SaaS Agency in Zurich — Build and Scale Your Product

#SaaS Agency in Zurich: What Changes When Your Buyer Is Regulated

Building a SaaS product in Zurich means designing software whose buyer will have to justify its use to a regulator. At Nehos, a multi-tenant SaaS MVP with subscription billing starts from €13,952 excl. VAT, a full platform from €50,000 excl. VAT, and the quote is delivered within 24 hours — in euros, or in Swiss francs on request.

That nuance is worth stating up front, because it drives everything downstream. In most European markets, a B2B SaaS vendor sells a subscription: the customer signs, logs in, and gets to work. In Zurich, the moment your customer is a bank, an independent wealth manager, an insurer or a reinsurer, your product isn't just software anymore — it becomes a line item in their outsourcing inventory. FINMA Circular 2018/3 governs this for banks and insurers: assessment of whether the outsourced function is material, audit and inspection rights that must extend to your own subcontractors, business continuity, an exit plan. Circular 2023/1 on operational risks and resilience, in force since January 2024, adds the identification of critical data and the definition of an acceptable disruption tolerance.

None of this can be patched in three weeks before signature. These requirements show up in your data model, in your deployment strategy, and in the list of third-party services your application calls in production. That's why we scope these questions during the first two weeks of a Zurich engagement, not during legal review.

#The First Trade-off: How Far to Share Your Tenants

Multi-tenant architecture is the decision that shapes a SaaS product, and in Zurich it gets negotiated as much as it gets designed. Here's the framework we lay out in the scoping workshop.

Tenancy model	Actual isolation	Infrastructure cost per client	What a Zurich risk committee says	When to choose it
Shared database, application-level segregation	Logical, depends on your code	Lowest	"Show us your cross-tenant leak tests" — often rejected in private banking	SaaS sold to SMEs and tech teams
One schema per tenant	Strong at the data layer, shared foundation	Moderate	Generally acceptable, with proof of backup separation	The best compromise for a vendor targeting the Swiss mid-market
Dedicated deployment per client	Complete, down to the network	High, scales linearly	The easiest answer to defend in due diligence	Two or three strategic accounts, not your default model

Our position is consistent: design the intermediate model from the MVP onward, and keep the ability to fully isolate an account that demands it. One codebase, one deployment pipeline, and the ability to answer yes to a due diligence questionnaire without rewriting the product. The costly mistake is starting with a fully shared database to move fast, then having to redo everything at the first banking contract — which in this city tends to arrive earlier than expected.

#Data Residency and the Subcontractor Chain

In Zurich, banking secrecy falls under Article 47 of the Banking Act — criminal law, not a contractual clause — and Article 271 of the Criminal Code penalizes acts performed on Swiss soil on behalf of a foreign state. The revised Federal Act on Data Protection, in force since September 2023, adds obligations of its own — including a right to human review of automated individual decisions that produce significant effects. In practice, for a SaaS vendor: your end customers' identifying data doesn't leave the country on the strength of a subcontracting agreement alone.

We therefore deploy Swiss SaaS products on Swiss infrastructure — Exoscale Zurich, Infomaniak — with a written map of every third-party service called in production. The usual blind spot isn't the database: it's the observability tool, the support platform and the transactional email service, which quietly exfiltrate customer identifiers without anyone having decided that. For vendors who also serve regulated French customers, the same application deploys on OVHcloud, up to SecNumCloud level. This sovereignty capability is available to your projects; our own website, by contrast, runs on standard infrastructure.

#Billing a Subscription from Switzerland: What Foreign Engines Miss

Subscription billing is the most underestimated component of a SaaS product, and the Swiss market adds three specifics.

Card payment isn't enough. Your medium and large Swiss customers pay by bank transfer, not by card. A subscription engine wired only for card payments forces your finance team to reconcile transfers by hand, every month, indefinitely. We implement the Swiss QR-bill — QR-IBAN and structured reference — with automatic payment reconciliation from ISO 20022 bank statements, alongside card payment for smaller accounts. A few days of development up front, dozens of hours saved every month after that.

Three currencies and a separate VAT regime. A Zurich vendor commonly invoices in CHF, EUR and USD. Swiss VAT runs at its own rate, and B2B services into the European Union follow the reverse-charge regime at the customer's end — a point to confirm with your accountant, but one that needs to be modeled into the billing engine from day one, not bolted on afterward. A subscription billed in the wrong currency or with the wrong VAT wording gets corrected by hand, credit note by credit note.

Access rights aren't billing. We use Stripe Billing as the payment engine, but the source of truth for entitlements — which plan, which limits, which active add-ons — stays inside your application. That's what lets you later switch payment providers, sell a multi-year contract negotiated off-platform, or grant access to a customer who pays by invoice. The full detail of our approach is on the custom SaaS development page.

