I did somewhere around 20 interview loops during my last serious job search. Some went well. Some went so poorly I still think about them in the shower. But here's what I took for granted at the time: every single one of those loops had a real human on the other side, evaluating real candidates for a role that actually existed.
That's no longer something you can assume.
48% of job postings in the tech and information sector are ghost jobs. Not roles that are slow to fill. Not positions stuck in budget approval. Roles that were never intended to result in a hire. 93% of HR professionals admit their employer posts them. 45% do it regularly. And you, the data engineering candidate grinding through your 12th application this week, have absolutely no way to tell which half of the market you're looking at. Job seekers spend an average of 12 hours per week on applications. That's 26% of a full work week burned on submissions, and roughly half of them are going nowhere.
The Ghost Job Numbers
The macro picture is damning once you actually look at it. Applicants per posting nearly doubled from 46 in 2021 to 95 in 2025. Completed hires dropped 20% in the same window, from 1.34 million to 1.05 million. Job posting volume stayed flat. More candidates chasing the same number of postings that produce fewer hires. The funnel is broken because it was designed to be broken.
Companies ghost-post for 3 reasons, and none of them have anything to do with filling a seat:
Investor optics. 43% of companies post roles to signal growth. A board deck showing "200 open headcount" looks like momentum. It's a press release disguised as a job listing.
Employee intimidation. 62% post to make current employees feel replaceable and work harder. That "Senior Data Engineer" listing on LinkedIn? It's not for a new hire. It's a message to the person already in the seat: we can replace you whenever we want.
Pipeline farming. Your resume goes into a database. If a real role opens in 6 months, they've got a pre-built candidate pool they didn't pay a recruiter for. You provided free labor and got nothing back.
70% of hiring managers view this practice as acceptable. Not reluctantly tolerated. Acceptable. This isn't a process failure or a breakdown in the system. It's the system working exactly as designed.
The data engineering market is genuinely growing, 15% annually, with role-specific hiring up 23% year over year. The demand is real. But the job board signal is so polluted with ghost postings that you can't distinguish demand from decoration. You're searching for real listings in a haystack where half the needles are plastic.
Where Hiring Actually Concentrated
Here's the contradiction nobody talks about honestly. The market isn't frozen. It's fractured.
Databricks posted 757 open roles after raising $5 billion at a $190 billion valuation in August 2026. 317 of those are senior positions. The company crossed $7 billion in annualized revenue with 80% year-over-year growth, is free-cash-flow positive, and hasn't done a single layoff. That's what real hiring looks like.
Now look at everybody else. 66% of CEOs surveyed, 350+ leaders managing $19 trillion in collective market cap, are freezing or cutting hiring through the rest of 2026. 80,000 tech jobs were cut in Q1 alone. Over 1,600 companies announced mass layoffs by late March.
The market isn't uniformly dead. It's concentrated. A handful of well-capitalized companies are hiring aggressively while the broader market ghost-posts for optics. Candidates chasing the 48% are burning unpaid hours on false positives while the real openings sit at Databricks, Anthropic, and a short list of others.
The bifurcation goes deeper than company size. Entry-level data engineering postings dropped 67%. Only 3% of 2026 listings explicitly target juniors. Meanwhile, senior roles routinely exceed 90 days to fill, and companies are running 42% more interview rounds per candidate than they did in 2021. What used to be a phone screen plus 2 interviews has bloated into 4 to 6 rounds, take-homes, and panel conversations. The average candidate spends 23.3 hours interviewing before receiving an offer. That's at companies that are actually hiring, not the ghost posters.
This creates a math problem most people don't think through. If you're applying broadly to 50 roles, roughly 24 are ghosts. Of the 26 that might be real, most are running 60 to 90 day cycles with 4 to 6 rounds each. The spray-and-pray approach collapses when half your targets are cardboard cutouts and the real ones take 3 months to close.
And here's the kicker: the best candidates leave the market in 10 to 14 days. Companies running 60 to 90 day hiring cycles aren't competing for top talent; they're fishing in the second tier by default. Every additional week in the cycle increases the chance that the person they want has already signed somewhere else. 43% of candidates have turned down jobs due to poor hiring experience alone.
The better strategy is unglamorous: identify the 5 to 10 companies in your niche that are provably growing. Check earnings calls. Check funding rounds. Check LinkedIn headcount trends over the last 6 months. Then commit to the long cycle at each one. That's where your 23 hours of interview prep should actually go.
