"We need a data scientist" is one of the most common briefs a Personalvermittlung receives, and one of the least precise. Sometimes the company actually needs a data engineer to build pipelines, sometimes an analyst to build dashboards, sometimes an ML engineer to put a model into production. When the brief is wrong, the placement fails no matter how good the candidate is. This guide walks through what a data scientist actually does, what the German market for the role honestly looks like, and how to separate candidates who create business impact from candidates who create notebooks.
What a data scientist actually does, and the roles it gets confused with
A data scientist turns a business question into a statistical or machine-learning problem, answers it with data, and translates the answer back into a decision or a product feature. That sits in the middle of a family of roles that overlap on tools but differ completely in output, and job descriptions in Germany mix them constantly.
| Role | Primary output | Hire this instead if you need… |
|---|---|---|
| Data scientist | Answered business questions: models, experiments, causal analyses that change decisions | — |
| Data analyst / BI analyst | Dashboards, reporting, descriptive analysis of what already happened | Recurring reporting and KPI visibility |
| Data engineer | Reliable pipelines, warehouses, and data infrastructure | Data that is currently scattered, slow, or untrustworthy |
| ML engineer | Models running in production under latency and reliability constraints | A prototype that must become a product |
| "AI specialist" (undefined) | Varies wildly; often a warning sign in a brief | A sharper role definition first |
The German market: crowded at the bottom, scarce at the top
The data science market in Germany and the wider DACH region has split. Universities and bootcamps produce a steady stream of entry-level candidates, so junior openings attract high application volume. Senior data scientists who have owned a problem from question to measurable outcome, and who can operate with stakeholders in German-speaking organizations, remain genuinely hard to find, and the GenAI wave has pulled part of that senior pool toward LLM-focused roles. The figures below are broad ranges we observe in the market, not guarantees; individual offers vary by industry, region, and company stage.
| Level | Permanent (gross annual salary) | Freelance (day rate) |
|---|---|---|
| Junior (0-2 yrs) | ~€45-58k | Rarely freelance at this level |
| Mid-level (2-5 yrs) | ~€55-75k | ~€600-850 |
| Senior (5+ yrs, proven impact) | ~€75-100k | ~€700-1,100 |
| Lead / principal | ~€95-130k | ~€900-1,300 |
What good candidates look like: signals worth vetting
Data science is a field where the vocabulary is easy to learn and the judgment is not. Portfolios full of Kaggle notebooks and tutorial projects tell you little about whether someone can handle messy company data and skeptical stakeholders. These are the signals that separate the two, and how to check them.
| Signal | What it tells you | How to verify |
|---|---|---|
| End-to-end project ownership | Can go from vague question to adopted result, not just model metrics | Have them walk through one project including the parts that failed |
| Statistical judgment | Knows when a simple baseline beats a complex model | Ask when they chose not to use ML, and why |
| Communication with non-technical stakeholders | Findings actually change decisions | Case discussion: explain a real analysis to a business audience |
| Working code and data hygiene | Analyses are reproducible, not one-off notebooks | Short code review of a realistic sample, not algorithm puzzles |
| Impact framing | Talks about outcomes (revenue, churn, cost), not only AUC scores | Ask what changed in the business because of their work |
Where recruitment goes wrong for this role
Data scientist searches fail in predictable ways, and most of the failure is built in before the first interview.
| Failure mode | Why it happens | What it costs |
|---|---|---|
| Wrong role in the brief | "Data scientist" used as a catch-all for data work | A hire who is miscast from day one and leaves within a year |
| Keyword screening | Recruiters match Python/TensorFlow keywords instead of judgment | Interview slots burned on candidates who can't scope a problem |
| Academic-only evaluation | PhD and publications treated as proxy for business impact | Strong researcher, struggling practitioner |
| No data foundation in place | Company hires the scientist before the data is usable | An expensive hire spending months doing data engineering |
| Slow process | 4-6 week pipelines while candidates hold parallel offers | The best shortlist candidates accept elsewhere |
Permanent hire vs. freelance data scientist
The right contract model depends on whether the work is a standing capability or a bounded question. Many companies get the best result by sequencing: a freelancer to validate that the data and the question are ready, then a permanent hire to own the capability.
| Dimension | Permanent hire | Freelance / embedded |
|---|---|---|
| Best for | Ongoing product analytics, experimentation culture, model ownership | Bounded analyses, PoCs, feasibility studies, maternity/paternity cover |
| Time to start | Often 3+ months including notice periods | Days to weeks |
| Cost shape | Salary + employer costs + recruiting fee, long-term | Higher daily cost, zero long-term commitment |
| Knowledge retention | Compounds inside the team | Leaves with the freelancer unless handover is contracted |
| Risk profile | Mis-hire is expensive and slow to correct | Mis-fit is visible and correctable within weeks |
How a specialized recruiter or network changes the picture
A generalist Personalvermittlung can source data scientist CVs; what it cannot do is tell you which of them can carry a business question to a result. A specialized, technically vetted network changes three things: the brief gets corrected before the search starts, candidates arrive pre-assessed on real work rather than keywords, and the process moves at the speed the senior market requires. Aiporate's model is built around exactly this: role scoping with the client, evidence-based technical vetting by people who have done the work themselves, and a vetted shortlist within 72 hours of a completed brief.
| Step | Generalist agency | Specialized network |
|---|---|---|
| Brief | Takes the job description as given | Challenges the brief: scientist, analyst, or engineer? |
| Screening | Keyword match on CVs | Structured technical assessment on realistic problems |
| Shortlist | 8-10 loosely matched CVs in weeks | 2-3 vetted candidates, benchmark: 72 hours |
| After placement | Guarantee clause | Fit feedback loop, option to start embedded before converting |
