Hiring guide · updated 2026-05-16

How to hire a data scientist in 2026: scope, channels, and comp

Three of four data scientist postings at Series B+ companies are mis-scoped, per recruiter survey data: the same job title covers analytics work, experimentation work, applied research, and modeling, and the channels and comp bands diverge sharply across those variants. Median total comp for a senior data scientist at top-50 US tech employers is $215,000 in 2026, and the median search runs 55 days. DataDriven.io's 14,200-user audience includes roughly 2,400 active data scientists spanning all three variants plus 1,500 analytics engineers, filterable by SQL depth, Python plus stats, and modeling problem completion for any DS variant search.

$215K
Median total comp
Senior DS, top-50 employers
55 days
Median time-to-fill
Senior data scientist, US
3 of 4
Postings are mis-scoped
"DS" used as a catch-all
<50%
DS candidates write production code
Plan engineering support accordingly

Citable claims from this report

Senior data scientists at top-50 US tech employers earn a median $210,000 to $240,000 total comp in 2026; Series B-D startups outside the top-50 pay $160,000 to $200,000.
n=38 Series B+ placements, Q1 2026
Three of four data scientist job postings at Series B+ companies are mis-scoped across the analytics, experimentation, and modeling variants, according to recruiters at Burtch Works and Selby Jennings.
22 specialist DS recruiter interviews, Q1 2026
Median time-to-fill for a senior data scientist at a Series B+ US company is 55 days in 2026, faster than the 70-day MLE median because the absolute DS pool is larger.
38 Series B+ placements, Q1 2026
Research and applied-scientist-flavored DS roles command a median $310,000 total comp at top-50 employers, versus $195,000 for analytics-flavored DS at the same employer tier.
38 placements segmented by variant, Q1 2026
94 percent of DataDriven.io's 14,200-user Q1 2026 cohort have executed graded SQL problems and 81 percent have executed graded Python problems, the floor skills for any DS variant.
Q1 2026 cohort, n=14,200 monthly actives

Before posting a data scientist role, decide which kind of data scientist you need. The three common variants attract different candidates, command different comp bands ($195,000 versus $230,000 versus $310,000 at top-50 employers for analytics, experimentation, and research variants), and source through different channels. Confusing them is the most expensive mistake in DS hiring at Series B+ companies in 2026.

Analytics-flavored DS writes SQL all day, builds dashboards in Mode or Hex, and runs ad-hoc analyses for product and operations. Often the same role another team calls analytics engineer or product analyst. Highest supply, lowest comp band. Source via verified-skill platforms, LinkedIn Recruiter with strict SQL filters, and niche analytics-flavored boards.

Experimentation DS designs A/B tests, owns the experimentation platform (often built on Statsig, Eppo, or in-house), and writes the analysis. Often partnered with product engineering. Mid supply, mid comp. Source via specialized DS recruiting agencies (Burtch Works, Selby Jennings) and DS community Slacks like Locally Optimistic.

Modeling and research DS builds predictive or causal models, often partnered with ML engineering for deployment. Overlaps with applied scientist. Lowest supply, highest comp. Source via arXiv (cs.LG, stat.ME), statistics PhD program networks at Stanford, Berkeley, and CMU, and Kaggle competition winners filtered by domain.

The list below is ordered for the most common case: experimentation DS at a Series B+ product company. Variant-specific adjustments flagged inline.

Eight channels for hiring a data scientist in 2026, ranked by signal quality and cost per qualified candidate.

  1. 2

    Specialized data science recruiting agencies

    Contingency agencies focused on DS hires. Examples: Burtch Works (DS-specialist, oldest), Selby Jennings, Storm2, Harnham. The good ones scope the role for you (analytics vs experimentation vs research) before sourcing. Compressed time-to-fill of 30 to 45 days.

    Strengths
    • Recruiter helps scope the role
    • 30-45 day time-to-fill
    • Specialist judgment on candidate quality
    Limits
    • 20-25% of first-year salary fee
    • Quality varies widely
    Best for: When you need to hire fast or are unsure how to scope
    Typical cost: 20-25% of first-year base salary
  2. 3

    LinkedIn Recruiter with stats + SQL filters

    Works well for DS because the title is heavily used on LinkedIn and the filters (degrees, current employer, named skills) are well-developed. Reply rates 4-9% for senior DS roles, higher than for MLE or AI engineer because the absolute pool is larger.

    Strengths
    • Widest absolute DS pool
    • Strong filtering by degree, employer, skills
    • Established workflow
    Limits
    • 4-9% reply rates on cold messages
    • DS-title saturation makes filtering harder
    • $10-15K/year per seat
    Best for: Volume sourcing across DS variants
    Typical cost: $10,000-$15,000/year per seat
  3. 4

    arXiv and Google Scholar for research-flavored DS

    Search arxiv.org (stat.ME, stat.AP, econ.EM) and scholar.google.com for authors of papers in your application domain (causal inference, recommender systems, NLP). Variant-specific: only useful for modeling and research DS, not analytics DS.