20 Hours of Unpaid Work for a Role That Doesn't Exist
Take-home assignments dominate 33% of top-tier company screening processes. The recommended length is 4 hours. What candidates actually report is 10 to 20 hours. And a 27% interview-to-hire ratio means 3 out of 4 people completing these assignments get nothing.
Now layer in the ghost job rate. If 48% of postings are fake and take-homes run 15 to 20 hours, engineers are routinely spending entire weekends building pipelines for roles that don't exist. I know an engineer who completed a weekend-long project for a mid-stage startup, got rejected within hours with zero feedback, and found the same role reposted 4 months later with the identical description. That's not a slow hiring process. That's resume collection with extra steps.
The screening process has a built-in failure rate that punishes qualified people by design. A properly prepared candidate has a 22% chance of failing a single technical phone screen due to variance alone. Not lack of skill. Variance. That's before the take-home, before the panel, before the "depth of reasoning" round where someone decides your answer was correct but insufficiently explained.
I've watched engineers with 10 YOE get rejected after passing SQL and system design because the panel cited "concerns about depth of reasoning." For a problem identical to multi-source ingestion work the candidate had built 3 times in production. The interview measures interview performance. The job measures engineering ability. These are 2 different skills, and conflating them is how companies reject qualified engineers and feel rigorous about it.
FAANG and peers openly admit to calibrating their processes to reject good candidates by design. They'd rather miss 10 strong engineers than risk one bad hire. That's a reasonable corporate strategy and a genuinely hostile candidate experience, especially when you're on loop number 15 and you just found out that loop number 12 was for a ghost.
The AI angle makes this even more absurd. If a model can spit out a clean solution to a medium LeetCode problem, what does asking it in a take-home actually tell you about the candidate? That they memorized something a machine produces on demand? The signal from take-homes was always thin. Now it's basically noise.
Fix Your Job Search Before It Eats You
Stop applying to ghosts. The red flags are learnable.
Check the company careers page. If the role exists on LinkedIn or Indeed but is missing from the company's own site, it's almost certainly dead. This is the single most reliable signal. Legitimate active openings always track on company websites; aggregator-only postings are resume harvesting.
Look at posting age. Average time to fill a role is 33 days. Anything live past 45 to 60 days with no updates is dormant. A posting reposted every 2 to 3 weeks with "urgent" language that never progresses to a phone screen is pipeline farming.
Check for recent layoffs in that department. A company that cut 15% of its data org last quarter and is now listing 8 data engineering roles is performing for its board, not hiring engineers.
Verify the role exists organizationally. Search LinkedIn for current employees with that title at that company. If nobody holds it, nobody recently left it, and there's no hiring manager posting about it, the headcount was never approved.
The 3-signal rule is your decision boundary: posting older than 45 days, missing from company site, no response after a polite follow-up email. Hit all 3 and move on. Don't spend another minute.
When you do land a real loop, the fundamentals haven't changed. Data modeling is still the core skill. SQL is still the lingua franca. The concepts transfer across tools; tool knowledge doesn't transfer across concepts. For sharpening what actually shows up in real panels, we built datadriven.io for data modeling interview questions that reflect what companies like Databricks and Snowflake ask today, not textbook exercises from a decade ago. System design for pipelines, not load balancers. The stuff that trips up engineers who are great at the job but haven't prepped for the game.
Because that's what this is. A game. 48% of the board is fake squares. The real squares take 60 to 90 days to land on. The dice are loaded by design. None of that is fair, and none of it matters. The engineers who get hired in this market are the ones who stop wasting energy on ghosts and start targeting the companies writing real offer letters.
I've been through 3 waves of "data engineering is getting automated away." Still here. Still employed. Still debugging the same categories of problems. The tools change every 18 months. The problems don't. Schema drift, late-arriving data, upstream teams breaking contracts without telling you. These are eternal. The market for people who solve them isn't going anywhere; the process for getting hired to solve them is just temporarily insane.
5 states have active legislation targeting ghost job practices as of May 2026. New York's bill calls them "dishonest and exploitative." Maybe regulation helps eventually. But your job search is happening now, not in 2028 when lawmakers might get around to enforcement.
Grind smart. Verify before you invest. Stop doing free consulting for companies that were never going to hire you.
What's the worst ghost job experience you've been through? I want the real war stories.
Top comments (0)