Data Engineering as a Service: when subscription beats hiring

You need pipelines that run, a warehouse you can trust, and someone owning the whole thing at 3am when something breaks. The default playbook is to hire a senior data engineer and hope you can keep them. That playbook is slower, riskier and more expensive than most Swiss SMBs realise.

Data Engineering as a Service is the third option. It is not consulting and it is not a contractor. It is a flat-rate subscription where a structured workflow turns your needs into shippable specs, a small team implements on a managed platform, and the price tracks the scope you actually keep active. This guide explains when subscription beats hiring, when it does not, and what the workflow actually looks like so you do not just trade one expensive lock-in for another.

The hiring math nobody likes to write down

A senior data engineer in Switzerland costs CHF 130k to 180k base, plus social charges, plus equipment, plus the manager time that supervises them. Realistic loaded cost: CHF 200k+ per year. That number assumes:

1. You can find one. Senior data engineers fluent in modern stacks (dbt, Airflow, Snowflake or BigQuery, Terraform) are scarce in Geneva and Zurich, and even scarcer in cantons outside the financial centres.
2. They ramp in three to six months. The first quarter is mostly context absorption. You pay full salary, you get partial output.
3. They stay. Median tenure for engineers in this segment is 18 to 24 months. Recruitment, onboarding and handover repeat on that cycle.
4. They are happy doing the unglamorous work. Most senior data engineers were hired to build models, not babysit a Stripe-to-warehouse pipeline that breaks every other week. Attrition follows.

The honest math: CHF 200k per year, 6 months of ramp, 18 months of useful output, a single point of failure on call.

The drowning-team math

If you already have a small data team, the failure mode is different. Two or three engineers cannot reasonably cover ingestion, transformation, warehouse hygiene, monitoring, governance, BI tooling and incident response, AND ship the analytical work the business actually asks for. Something gets dropped.

What gets dropped is usually the work that moves the needle. Pipeline maintenance is loud and visible. New attribution models, customer-segment dashboards and ML features are quiet and skippable. The team ends up firefighting instead of building. You do not need to fire anyone. You need to give them air cover for the boring layer.

What "as a service" actually means at Jelzia

Done well, Data Engineering as a Service is a workflow, a platform and a pricing model wrapped together.

The workflow: single-WIP through a Kanban you share with us.

When you sign up, you are added to a Linear board (or equivalent Kanban) that we run jointly with you. An embedded agent in the board helps you turn rough needs into specific, executable specs. The agent's job is to make sure that what enters the backlog is actually implementable, not a vague wish.

You can pile as many items as you want into the backlog. The discipline is that only one item lives in "To Do" at any moment. We pull that one item, implement it, run it against every Definition of Done criterion you set, and hand it back to you for validation. You have a clear window to validate. If you let the window pass, the implementation is considered accepted (with a follow-up safety net the next time you spot something off). Then the next item from the backlog moves up.

That single-work-in-progress rule is the most important part. It is what stops the chaos that kills internal data teams: ten half-done things in flight, no clear owner, no one knowing what is shipping when. With single-WIP, the answer to "what is happening this week" is one named ticket.

The platform: JDP, in-house and invisible to you.

Implementation runs on JDP, the Jelzia Data Platform. JDP is our unified, opinionated stack for ingestion, transformation, orchestration, monitoring and governance. We operate it, we optimise it, we keep it boring. You do not have to think about which orchestrator, which warehouse vendor, which monitoring tool: that is our problem. The data remains yours throughout (more on that below).

You should not have to care about JDP day-to-day, but it is worth knowing it exists. It is what lets us deliver in days instead of weeks once a spec is clear.

The pricing: flat per active scope, sub-linear.

Pricing is not hourly and it is not "one annual lump for unspecified work". It is a flat monthly rate tied to the scope of what you keep active. Add a pipeline, the rate goes up. Add a second pipeline, the rate goes up by less than the first one. Trim something out of scope, the rate comes down. The structure rewards consolidation: the more we run for you, the better your effective price per unit of work.

That makes the bill predictable in a way an in-house team's loaded cost is not. You always know what next month's invoice looks like before next month starts.

