Vendors claim data sovereignty. Procurement teams nod along. Then the proof of concept begins, and suddenly the definition gets blurry. What looked like full control on a slide deck turns out to mean “your data stays in our cloud region”, which is not quite the same thing. Testing data sovereignty claims during a PoC is one of the most underrated parts of platform evaluation, and one of the most consequential, especially for organisations operating under frameworks like the Digital Operational Resilience Act (DORA) or the NIS-2 Directive.
This post walks through how to structure a PoC so that sovereignty is tested, not assumed. We look at what to probe, what signals matter early, and how to turn your findings into something you can actually act on. At the end, you’ll see how the Stackable Data Platform (SDP) approaches these questions in practice.
What data sovereignty actually requires in a PoC
Data sovereignty is not a feature. It is a set of conditions: you control where your data lives, who can access it, how it moves, and what happens to it when you stop using a service. A PoC needs to test all four dimensions, not just the one the vendor highlights in their demo.
In practical terms, this means the PoC environment should mirror your production constraints as closely as possible. If your organisation requires data to remain on-premises or within a specific jurisdiction, the PoC should run there, not in a vendor-managed sandbox. If your compliance posture requires audit logs, role-based access, and encryption at rest, those controls need to be active during the test, not bolted on afterward. A PoC that runs in a vendor’s managed environment with default settings tells you very little about sovereignty in your actual infrastructure.
Key sovereignty signals to look for early
Some signals appear within the first few days of a PoC and are worth watching closely. The first is data egress behaviour: does the platform send telemetry, usage data, or metadata to external endpoints by default? Check network traffic early. Platforms that phone home without explicit configuration are a red flag for sovereignty-conscious organisations.
The second signal is configuration transparency. Can you read, version, and reproduce every configuration decision in plain text? Infrastructure-as-code approaches, where the entire platform state is expressed in declarative files, make this straightforward. Platforms that rely on opaque GUIs or proprietary configuration stores make it much harder to verify what is actually running and why.
A third early signal is identity and access management integration. Does the platform support your existing identity provider, or does it require you to use its own? Sovereignty over access control is as important as sovereignty over data storage.
How to structure PoC test scenarios for sovereignty
Structure your PoC around scenarios that stress-test the boundaries of control, not just the happy path. Three scenario types are particularly useful here.
Portability scenarios
Can you export your data, configuration, and pipeline definitions in open formats? Run a scenario where you attempt to migrate a workload out of the platform entirely. If that process requires vendor assistance, proprietary tooling, or significant rework, you have found a sovereignty constraint worth documenting.
Failure and recovery scenarios
What happens when the vendor’s support or managed service is unavailable? Test whether your team can operate, debug, and recover the platform independently. Sovereignty includes operational independence, not just data location.
Access control scenarios
Simulate a situation where you need to revoke access for a specific user, role, or external party. How quickly can you do it? Is the change auditable? Does it propagate consistently across all components? These scenarios reveal whether access governance is genuinely in your hands or partially delegated to the vendor.
Evaluating vendor lock-in risk during the PoC
Vendor lock-in and data sovereignty are closely related but not identical. Lock-in is the mechanism; loss of sovereignty is often the consequence. During a PoC, the clearest lock-in signals come from proprietary formats, closed APIs, and licensing terms that restrict portability.
Ask directly: what happens to your data and configuration if you cancel the contract? What formats are used for data storage? Are the APIs open standards or vendor-specific? For open-source data platform evaluations, check whether the open-source components are genuinely upstream or whether they are forks with proprietary extensions that create dependency over time. A platform built on unmodified, community-maintained open-source components is significantly easier to migrate away from if your needs change.
Also examine the licensing model carefully. Some platforms offer open-source tooling but wrap it in licensing terms that restrict where and how you can run it. The PoC is the right time to have that legal conversation, not after you have built production pipelines on top of it.
