped into affordability, stability, willingness and assets at least contain the material for a decline reason. Nothing tells an applicant that cashflow analysis shaped the outcome or offers a route to correct a misread transaction.

Sitting on the customer's own loan outcomes, bank transactions and credit reports, the company has no bureau or data broker in its path to disclose. The infrastructure and decisioning partners are named, but the scoring and categorization models are not described and no subprocessor list appears.

Customers pass existing data to the interface untransformed, and results land as tables inside their own warehouse through native sharing, ready for training and dashboards. A decisioning platform is a named small business partner, though no origination, core banking or servicing system is.

Output rests in the customer's own environment under existing access controls rather than in a vendor held store. Where processing itself happens, and any region choice or residency commitment, is unpublished.

Pricing, packaging and basis of charge are all unpublished, and access starts with a demo request. For a product measured in monthly evaluations, whether a lead scoring call costs the same as an underwriting call would shape how a lender designs its funnel.

Consumer and small business lenders are served across lead scoring, underwriting, limit setting, servicing and collections. Buyers are lenders specifically, and no footprint outside one market is evidenced.