Last year I was helping a friend prep for senior data engineer interviews. He'd been building pipelines at a Series B for four years, solid production experience, and wanted to know what number to put in the salary field. So he did what everyone does: checked Glassdoor, Indeed, PayScale, and Levels.fyi.
He got four numbers. They disagreed by over $120,000.
Glassdoor said $133K. PayScale said $100K. Indeed's senior number sat at $216K. Levels.fyi split the difference at $157K. Same title, same country, same year; four answers that can't all be right. And here's the thing: none of them are lying. They're just counting different people, over different time windows, with different biases baked in. The result is that candidates trying to benchmark their data engineer salary are pricing themselves against a number that doesn't represent their actual market.
This is a problem. In a hiring environment where 52,050 tech workers got laid off in Q1 2026 alone, where senior roles take 60 to 90 days to fill, and where title inflation has made "data engineer" mean three different jobs depending on who's posting, getting your number wrong has real career cost. You either leave $30K on the table or you overshoot and get ghosted. Both outcomes trace back to the same root cause: the data you're benchmarking against is broken.
Why Every Salary Site Disagrees by $120K
Each major compensation source has its own rot problem. Understanding the bias is more useful than trusting the number.
Glassdoor reports $133,484 average across 32,984 submissions. The issue: it's entirely self-reported, and higher earners submit more frequently. The person who just got a $180K offer is more motivated to log it than the person who accepted $115K and moved on. The sample skews up.
PayScale reports roughly $100K. That sounds low because it is; 71% of their data engineering respondents are mid-level or junior. PayScale validates every data point and refreshes on sub-90-day cycles, which makes it the most accurate floor for what actually clears at offer stage. But candidates see $100K and panic. They shouldn't. They're looking at a junior-weighted average.
Indeed sits at $216K for senior roles. The problem here is temporal: Indeed averages job postings going back 36 months. Their June 2026 number includes postings from June 2023, before the layoff waves, before signing bonus compression, before the market shifted. You're benchmarking against fossil data.
Levels.fyi pegs the median at $157,450, but this population skews heavily toward top-tier tech companies and excludes non-tech firms where data engineers earn 20 to 35% less. Google's median is $278K. Capital One's is $130K. That's a $148K spread for the same title on the same platform.
The salary data isn't wrong. It's measuring different populations, different time windows, and different definitions of the job. Once you know which population you're in, the number becomes useful. Until then, it's noise.
The practical damage is real. A mid-market data engineer sees Glassdoor's $133K, anchors there, and never learns that the number includes FAANG outliers pulling the average up. Or worse, they see Indeed's $216K senior figure and counter-offer at a number that makes the hiring manager close the tab.
Role Title Chaos Is Pricing You Against the Wrong Pool
Here's the less obvious problem: the job you're benchmarking might not even be the job you're doing.
Analytics engineers earn $155K to $195K median in 2026. ML engineers command a 38% salary premium over data engineers at mid-career. Data scientists occupy yet another band. These are different roles with different compensation structures. But companies routinely mislabel them.
Analytics engineer postings grew 114% from 2023 to 2024, yet dbt Labs openly admits the title boundaries are blurring. Analysts drift into dbt modeling. Data engineers adopt dbt as standard tooling. The result: a "$150K dbt role" could be transformation work (analytics engineer) or pipeline infrastructure (data engineer), and the salary sites have no idea which one they're counting.
37,000 data engineering jobs post monthly on average, but a significant portion of those are mislabeled analytics engineer, ML engineer, or data scientist roles. When a company posts "Senior Data Engineer" but the job is really dbt plus Snowflake plus stakeholder dashboards, that's an analytics engineer role at data engineer pricing. The candidate benchmarks against infrastructure DE salaries ($115K to $160K) when they should be benchmarking against analytics engineer salaries ($155K to $195K). That's a $30K to $40K miss.
The reverse kills you too. An analytics engineer who sees ML engineer salary data and anchors at $190K gets rejected as "overpriced" for the actual scope.
The litmus test isn't the title. It's the job description. If it says dbt, Snowflake, and "stakeholder reporting," you're an analytics engineer regardless of what the posting calls you. Benchmark accordingly.
2023 Job Ads Are Still Haunting Your 2026 Number
Indeed's 36-month lookback window deserves its own section because the implications are worse than they look.