#Embedding AI in the Product Without Wrecking Your Compliance Story

This is the number one request from vendors in the region today: adding an AI layer to the product — semantic search over a document corpus, form pre-filling, case summarization, drafting assistance. Zurich has the rare feature of concentrating supply and demand on the same tram line: ETH and the university train the talent, Google runs its largest engineering centre outside the United States here, IBM Research keeps its European lab in Rüschlikon, and the Swiss AI initiative run with EPFL on the CSCS's Alps supercomputer has released Apertus, an open-weight language model with published weights and training recipe.

For a vendor, this availability of open-weight models isn't an academic curiosity — it's a purchasing option. It lets you host inference within the same perimeter as your data, which answers the only question that matters in committee: who can read what we send to the model, and under which jurisdiction. Three rules guide our integrations:

• Your model provider becomes a subcontractor in your customer's outsourcing file. An undocumented foreign API is enough to sink a due diligence review.
• Any feature that influences a decision affecting a person — scoring, pricing, case selection — must keep a human decision point and a record of who decided.
• If you sell into the European Union, the EU AI Act applies to you even if your company and servers are in Switzerland, as soon as the system's output is used within the Union. Switzerland, for its part, has chosen a sectoral approach rather than a horizontal law: most Zurich groups therefore build to the stricter of the two regimes.

The detail of our local practice on these topics is on the AI agency Zurich and AI agents Zurich pages, which cover banking and insurance use cases. To size the opportunity before you build, our free diagnostic tools — ROI calculator, AI Act risk classification — give a first order of magnitude in a few minutes.

#Taking Over and Modernizing an Existing SaaS

A significant share of our engagements don't start from a blank page. The typical scenario in this region: an ETH spin-off or an internal product team has built software that works and has found its first customers, but industrialization never happened. The signs are always the same — every new client triggers a manually cloned deployment, schema migrations happen in the evening, secrets live in the codebase, and nobody knows which features are actually used because there's no usage telemetry.

We start with an audit that answers four questions: what's the real infrastructure cost per client, how much human time does onboarding one more client cost, what's the security exposure, and which parts of the code can be kept. Then we modernize in stages, without service interruption: the product layer, tenancy, billing and industrialization move to us, your core business or scientific logic stays with you. Multilingual support deserves a mention of its own in this market: a Swiss SaaS targeting the mid-market needs to exist in German and English, often French, sometimes Italian — and this requirement gets very expensive if it's bolted on after the fact instead of planned into the content model.

#What Does Building a SaaS Product Cost?

Unlike most agencies, we publish our prices rather than hiding behind "it depends" — the full grid is on the Nehos pricing page:

• SaaS MVP: from €13,952 excl. VAT. Two-week product scoping, six to eight weeks of development, multi-tenant architecture, subscription billing, authentication, guided onboarding, CI/CD, and usage analytics from day one in production.
• Full SaaS platform: from €50,000 excl. VAT. Enterprise SSO (SAML/OIDC), documented public API, ERP and CRM integrations, advanced admin back-office, training for your product team.
• Maintenance and ongoing development: from €750 excl. VAT per month.

These are low-end entry points. The firm quote depends on the tenancy model chosen, the number of integrations, and the documentation standard your regulated customers require — the factor vendors underestimate most often. Billing in euros by default, in Swiss francs on request.

#Working with Nehos from Zurich

Our product and engineering team works out of our Toulouse headquarters, in the same time zone as you: no day lost waiting for an answer, decisions settled by the afternoon. Foued Cherni and Chokri Siala travel to Zurich for the moments that matter — initial scoping, steering committee meetings, architecture review ahead of a customer due diligence. The rest of delivery runs remotely, on a staging environment you can access continuously: you watch the product get built instead of discovering it at delivery.

The economic argument is worth stating honestly. Zurich has one of the highest engineering costs in the world, and your scale-ups are competing for talent against Google, Apple and Microsoft in the same pool. Outsourcing the platform layer — tenancy, billing, infrastructure, industrialization — lets you concentrate local hiring where your differentiation actually lives: your domain expertise and your intellectual property. That's not a universal answer, and we'll tell you plainly if your situation calls for building it all in-house instead.

#Why Nehos

• 47 specialists across product engineering, AI and data, with 200+ projects delivered since 2014.
• An average 340% ROI at 12 months across our engagements, because every project starts from a costed use case rather than a feature wishlist.
• Published prices and a quote within 24 hours — you know whether a project fits your budget before you ever call us.
• A genuine sovereignty capability available to your projects — Swiss hosting for your FINMA-regulated clients, OVHcloud up to SecNumCloud level for your regulated French clients.
• An editorial and acquisition practice for the post-launch phase: we ran a twelve-month content marketing and LinkedIn ambassador programme for a French B2B SaaS scale-up, generating 7 attributed deals worth €320k in ARR (anonymized case study details).

To place your project within our view of the vendor market, see our SaaS and scale-up sector page and the SaaS software guide, which covers the trade-off between buying an off-the-shelf solution and building your own. The overview of our local activity is on the Nehos Zurich hub page.

A simple test before you talk to us: ask your team how many human hours it currently takes to bring a new client onto your platform. If the answer is more than an hour, you don't have a features problem — you have an architecture problem, and that's exactly where a well-run SaaS engagement earns its keep.