    Strengths
    • Targets researchers LinkedIn misses
    • High response on specific paper references
    • Strong signal of depth
    Limits
    • Slow, manual sourcing
    • Many researchers want academic-adjacent roles only
    • Only useful for one DS variant
    Best for: Research-flavored DS, applied scientists
    Typical cost: Recruiter time only
  4. 5

    Statistics PhD program networks

    Top stats and biostats programs (Stanford, Berkeley, CMU, U Washington, Harvard, Wisconsin, Toronto) have placement coordinators and active alumni networks. Best for research and modeling DS hiring. Relationship to maintain across multiple hiring cycles, not a one-shot channel.

    Strengths
    • High-quality candidate pool
    • Strong on causal inference and Bayesian work
    • Relationships compound year over year
    Limits
    • Slow ramp
    • Geographically clustered
    • Best candidates have multiple offers
    Best for: Research and modeling DS at large data orgs
    Typical cost: Recruiter time + occasional sponsorship
  5. 6

    Kaggle competition winners (modeling DS only)

    Top finishers in Kaggle competitions have publicly proven modeling skill. Variant-specific: useful for modeling DS roles, less for analytics or experimentation DS (where Kaggle's competition format does not match the day-job).

    Strengths
    • Publicly verifiable modeling skill
    • Strong signal on practical problem-solving
    • Domain-filterable
    Limits
    • Only useful for one DS variant
    • Competition-flavored work differs from production
    • Top finishers often already employed
    Best for: Modeling and applied-research DS
    Typical cost: Recruiter time only
  6. 7

    Data science Slacks and Discords

    Locally Optimistic (the largest analytics-DS community), dbt Slack (analytics-DS overlap), MLOps Community Slack (modeling-DS overlap), and a handful of paid newsletters with private communities (Data Engineering Weekly, Benn Stancil's Substack). Job channels exist; participation precedes successful hires.

    Strengths
    • Direct line to working DS practitioners
    • High-signal small communities
    • Strong for analytics and experimentation DS
    Limits
    • Requires real ongoing presence
    • Small absolute reach per post
    Best for: Companies with strong DS founder or leader presence
    Typical cost: Mostly community time; some featured placements $500-3,000
  7. 8

    Generic job boards

    DS is one of the few data roles where generic boards (Indeed, Glassdoor, Built In Data) still produce reasonable mid-level inbound, because the DS title is broadly recognized. Senior DS hiring still works better through other channels.

    Strengths
    • High inbound volume for mid-level DS
    • Low cost per posting
    • Wide reach
    Limits
    • Low signal-to-noise on senior DS
    • Heavy resume screening
    • Best senior candidates not here
    Best for: Mid-level DS roles
    Typical cost: Free to $250 per posting
Comp band by DS variant (2026, senior IC at top-50 employers)
Research / applied scientist DS $310K
Modeling DS $260K
Experimentation DS $230K
Analytics DS $195K
Analytics engineer (compare) $185K
DataDriven Partners benchmarks, published comp aggregators, Q1 2026

The three questions to scope the role

1. What percentage of the day is SQL? If the answer is more than 50%, you are hiring analytics DS, which overlaps with analytics engineer. Consider the AE title instead if most of the SQL is dbt modeling.

2. Will the candidate own production code? If yes, scope for engineering rigor (testing, code review, CI). If no, you are hiring an analyst with stats depth, not an MLE. Plan engineering support to deploy any models the DS builds.

3. Is the goal to ship features or to inform decisions? Shipping features means modeling DS or applied scientist (overlap with MLE). Informing decisions means analytics or experimentation DS. These attract different candidates and different comp bands.

At-a-glance DS hiring channel comparison

Channels mapped to DS variants. The "best variant" column matters more for DS than for any other data role.

ChannelBest DS variant?Cost?Time to fill?Signal?
Specialized DS agencyAny (recruiter scopes)20-25% salary30-45 daysHigh
LinkedIn RecruiterAll variants$10-15K/yr45-75 daysMedium-high
arXiv / Google ScholarResearch / modelingRecruiter timeLong tailVery high
Stats program networksResearch / modelingRecruiter timeLong tailVery high
Kaggle winnersModeling onlyRecruiter timeLong tailHigh
DS communities (Locally Optimistic)Analytics, experimentation$500-3KLong tailIndirect
Generic boardsMid-level any$0-250VariableMedium (mid only)

Time-to-fill and signal reflect senior DS hires at Series B+ companies in 2026.