Compliance and regulatory checkpoints to include
If your organisation operates under DORA, the NIS-2 Directive, or sector-specific regulations, the PoC needs explicit checkpoints for each relevant requirement. This is not about passing a compliance audit during the PoC; it is about verifying that the platform architecture supports compliance in production.
Key checkpoints include audit log completeness (are all access and configuration events logged in a tamper-evident way?), data residency controls (can you enforce that data never leaves a specific region or infrastructure boundary?), and supply chain transparency (can the vendor provide a software bill of materials, or SBOM, for every component in the platform?). The last point is increasingly relevant as regulators pay closer attention to software supply chain risk. A data security demo or reference architecture from the vendor can help you understand how these controls are implemented, but you need to verify them in your own environment.
Common gaps between claimed and tested sovereignty
The most common gap is between “data stays in your region” and genuine operational control. A vendor can truthfully say your data is stored in a specific geography while still requiring that all management operations route through their control plane, that encryption keys are held by them, or that upgrades can only be applied through their tooling. Each of these is a sovereignty constraint that would not appear in a standard feature comparison.
A second common gap is around dependency on managed services. A platform may be cloud-agnostic in principle but deeply integrated with a specific cloud provider’s managed database, object storage, or networking service in practice. During the PoC, map every external dependency and ask whether each one can be substituted or self-hosted if needed.
A third gap appears in multi-tenancy and isolation. Vendors often claim strong tenant isolation, but PoC environments rarely test this under realistic load or with realistic access patterns. If multi-tenancy is relevant to your architecture, include explicit isolation tests in your PoC plan.
Turning PoC findings into a sovereignty scorecard
A sovereignty scorecard gives you a structured way to compare platforms and communicate findings to stakeholders who were not in the room during testing. Keep it simple: five to seven dimensions, each scored on a consistent scale, with evidence attached.
Useful dimensions include data portability (can you export everything in open formats?), operational independence (can your team run and recover the platform without vendor involvement?), access governance (are all access controls in your hands and fully auditable?), supply chain transparency (is the software bill of materials available and verifiable?), regulatory alignment (does the architecture support your specific compliance requirements?), and configuration reproducibility (is the full platform state expressible as code you own?).
For each dimension, record what the vendor claimed, what you tested, and what you found. Gaps between claimed and tested behaviour are the most valuable output of the PoC. They are also the most useful input for contract negotiations and for setting realistic expectations about what “data sovereignty” will actually mean in production.
How Stackable supports data sovereignty testing
The SDP is built specifically for organisations that need to verify sovereignty, not just trust a vendor’s word for it. Here is what that looks like in practice:
- Fully open source, no proprietary extensions: Every component in the SDP is available under an open-source licence with publicly accessible source code. There are no hidden forks or closed APIs that create dependency over time.
- Kubernetes-native, runs anywhere: The SDP runs on-premises, in any cloud, on the edge, or in hybrid environments. Your PoC can run in your own infrastructure from day one, which means your test results reflect your actual production conditions.
- Infrastructure-as-code by default: All configuration is declarative and version-controlled. Every platform state is reproducible, auditable, and owned by you, not stored in a vendor’s control plane.
- No telemetry or phone-home behaviour: The SDP does not send usage data or metadata to Stackable by default. Network egress behaviour is transparent and verifiable.
- Modular architecture: Add or remove components like the Stackable Operator for Apache Kafka® or the Operator for Trino independently. No component creates a hard dependency on the rest of the platform.
- Supply chain transparency: Stackable provides SBOMs and maintains a fully traceable software supply chain, supporting the compliance checkpoints that matter most under DORA and the NIS-2 Directive.
If you want to run a structured sovereignty PoC against the SDP in your own environment, get in touch with the team or explore the available demos to see how the platform behaves under realistic conditions.
Related Articles
- What is medallion architecture (bronze/silver/gold) in a data lakehouse?
- Why is data governance harder in a data mesh?
- What are the biggest risks of a data platform migration in 2026?
- How do you handle data governance during a platform migration?
- What happens to your data pipelines during a platform migration?