The Seniority Skills Ladder: How Skill Requirements Change as You Level Up

Only 3-7% of the openings we track are explicitly junior. Around half are senior. The rest sit at staff, principal, or director level. The market is structurally senior-heavy, and the skill requirements at each level look meaningfully different from each other.

We analysed skills data across 5,400 job postings from company career pages, covering data, product, and delivery roles in London, New York, Denver, San Francisco, and Singapore. The patterns below show what companies expect at each level, and where the biggest shifts happen.

What seniority actually changes about skill requirements

The instinct is that seniority means more skills, or harder skills. The data tells a different story. Across every role we track, leveling up is defined as much by what drops off your requirements as by what gets added. SQL drops 9 percentage points from junior/mid to senior+ roles across all subfamilies combined. Pandas drops 15 points for data scientists. Excel drops 7 points for product managers. Prompt engineering drops 16 points for AI/ML product managers.

These skills don't become irrelevant. They become assumed. A senior data engineer still writes SQL. But the job spec no longer bothers to say so, because the role's identity has shifted from executing queries to designing the systems those queries run against.

The skills that climb are consistently more abstract: AI (+5 points across all roles), machine learning (+3.4), data analysis (+3.1), A/B testing (+3.1), product strategy (+3.0). Seniority, in the data, is a move from tools to judgment.

Where the ladder diverges by role

The universal pattern holds across the board, but the specific skills involved differ sharply by role type. Here's where the most dramatic shifts happen.

Data scientists: from library proficiency to causal thinking

At junior and mid levels, data science job specs read like a toolkit checklist. Pandas (25%), statistics (25%), NumPy (12%), SciPy (7%). By senior and above, every one of these drops by 5-15 percentage points. What replaces them is instructive: causal inference rises from 17% to 25%, and A/B testing from 8% to 15%.

The shift captures something real about what senior data scientists do. Early in the career, the challenge is executing analysis correctly. Later, the challenge is designing the right analysis to answer a business question. Causal inference and A/B testing are frameworks for making decisions under uncertainty, which is a fundamentally different skill from knowing how to wrangle a DataFrame.

ML engineers: from frameworks to operations

Junior and mid-level ML engineer postings lean heavily on frameworks: PyTorch (50%), TensorFlow (37%), Scikit-Learn (15%), Docker (16%). At senior and above, PyTorch drops 8 points, TensorFlow drops 7, Scikit-Learn drops 8. What climbs instead is MLOps, from 6% to 12%, alongside a broader emphasis on machine learning as a concept (22% to 29%) over any specific implementation tool.

ML engineers also have the highest staff-level representation of any role we track, at 26% of postings. And the salary jump from senior to staff is substantial: $180k-$251k at senior, $215k-$302k at staff in US markets. That's a $35k-$51k jump at the median, the largest absolute increase in the dataset.

Product managers: AI as the senior baseline

For core product managers, the most notable shift is around AI. At junior and mid levels, 13% of postings mention AI as a required skill. At senior and above, it's 20%. Among AI/ML product managers, the same pattern plays out more dramatically: AI goes from 30% to 41%.

What declines tells the other half of the story. Agile drops 10 points for core PMs (29% to 19%). Scrum drops 8. Excel drops 7. Figma drops 5. Backlog management drops 6. These are the mechanics of product development. At senior levels, companies assume you can run a sprint and want to know whether you can set direction for a product that incorporates AI.

The salary progression for core PMs is unusual. Mid to senior is a modest step: $153k-$190k to $165k-$220k. The real jump comes at director+: $208k-$286k, a $43k-$66k increase over senior. For PMs, the financial case for leveling up is strongest at the director transition.

Programme managers: where Agile becomes the senior badge

Programme management shows the steepest single skill climb in the entire dataset. Agile goes from 17% at junior/mid to 38% at senior and above, a 21 percentage point jump. JIRA follows a similar trajectory, climbing 17 points. Scrum rises 13 points. PMP certification rises 11 points.

This is the inverse of what happens in product management, where Agile and Scrum decline at senior levels. In delivery roles, these methodologies become more important with seniority. The interpretation: senior programme managers own the delivery framework for entire organisations. They're defining how work gets done, which is a different relationship with Agile than following someone else's process.

The staff specialisation pattern

One of the clearest patterns in the data is that staff and principal roles mark a point of specialisation, where the toolkit narrows and deepens.

For data engineers, Kafka jumps from 22% at junior/mid to 28% at senior+, and Databricks climbs from 11% to 20%. These are distributed systems and large-scale data platform skills. For analytics engineers, GitHub and Hex emerge at senior+ (neither appears meaningfully at junior/mid). For ML engineers, MLOps doubles from 6% to 12%.

The pattern suggests that staff-level roles are defined by architectural ownership. You're no longer learning a wider range of tools. You're going deeper into the systems that define your domain.

What this means for career planning

The data suggests a consistent sequence across roles: learn the toolkit early, prove you can execute with it, then gradually shift your identity toward the problems you solve and the systems you design.

If you're mid-level and aiming for senior, the gap is rarely about learning another framework. It's more likely about building experience with the skills that sit above the tools: causal inference for data scientists, MLOps for ML engineers, product strategy for PMs, delivery methodology ownership for programme managers.

If you're senior and eyeing staff, the data points toward specialisation. Staff roles go deeper, with a narrower focus on the systems that define a domain. Kafka and Databricks for data engineers. MLOps and production systems for ML engineers. These are bets on a domain, and the market rewards the specificity.

And if you're a PM evaluating where the salary ceiling is, the data is clear: core PM compensation jumps most dramatically at the director+ transition ($208k-$286k in US markets), while ML engineering sees its biggest jump at the staff level ($215k-$302k). The two disciplines reward different inflection points.

---

Based on 5,367 job postings from company career pages, tracking product, data, and delivery roles across London, New York, Denver, San Francisco, and Singapore. Salary data from US markets (New York, San Francisco, Denver) where sample sizes exceed 30. Data covers November 2025 through February 2026. Full interactive dashboard at richjacobs.me/projects/hiring-market.