In 2023, the median data engineer salary was $117,446. By June 2026, Indeed reports $136,776. That looks like 16.5% growth over three years, which isn't terrible. But the number is being held down by every posting from 2023 and 2024 that's still sitting in the average.
Here's what makes this especially misleading: 68% of tech job postings included explicit salary ranges in 2025, up from 45% in 2023. Pay transparency laws made the data more granular. But Indeed weights all 36 months equally. A vague salary range guess from a 2023 pre-transparency posting counts the same as a precise, legally mandated range from 2026. Higher sample size, stale composition.
Then there's the ghost job problem. One-third of employers admit to posting inactive roles. Greenhouse data found 18 to 22% of listings are never filled. Stale 2023 postings are more likely to be dormant, and they're inflating the denominator. You're benchmarking against jobs that don't exist anymore.
The senior role divergence tells the real story. The $60K gap between mid-level ($133K) and senior ($175K) data engineers in 2026 suggests the market has repriced for experience. But the aggregate average is anchored by fossils. If you're mid-career, the number you see is artificially low.
Layoffs Created a Tier, Not a Glut
52,050 tech workers laid off in Q1 2026. A 40% jump over Q1 2025. Oracle cut 21,000. Amazon cut 16,000. Dell cut 11,000. Sounds like a buyer's market.
It's not. Or at least, not uniformly.
Those 52K cuts coexist with 67,000 active software engineering job postings in the same quarter, the highest posting volume in three years. Companies are cutting commoditized roles while hoarding data engineers, ML engineers, and security specialists. A junior full-stack engineer is in a buyer's market; a senior data engineer with Airflow and Spark experience is not. The "layoff market" narrative breaks down completely by skill tier.
But here's the asymmetry that actually matters: companies take 60 to 90 days to fill senior roles because they're running multiple candidates in parallel. Individual candidates spend 3 to 9 months searching. The employer can wait. The candidate runs out of severance. That's where negotiation leverage shifts; not because the market is soft, but because one side has a deadline and the other doesn't.
The data on negotiation is striking. Data engineers who negotiate earn $24,479 more annually, an 18.83% increase. 85% of counter-offers get at least partial acceptance. 70% of hiring managers expect you to negotiate. Only 44% of candidates actually do it. The $120K gap between salary sources is partly a measurement problem, sure. But it's also partly behavioral. The spread between 25th and 75th percentile reflects negotiation winners vs. passive accepters, not just market fragmentation.
Engineers with current cloud and security skills close offers in 2 to 4 weeks. Everyone else faces the full timeline. Skill specificity determines leverage more than market conditions.
What Number to Actually Put in the Field
Stop averaging the averages. Here's the hierarchy of sources, from most to least useful for your career planning:
Levels.fyi is best for FAANG and top-tier tech. Filter by company, level, and location. The by-company variance is massive ($278K at Google vs. $130K at Capital One), so the aggregate median is useless. You need the company-specific number.
Glassdoor is useful for the 25th to 75th percentile range at your target company, if they have enough submissions. The $141K to $219K senior DE range tells you more than the $175K mean.
PayScale is the most accurate floor. If you're at a non-tech company or early in your career, this is closer to your reality.
Indeed is the least useful for current benchmarking. The 36-month window buries the signal.
The actual number you put in the field should be the Levels.fyi or Glassdoor 75th percentile for the specific company, then negotiate. If 70% of hiring managers expect negotiation, pricing yourself at the median is pricing yourself to get negotiated down.
And one more thing the salary sites never show you: base salary is barely over half the employer's total cost to hire. That $200K base offer costs the company $240K to $290K when you add payroll tax, benefits, recruiting fees (18 to 25% of first-year base), and onboarding ramp. They have more room than you think. The question is whether you know enough about your own market to ask for it.
If you're prepping for the senior and staff loops where compensation actually diverges, strip back the "system design for software engineers" mentality; we built system design for data engineers with datadriven around pipeline architecture problems, not the load-balancer trivia that SWE prep loves and DEs never face on the job.
The salary data is broken. The titles are broken. The timelines are longer. None of that changes the fact that data engineering compensation is strong and growing for engineers who know what they're actually worth. The trick is figuring out which population you belong to, not which average to believe.
What's the biggest gap you've seen between what a salary site reported and what you actually earned or were offered?
Top comments (0)