Competitive Hiring Benchmarks: Is Your Job Spec Aligned with the Market?

If you're writing a data engineer job description and listing 20 required skills, the market median is 11. If you're offering onsite-only for an ML engineer role, 74% of competing postings offer remote or hybrid. If your data scientist listing doesn't mention Python, you're an outlier among 84% of employers who do.

We analysed 5,400 job postings from company career pages across data, product, and delivery roles to build a set of benchmarks that hiring managers can use to pressure-test their own specs. The point is to show you what other companies are actually listing right now, so you can calibrate accordingly.

The skills consensus

Some skills are so widely listed that omitting them signals either a genuinely unusual role or a description that hasn't been calibrated to the market.

For data engineers, Python (85%) and SQL (73%) are the baseline. Below that sits a clear infrastructure tier: AWS (42%), Airflow (39%), dbt (35%), Spark (34%), Snowflake (34%). Any combination of these is standard. Listing all of them is asking for a lot. ML engineers follow a different pattern: Python (80%) is universal, but the next tier is ML-specific frameworks. PyTorch (43%), TensorFlow (31%), and LLMs (27%) form the core. SQL, which dominates every other data role, appears in only 19% of ML engineer postings.

Analytics engineers have the tightest skill consensus of any role in the dataset. SQL appears in 93% of postings, dbt in 70%, Python in 63%. The role is narrow and clearly defined, which makes hiring for it more straightforward but the candidate pool smaller.

For product managers, the picture is softer. No single skill crosses 45% across all PM variants. The closest are Agile (42% for project managers, 33% for programme managers), JIRA (38% and 35%), and stakeholder management (36% and 37%). Technical skills like SQL appear in 12-32% of PM postings depending on the variant.

How many skills is too many?

The number of skills in a posting varies meaningfully by role, and it reveals something about how companies think about the hire.

Data engineers list the most skills at a median of 11 per posting, followed by ML engineers at 10. Data scientists and analytics engineers sit at 9. Project and programme managers are the leanest at 7.

The gap between data roles and management roles makes sense: engineering hires are evaluated against a specific technical stack, while management roles are evaluated against experience and judgment, which are harder to enumerate as discrete skills. But within the data category, there's a lesson. If your data engineer posting lists 18 skills and the market median is 11, you're likely either describing two roles or filtering out candidates who could do the job but don't tick every box.

Salary positioning

US salary data (New York, San Francisco, Denver) gives a clear competitive picture for the roles where we have sufficient data.

ML engineers command the highest ranges at $185k-$265k, followed by data engineers at $166k-$230k. Data scientists sit at $162k-$222k. Analytics engineers are closer to data scientists at $160k-$200k, despite a more specialised and narrower skill set.

On the delivery side, programme managers earn $152k-$200k, a 32% premium over project managers at $115k-$160k. These two roles share nearly identical skill requirements (10 of 15 top skills overlap). The premium is for scope and organisational altitude, and the market prices it consistently.

Working arrangements

Hybrid is the modal arrangement across every role in the dataset, at 27-41% of postings. Remote sits at 22-33%, with data engineers most likely to offer it (33%) and programme managers least (22%). Onsite-only is relatively rare at 11-18%.

The practical implication: if you're insisting on full-time office presence for a data or ML role, you're competing in a pool that represents roughly 15% of the market. That's a choice you can make, but it should be a deliberate one, with compensation or other factors strong enough to offset it.

The interesting thing about benchmarks is how rarely people check them before writing a spec. Data engineers list a median of 11 skills. Management roles list 7. That gap says something about how the market distinguishes between hiring for technical capability and hiring for judgment and coordination. Your spec isn't just a filter. It's a signal about what you think the role actually requires.

---

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