Credibur Launches Verification Agent for Debt Facilities
< All News & Insights

Embedded Lending: Why the Credit Process of the Future Does Not Start at the Bank

Published on Oct 29, 2026
5 min read
Natalie Schleiz, VP Commercial at Credibur
Teaser Image

Platforms often know their merchants better than any bank does. The three routes to an embedded lending offer in Europe, and where banks fit in.

A restaurant selling through Wolt can receive a pre-approved financing offer directly in its merchant dashboard, based on its sales. Its house bank only sees the incoming payment.

That is embedded lending: the loan originates where the business happens. In Europe this is still the exception. Around three quarters of corporate credit still comes from banks, compared with roughly one fifth in the US. McKinsey expects embedded channels in Europe to carry 20 to 25 percent of lending by 2030, up from 5 to 10 percent today.

The reason is information. A marketplace knows a merchant's order volume, return rate and seasonality on a continuous basis, months before the annual accounts are available. Berg et al. (2020) show for a German online retailer that the digital footprint predicts payment defaults about as well as a credit bureau score, and the two combined do markedly better.

The bank does not see this data because it is not generated in the bank's process. So the credit process increasingly starts on the platform. Where banks stand in this model depends on the route the platform takes.

Refer, buy in or build

The simplest route is referral. The platform passes its merchants to a lender, earns a commission and in Germany usually needs no more than a license under Section 34c of the Trade Regulation Act (GewO). The bank remains the lender in the classic sense. But a share of merchants drops off at the handover, and the platform learns little about what gets signed.

The second route is buying in from a partner who brings the license, the capital and the servicing, while the platform keeps the customer relationship. This is how the Wolt offer mentioned above works: technology and funding come from finmid. Amazon takes a similar path in Germany, where the offer appears directly in the seller portal and the loans today come from HypoVereinsbank and Banxware. In both cases the partner sets the credit criteria and keeps a large part of the margin, and how it handles a missed installment reflects back on the platform. The bank is either directly behind the offer or one level further back as the partner's refinancing source.

The third route, an own loan book under an own license, is rarely the first step. It usually follows a partner model that has built up origination data and relationships with capital providers over years. The platforms we work with then build what makes up their edge themselves and buy in the rest.

The first test: business model and data

Which route fits depends above all on the role of the financial product. Is credit a revenue line of its own, or a lever that increases retention and repeat purchases of the core product? Then it matters whether the platform controls the payment flow and can withhold repayments from payouts, and whether the board carries the credit risk or only the product team. For capital providers these are the first checkpoints as well. Two patterns almost always mean no: a need that occurs rarely and does not show up in the platform's own data, and a customer base too small for a portfolio to season. In those cases, refinancing usually cannot be found.

The underestimated part: refinancing and operations

Once a platform scales, it needs institutional capital, for example through forward flow, a warehouse facility or a securitization. A capital provider scrutinizes the operational risk as closely as the credit risk: whether the portfolio is what it is claimed to be, and whether the book can continue to be serviced if the originator fails. This leads to requirements that, in my experience, originators almost always underestimate: loan-level data from day one, a running calculation of borrowing base and covenants instead of periodic reports, a named backup servicer and a documented, explainable credit model.

For banks, this is exactly the point of entry. They reach lending volume here that stays closed to them through their own channels. If they see the data continuously and at loan level, they can manage this volume like their own book, with eligibility criteria, concentration limits and covenants. The credit process of the future does not start at the bank, but it does not work without the bank's discipline.

I will be discussing these questions on November 9 at the annual conference "Kreditprozesse der Zukunft" (Credit Processes of the Future) in Düsseldorf, on the panel "Kredit als Feature" (Credit as a Feature). If you are there, come and find me. If not, schedule a call with me here.

About the author

Natalie Schleiz is VP Commercial at Credibur GmbH in Berlin, an infrastructure platform for non-bank lending and structured credit facilities in Europe. She is responsible for commercial strategy and the go-to-market of the product. Before Credibur, she built and scaled sales teams that brought fintech products in anti-money laundering, fraud prevention and financial infrastructure into regulated markets.

This article was first published in German as a guest contribution on Handelsblatt Live on October 29, 2026.