Regulated institutions need data that works, not more tools that don't talk to each other.
Belo Delta builds data science solutions for financial institutions navigating a fragmented, expensive compliance landscape: starting with Sonda, a modular compliance platform covering four regulatory obligations from a single data foundation, hosted entirely on Dutch and EU infrastructure.
Compliance tooling is broken for smaller regulated institutions
The Dutch market alone has 131 regulated financial institutions carrying AML/CFT monitoring obligations. Most of them are currently managing a patchwork of disconnected tools — each covering one obligation, none sharing data, most hosted on US-controlled infrastructure.
Blockchain analytics tools cover on-chain crypto activity but not euro payment flows. Fiat AML tools cover payments but have no crypto capability. DORA tooling addresses ICT resilience but has no connection to transaction monitoring. Every obligation has a separate tool, a separate vendor, and a separate data silo — with no ability to generate unified regulatory reporting from the combined dataset.
Six regulatory frameworks are converging simultaneously: the Wft, the Wwft (transitioning to the AMLR in July 2027), PSD2/PSR, MiCAR, DAC8, and DORA. Each draws on the same underlying transactional data. Institutions that can't afford to build in-house, and can't justify enterprise pricing, are currently left to accept the gaps, or manage with spreadsheets.
Three signals that make this urgent
The case for an EU-sovereign, integrated compliance platform is no longer theoretical. Developments in Dutch public policy, supervision, and the market itself confirm it.
Data sovereignty is now enforced Dutch policy
In May 2026 the Dutch government prohibited the acquisition of Solvinity, the cloud provider behind DigiD, by a US-listed firm, and in July 2026 the Rotterdam court refused to suspend that prohibition. The exposure arises from a US company's legal position, not from where a server stands: US law can compel disclosure of data wherever it is stored. Sonda is hosted exclusively on Dutch and EU infrastructure, with no US-owned company at any layer.
Supervisors are watching this market
The AFM has publicly named CASP compliance monitoring as a supervisory priority following MiCAR's full application, and every new authorisation in the register is an institution that must stand up transaction monitoring, often for the first time. From July 2027 the AMLR raises the bar further, replacing the reporting of unusual transactions with a stricter suspicious-transaction standard.
The market leader has confirmed the gap
The convergence of crypto and fiat monitoring is real enough that the largest blockchain analytics vendor paid a reported USD 150 million in 2025 to begin addressing it, from the fraud side, on US infrastructure, at enterprise pricing. The position Sonda occupies, broad coverage on EU-only infrastructure at an accessible price, remains unclaimed.
Three things we wouldn't compromise on
Every project starts from these — not as marketing language, but as the actual constraints we design within.
Hosted in the Netherlands
Your data stays on Dutch and EU infrastructure, end to end. No dependency on US-based cloud providers, and no ambiguity about which jurisdiction governs it.
Privacy first, by design
Built to minimise what's collected and retained, not maximise it. Security and data minimisation are design constraints from day one, not a checklist added before audit.
One foundation, many uses
The same underlying data work can power several things at once — monitoring, insight, efficiency — instead of building a separate, disconnected tool for each.
A tool, not a headline
AI shows up in our work where it genuinely earns its place — not because it's expected, and not as a label on something that didn't need it.
-
→
Used where it adds real lift
Pattern detection, anomaly scoring, and language-heavy tasks are places machine learning genuinely outperforms fixed rules — that's where it's applied.
-
→
Held to the same rigour as everything else
Models are validated, explainable, and checked against simpler baselines — the same discipline a research background demands, applied to production work.
-
→
Never the unaccountable part
Where decisions need to be defensible — to a regulator, a customer, or your own team — a model's output is a signal a person can inspect, not a black box you take on faith.
"Most regulated institutions are not short of data. They are short of a way to make that data do more than one job at once — without rebuilding from scratch every time a regulation changes."
Belo Delta Sonda
A modular compliance platform for regulated financial institutions — built on a single data foundation, covering multiple regulatory obligations simultaneously, hosted entirely on Dutch and EU infrastructure.
Sonda
Compliance platform for CASPs and payment institutions · EU-hosted · Modular · Subscription-based
AML/CFT Transaction Monitoring
EUR payment flow monitoring and crypto transaction screening from a single data layer — covering the fiat leg that blockchain analytics tools leave invisible. Detection logic is built around your institution's specific risk profile, not a generic rule library.
Fraud Detection and Prevention
Real-time anomaly detection for authorised push payment (APP) fraud, mule account patterns, and transaction-level risk signals — aligned with PSD2/PSR fraud liability obligations and MiCAR consumer protection requirements.
DAC8 / CARF Reporting
Automated generation of DAC8-compliant reports for the Belastingdienst, built directly from the transaction data Sonda already collects for AML/CFT purposes. No separate data collection layer required. First reporting deadline: 31 January 2027.
