When someone moves country, their financial record does not move with them. A person who repaid a loan for six years in Warsaw, Madrid or Bogotá arrives in the United Kingdom and, to a lender here, has no past at all.
The consequence is not only unfair, it is commercially untidy. Some applicants really are higher risk. Others are simply harder to observe. A lender that cannot tell the two apart pays for both, through declines that would have performed and manual reviews the data could have settled.
CashKi is the layer between a verified identity and a credit decision. We take financial evidence created under one system, bind it to the right person, record where it came from and how it was checked, and translate it into terms an underwriter in another country can actually use.
An applicant whose foreign record cannot be read is a manual review, and often a decline. The same applicant arrives as structured attributes your existing process can consume. The first thing that changes is not who you approve, it is how many files reach a decision at all.
Every attribute carries its source, the method used to verify it, the date it was checked, and how confidently it was matched to the applicant. Enough to explain an adverse decision to a customer and to sit in your model-risk file without an asterisk.
The foreign record, the bank data, the employment evidence and the documents arrive under one provenance standard and one set of definitions. Yours, not ours: where a foreign concept has no defensible equivalent, we return it unmapped rather than quietly translate it into something that looks familiar and is not.
And no score. The weighting stays inside your model, where the regulatory responsibility already sits. We think a verified attribute you can interrogate is worth more to a credit team than a number from a vendor you have no reason to trust yet.
Whether this changes decisions or only the cost of reaching them is an empirical question, and not one we can answer for you. It is what a retrospective test against your own declined applications would establish.
Nothing moves without the person it belongs to asking for it. Consent is specific to the attributes requested, limited to the purpose given, time-bound, and can be withdrawn.
We also send as little as the decision requires. Where a lender needs to know that income exceeds a threshold, it receives that answer and not the underlying transactions. The default is the narrowest disclosure that still answers the question, because a financial history is not made safer by travelling in full.
Early, and honest about it. CashKi is pre-revenue. No lender has yet used our output in a decision, and we make no claim about predictive value until one has.
There is an open enquiry with the Financial Conduct Authority's Innovation Pathways regarding where this activity sits in the regulatory perimeter.
We are looking for one UK lender willing to test the idea retrospectively, against applications it has already declined.
Two people: an economist in London working on the regulatory and competition side of the question, and a technical co-founder in Kraków building the system. No outside investment.
The thinking behind CashKi is published openly and built on the FCA's own market study and consultation record rather than on proprietary data. These are self-published essays. They are not peer-reviewed.
Copies available on request, or write and I will send them.