Data Department worked embedded with a European fintech lending marketplace for three to four months in 2023. The marketplace offers revenue-based financing to software-as-a-service (SaaS) companies. We analysed its financing approach and built two modules: one that generates repayment plans from the parameters of a financing option, and one that scores credit risk using data pulled from the borrower's own systems.
What is revenue-based financing?
Revenue-based financing is a way for a company to raise money without giving up ownership. In this marketplace's model, a SaaS company receives funding now and, in return, sells a share of its future recurring revenue. It does not sell shares in the company (equity).
In general terms, the appeal is simple. Software companies with steady subscription revenue can fund growth, such as hiring or marketing, without diluting their owners. The lender's question is equally simple: how likely is this company to keep earning the revenue that pays the financing back?
The decision
A lending marketplace makes two decisions for every company that applies:
- What would repayment look like for a given financing option?
- How risky is this borrower, and should the option be offered at all?
Both decisions need to be consistent and repeatable across many applicants. Doing them by hand in spreadsheets is slow and hard to audit.
What we built
| Module | What it does |
|---|---|
| Repayment plan generator | Takes the parameters of a financing option and generates the repayment plan |
| Credit-risk scoring | Estimates the probability of default, meaning the chance a borrower fails to repay, using logistic regression |
| Data inputs | Accounting, banking and commerce data pulled from the borrower's systems |
| Data connection | A financial-data API (application programming interface) that retrieves that data from those systems |
Why logistic regression. Logistic regression is a well-understood statistical model that turns a set of inputs into a probability. In lending, a model whose reasoning can be explained to credit teams and reviewers is usually worth more than a slightly more complex one that cannot.
Why borrower-system data. Accounting, banking and commerce data show how a company actually earns and spends. For a lender whose repayment depends on future revenue, that is more direct evidence than a self-reported application form.
Results
- Engagement: embedded with the client team for three to four months in 2023.
- Built: repayment plan generation and credit-risk scoring modules.
We do not publish performance figures for this project. The value was an explainable basis for two core lending decisions.
Lessons
These lessons come from how we approach credit and risk work in general.
Explainability matters in lending. Credit decisions affect real businesses and are often reviewed. A simple model that people can follow is easier to trust, challenge and improve.
The data connection is half the work. Pulling consistent data from many different borrower systems is often harder than the model itself. Plan for it early.
See our financial services and insurance page for related work, and AI (artificial intelligence) strategy and fractional Chief AI Officer for how we work alongside client teams.
Frequently asked questions
What did Data Department build for the lending marketplace?
Two modules: a repayment plan generator driven by financing-option parameters, and a credit-risk score (probability of default) built with logistic regression.
What data did the risk model use?
Accounting, banking and commerce data pulled from the borrower's systems through a financial-data API (application programming interface).
Are there performance figures for this project?
No. We report what was built, not accuracy or business figures.