When subscription beats hiring

Subscription wins when at least three of these are true for your business:

• You generate less than CHF 50M revenue per year. Below that scale, you cannot keep two senior data engineers usefully busy, and one is a single point of failure.
• Your data work is largely orchestration and maintenance, not novel ML research. If the heavy lift is "make Stripe and Shopify produce reliable revenue numbers in a warehouse", a structured workflow on a unified platform has done it 30 times already.
• You need predictability. A flat monthly rate beats a salary plus recruiter fees plus the surprise of a senior leaving in month 14.
• You are not yet ready to commit to a permanent in-house team. Subscription gives you a graduated path: subscribe today, hire when the volume genuinely justifies it.

When hiring still wins

Be honest with yourself. Hiring beats subscription when:

• Your data is genuinely novel and central to your competitive moat. Algorithmic-trading firms, biotech research labs and ML-first products are usually better off with permanent staff.
• You already have three or more data engineers and the volume of original work justifies that headcount. Adding subscription on top is duplication.
• Your security posture forbids any external party processing production data on a shared platform. This is rare in Swiss SMB but real in defence and parts of pharma.

If two or more of these apply, build the team. DEaaS is not a religion.

What to look for in a managed data partner

If you are evaluating providers (Jelzia included), the questions worth asking are not the ones a typical RFP asks. Ask:

1. Show me the workflow. Where do specs live, who writes them, who validates? "Slack and trust me" is not a workflow.
2. What is your work-in-progress limit per client? No limit means scope chaos in disguise.
3. Who owns the data, where does it physically sit, and what happens on day one of cancellation? A good answer names the owner (you), the location, and the cancellation playbook in one breath.
4. What is your delivery time for a small, well-scoped change once credentials and access are in place? Days is plausible. Weeks for trivial work is a red flag.
5. How does pricing change when scope grows or shrinks? A clear scaling rule beats opaque "let us put a quote together".
6. What is the cancellation clause and the data exit motion? Month-to-month cancellation and a documented data export beats annual prepay every time.

The Jelzia way

We run the subscription light. One named lead per client, one shared Kanban board with the spec agent embedded, single-WIP discipline, flat scope-based monthly rate. Implementation runs on JDP. The data stays in your account or on storage you nominate; we never use your data for anything but the work you ticketed.

If you want to stop, we ask you what you want done with the data: export to a destination you choose, delete, or transfer. Your call, no negotiation.

If that sounds like the answer for your stage, the DEaaS service page has the scoped pricing and the included workflow. The fastest path to a yes-or-no is a 30-minute discovery call: book one here.

FAQ

Who owns the data? You do, always. Jelzia operates the technical layer (pipelines, platform, monitoring) but the data is yours throughout the engagement and on the way out.

What happens to the data if we cancel? We ask you what you want: export to a destination you choose, delete in place, or transfer. Your decision, executed on your timeline.

Is this just a contractor in disguise? No. Contractors bill by the hour and stop when the budget runs out. DEaaS is a flat-rate ongoing service with a structured Kanban workflow, single-WIP discipline, and a managed platform underneath. The difference matters most when something breaks at 11pm on a Saturday.

What is the embedded spec agent? An assistant that lives inside the Kanban board and helps you turn fuzzy needs ("we want better reporting on churn") into specific, executable specs ("daily refresh of churned-customer cohort, segmented by acquisition channel, into the marketing schema"). It catches under-specified work before it enters our To Do column.

How fast can a piece of work ship? For a small, well-scoped change with credentials and access already in place, a few days is realistic. The single-WIP rule means there is never a queue of ten things ahead of yours.

What if we already have a data engineer? Then DEaaS covers the platform and operations layer so they can focus on novel work. Several of our subscriptions look like this.

Can we start with subscription and hire later? Yes, and it is the most common path. We typically run for 12 to 18 months while the business grows enough to justify a permanent first hire, then transition with a documented handover of the work in scope.

---

Building a data team is one of those decisions that compounds: the cost of getting it wrong is paid in delayed analytics, frustrated hires and a CTO who spends Saturdays debugging Airflow. Subscription is not the right answer for every business. For Swiss SMBs that need reliable data infrastructure with a clear workflow and predictable scope-based pricing, it is the answer the market has been waiting for. If you want to talk through whether it fits your stage, book a discovery call and we will tell you honestly either way.