DORA Compliance Support
ICT third-party risk register, incident logging, and operational resilience documentation — structured around DORA's Register of Information requirements and major incident reporting obligations.
Subscription model — ongoing calibration
A static compliance tool is a compliance liability. Sonda subscriptions include scheduled reassessments of your detection rules and models, coordinated with your team and aligned to your institution's risk appetite and operational reality.
- 2 or 4 reassessment cycles per year, depending on subscription tier
- Rules and models reviewed against current typologies and regulatory updates
- Changes proposed, tested, and confirmed with the customer before implementation
- False-positive rates and detection performance tracked across cycles
Audit trail and quality assurance
Every rule or model change in Sonda is governed by a documented quality assurance process before it goes live — giving your compliance team the evidence trail it needs for internal governance and regulatory scrutiny.
- Full audit trail of all rule and model changes, timestamped and attributed
- Data quality tests run automatically before each change is implemented
- Performance benchmarks compared pre- and post-change
- Change documentation exportable for regulator or internal audit review
Company established
Registration with the KvK, technical architecture scoped, EU hosting providers identified, brand and website live.
Proof of concept
Module 1 for euro payment flows: configurable rule engine, ML-assisted anomaly scoring, alert and case interface, full audit trail.
Pilot and first revenue
First pilot customer, real-world testing, and the first Module 1 subscription with calibration sessions.
Commercial operations
Live product, structured calibration cycles, and Module 2 development in parallel with a growing customer base.
Priced for the institutions the market overlooks
Sonda is subscription-based and modular. Every subscription includes at least two documented quality assurance and calibration sessions per year, because a static compliance tool is a compliance liability. The figures below are total annual costs; there are no additional platform fees.
| Configuration | Total annual cost |
|---|---|
| Module 1 (AML/CFT monitoring), standard tier with 2 calibration sessions | €15,000 to 20,000 |
| Module 1 + Module 2 (fraud detection), standard tier | €20,000 to 25,000 |
| Module 1 + Module 3 (DAC8 reporting), standard tier | €17,500 to 22,500 |
| Full platform, Modules 1 to 4, enhanced tier with 4 calibration sessions | €37,500 to 42,500 |
Even the full platform sits well below the €50,000 to 150,000 annual cost of a comparable multi-vendor stack, and unlike that stack it covers all four obligations from one data foundation, with no integration burden between systems.
Applied data science consultancy is available separately at €150 per hour, or on a fixed-price basis for defined scopes. Quotes via info@belodelta.com.
Built for three audiences, in order
Sonda is developed for a deliberately sequenced market, starting where the need is most urgent and the fit is strongest.
Fintech startups building compliance from scratch
Institutions preparing for or holding a new MiCAR CASP or Wft payment institution licence, standing up their monitoring framework for the first time. No legacy tooling to displace, the highest urgency, and a data foundation that is DAC8-ready from day one.
Licensed CASPs and PSPs with unmet needs
Above all the gap no current tool closes: monitoring crypto-asset transactions and the euro payment flows that accompany them from a single data layer. Module 1 can run alongside an existing crypto tool rather than replacing it.
Non-financial Wwft-obliged entities
Accountants, notaries, estate agents and other Wwft-obliged organisations need lightweight, affordable compliance tooling. This audience is on the roadmap, not in the current build, and is recorded here as the direction after the core platform.
The same data, working harder
The data collected for compliance purposes is rarely put to work for anything else. Belo Delta's broader data science capability applies the same underlying dataset to problems beyond regulatory obligation.
This capability is available today as applied data science consultancy, contracted through Belo Delta, independent of the Sonda platform.
-
→
Customer insight
Understanding who your customers actually are and what drives their behaviour — turning scattered transactional records into something a product or commercial team can act on.
-
→
Growth signals
Spotting where the real opportunities are hiding in your own data, before you spend on finding out the hard way.
-
→
Process efficiency
Finding the slow, manual, or repetitive parts of how a company runs — and showing where data can quietly take the weight off.
Two backgrounds, one focus
Computational science background
A PhD from the University of Groningen and postdoctoral experience in applied data science, nine peer-reviewed publications, and seven years of quantitative modelling training, now directed at commercial data problems. Full-time on the build.
Regulatory and AML/CFT background
Nearly two decades across law enforcement and the private sector in financial crime and AML/CFT, including a current board role at a Dutch crypto-asset service provider, assessed and approved by the AFM. Part-time, defining what the product must detect.
We're building this in the open, early.
If you work in compliance or risk at a Dutch or EU financial institution, or if your company is sitting on data that could work harder — we'd like to hear from you.
info@belodelta.com