What buyers want to know

The answer, in business terms

This page exists to make security reviews, DPIAs, and vendor assessments faster. Here is the architecture commitment behind the product.

Swiss-only execution

Address parsing, locality resolution, street matching, and scoring are executed on our Swiss-hosted stack.

No third-party inference path

Submitted address payloads are not sent to external generative models, LLM APIs, or foreign inference providers.

Built for regulated workflows

The result is easier privacy governance for teams in finance, health, public sector, SaaS, and enterprise environments.

Explicit commitment

Do submitted addresses leave our servers or Switzerland?

No.

When you call HelvAddress normalization, the address stays within our Swiss processing environment. We may rely on technical providers for infrastructure or service operations, but they do not run the normalization engine and they do not receive the request body for inference, enrichment, or external matching.

No outbound call to third-party AI providers
No outbound call to external geocoding APIs
No handoff to foreign-hosted normalization services
A clearer position for privacy, secrecy, and data residency requirements
Operational guardrails

Normalization logic runs on Swiss-hosted HelvAddress infrastructure.

The request flow excludes third-party generative AI, LLMs, and external inference APIs.

Official Swiss data sources and internal algorithms power matching and qualification.

This separation helps keep procurement, legal, and security assessments straightforward.

Hosting provider & references

Our web and hosting infrastructure is operated in Switzerland with Infomaniak. For security, privacy, and procurement teams, we provide the official provider references below.

Infomaniak privacy policy
Infomaniak legal / compliance information
High-level architecture

How the engine works without exposing your data

We are transparent about the operating model without disclosing proprietary implementation details. The important point: the full normalization flow stays inside the Swiss HelvAddress stack.

High-level Swiss-only normalization flow Input Free-text address Input prep Harmonize writing Component detection ZIP, locality, street, number Swiss context Language & geography Federal matching RegBL + swisstopo Result qualification Score, precision, alternatives External APIs Not in request path Structured output Trusted Swiss result
What happens at a high level
1
1. Prepare the input

The engine standardizes the raw input so it can reason on a stable, consistent representation.

2
2. Detect useful components

It identifies locality, postal code, street, house number, and contextual clues even when the order is irregular.

3
3. Resolve Swiss context

It uses Swiss geographic and linguistic context to reduce ambiguity as early as possible.

4
4. Match and refine

It searches the correct territorial scope and can refine the match down to building level when the data allows.

5
5. Return a qualified result

The API returns a normalized address, precision level, confidence score, and relevant alternatives when needed.

Why this matters commercially

A stronger answer for enterprise buyers

Data sovereignty is not just a legal footnote. It reduces friction in sales cycles and lowers operational risk.

Faster security reviews

Security and privacy teams get a direct, explicit answer instead of vague wording about third parties.

Less vendor risk

Fewer subprocessors in the inference path means fewer unknowns around confidentiality, transfer risk, and data handling.

Better fit for sensitive use cases

A Swiss-only processing model is easier to justify for regulated sectors, privacy-sensitive workflows, and critical address data.

Common questions

Questions we hear from data, security, and procurement teams

Short answers you can reuse internally.

Do you use third-party generative models or LLM APIs to normalize addresses?

No. The normalization path does not call third-party generative models, LLM APIs, or external inference endpoints.

Are submitted addresses sent outside Switzerland or outside Europe for processing?

No. Submitted addresses are processed on our Swiss infrastructure and are not sent abroad for normalization or inference.

What about hosting or operational service providers?

We may use trusted technical providers to help operate the platform, but they do not power the normalization engine and do not receive address payloads for external processing.

What powers the matching engine?

Official Swiss federal datasets, Swiss-specific heuristics, and in-house matching logic designed for multilingual, real-world address input in Switzerland.

Need a crisp answer for your DPO, security lead, or procurement team?

Send us your questions. We can explain the operating model clearly and help your team evaluate fit.

Contact us Try the live demo Explore normalization