5%
Of DataDriven.io's 14,200 active users in Q1 2026 self-identify as data scientists. 94% of all users have executed graded SQL problems and 81% have executed graded Python problems (the floor skills for any DS variant).
DataDriven Partners platform telemetry, Q1 2026 cohort, n=14,200 monthly actives · 2026-05-16

Data scientist variants

Confusing one variant for another is the most common reason a DS hire under-performs. Hire for the actual work.

Analytics DS
SQL-heavy, dashboards, ad-hoc analyses for product and ops. Overlaps with product analyst and analytics engineer. Lowest comp band among DS variants. Source via verified-skill platforms and SQL-focused boards.
Experimentation DS
A/B test design, causal inference, ownership of the experimentation platform. Often partnered with product engineering. Mid comp band. Source via specialized DS agencies and DS community Slacks.
Modeling DS
Predictive or causal models, often partnered with MLE for deployment. Overlaps with applied scientist. Higher comp band than analytics or experimentation. Source via Kaggle, arXiv, and stats programs.
Research / applied scientist DS
New model architectures, methodology research, often with a PhD background. Highest comp band among DS variants. Source via arXiv, stats program networks, and academic conferences.
Analytics engineer (compare)
Owns dbt models and the metrics layer. Not a DS variant but the most-confused adjacent role. If most of the work is dbt modeling, hire an analytics engineer instead.

What to test for in a DS interview

Standard four-block loop. SQL round: window functions, aggregations, common patterns. Stats and experimentation round: A/B test design, p-values, power analysis, common pitfalls (peeking, multiple comparisons, novelty effect). Past-project deep-dive: 60 to 90 minutes on a real analysis the candidate has shipped. Take-home or live coding in Python or R with a real dataset.

For modeling DS roles, replace block four with a modeling round (build a model on a provided dataset, defend choices). For research-flavored DS, add a paper-discussion round where the candidate walks through a recent paper of their choosing.

One opinionated recommendation. Scope the new role against what the existing person actually does, not the job description they were hired against. Three of four DS hires drift from their JD by the 18-month mark, per our partner interviews with Burtch Works and Selby Jennings recruiters in Q1 2026. Replacing a departing DS using the original posting frequently recruits a candidate the team no longer needs.

Frequently asked

What is the difference between a data scientist and an analytics engineer?
Analytics engineer owns dbt models and the metrics layer. Analytics-flavored data scientist writes SQL on top of those models to answer business questions. If most of the work is dbt modeling, hire an analytics engineer; if most is new analysis on existing models, hire an analytics DS.
Should I hire a data scientist or an ML engineer?
Hire an MLE if the goal is shipping a production model. Hire a DS if the goal is informing a business decision with data and stats. Many companies hire DS expecting MLE work and end up with models that never ship.
What is the right comp band for a senior data scientist in 2026?
At top-50 US tech employers, median total comp is $210,000 to $240,000. At Series B-D startups, $160,000 to $200,000. Modeling and research DS run 15 to 25 percent higher than analytics DS. Research DS at frontier labs runs $280,000 to $360,000.
How long does it take to hire a senior data scientist?
Median 55 days from req-open to signed offer at a Series B+ US company. Specialized agencies compress that to 30 to 45 days. DS hiring is faster than MLE or AI engineer because the absolute pool is larger.
Do data scientists write production code?
Fewer than half do. Stats and economics backgrounds typically write less production code; CS backgrounds write more. Plan engineering support to deploy any models the DS builds, especially for analytics and experimentation variants.
What is the difference between a data scientist and an applied scientist?
Applied scientist is the title commonly used at FAANG-tier AI labs for research-flavored DS work, where the deliverable is often a new model architecture. Outside FAANG, 'research data scientist' or 'senior data scientist (modeling)' covers similar work.
Should I use Kaggle as a DS hiring signal?
Only for modeling DS roles. Kaggle format diverges from analytics and experimentation DS work. For modeling DS, top-100 finishers in domain-relevant competitions are a strong signal.
Where should I not advertise a data scientist job?
Generic recruiting newsletters that bundle every tech role, ML-specific job boards (those candidates want MLE work), and any platform promising 'vetted candidates' without disclosing the vetting process.

Sources cited

  1. How to Hire Data Engineers in 2026 · Kore1 · 2026
  2. AI/ML Talent Shortage Strategies for 2026 · CalTek Staffing · 2026
  3. Locally Optimistic data science community · Locally Optimistic · 2026

Related guides

Hire from a skill-verified audience of 14,200 engineers.

14,200 active data, ML, and AI engineers practice on DataDriven.io. 41 percent are senior IC, 78 percent are interviewing within 30 days. Filter by skill, seniority, and geo, then place a pinned listing on every problem page matching the role.