Scaling PVS Integrations and Embedded Finance at Nelly Solutions

Nelly Solutions is a Berlin healthtech/fintech (Series B, €50M, Lakestar and Cathay Innovation) that digitises patient intake, e-signatures, and payments for medical and dental practices — over 1,200 practices, more than 2 million patients. The product replaces clipboards and fax machines by plugging directly into the practice management systems (PVS) clinics already run on. I owned that integration layer, and the embedded-finance products on top of it.

Germany's PVS landscape is fragmented: Charly, InSuite, and dozens of others, each with its own data model and its own update cadence. A new connector meant weeks of custom work. A broken one meant a practice's front desk losing trust in the product the same morning. Onboarding stalled on integration capacity, and international expansion was fantasy until that changed.

The integration programme I led moved the work from our engineers to a system. Versioned APIs with stable contracts, semantic versioning, and a published deprecation policy gave PVS vendors something they'd never had from us: predictability. A sandbox with conformance test suites let vendors certify their own integrations instead of queueing for our help. SLOs per connector — sync time, failure rates, on dashboards — turned reliability from a hope into a number someone owns. A tiered partner programme with release calendars kept the biggest vendors close.

The unglamorous part of stable contracts is that they bind you too: no more quick per-vendor hacks, and every breaking change costs a deprecation cycle. That discipline is exactly what made 30+ integrations operable by a team that stopped growing linearly with connector count.

The other number worth explaining is 0.7% monthly MRR churn across those 1,200+ practices — for healthcare SaaS, low enough to invite questions about the methodology. It came from watching earlier. A practice heading for the exit shows it in the data weeks in advance: logins thin out, support tickets stack up, features that used to be daily go quiet. We built detection around those signals and intervened while accounts were still saveable, instead of learning about problems from a cancellation notice.

The redesigned intake flow ships dynamic forms in 30+ languages with conditional paths, QR-based remote pre-check with save-and-resume, and e-signatures feeding a consent ledger with revocation handling and audit trails. Staff tooling chases the unsigned documents so people don't have to.

On the finance side: payments via Adyen — POS, payment links, Apple Pay, Google Pay, SEPA — reconciling automatically back into the PVS. Factoring with intake-linked risk checks, one-click enrolment, 24-hour payouts, in recourse and non-recourse models. Patient installment plans up to 36 months, tied to treatment plan acceptance. Daily accounting journals and VAT summaries export straight to the practice's tax advisor.

Italy and Spain created the standard scale-up temptation: fork the codebase per country and move fast. I held the line against it. Reusable country packs — toggleable features, price books, notification templates, legal copy — carried Italy live (dental-first, local tax and invoicing rules, Italian IBANs) and left Spain prepared, on one architecture. Six functions had to ship together for that — Product, Engineering, Risk, Legal, Operations, banking partners — so releases ran on monthly trains with freeze windows for critical connectors, a passed DPIA per release, and incident playbooks covering rollback and partner communication.

• 1,200+ practices connected, with measurably higher connector reliability and faster integration cycles.
• Intake completion rose, unsigned consents dropped, and practices saw shorter time-to-cash through factoring.
• Italy live and Spain ready on the same platform — no forks.
• Support load fell as observability and partner self-serve replaced manual firefighting.