How to Get Hired in Tech.
The resume + interview guide that beats the funnel.

Hard, sourced funnel numbers for FAANG specifically are scarcer than the internet suggests. Most percentages floating around Reddit are uncited. What we can verify paints a clear picture of brutal selectivity at the top of the funnel and surprisingly noisy decisions at every later stage.

Aline Lerner's analysis of ~300 candidates at TrialPay (a unicorn-tier company at the time) found roughly 1 in 50 interviewed candidates received an offer, and 1 in 500 applicants overall. That's a ~2% offer rate among people who made it past the resume screen, and ~0.2% end-to-end.

At the resume-review stage, interviewing.io's controlled study found recruiters picked the stronger candidate correctly only 55% of the time — barely better than a coin flip. Two recruiters reviewing the same resume disagreed by an average of 41 percentage points. Translation: a meaningful chunk of 'rejected at recruiter screen' is noise, not signal.

Almost every large employer you'd want to work for runs your resume through an Applicant Tracking System before a human sees it. Jobscan's 2025 ATS Usage Report found 97.8% of Fortune 500 companies use a detectable ATS (489 of 500).

Vendor breakdown matters because each ATS parses resumes differently. Fortune 500 leans on Workday (~39%) and SuccessFactors (~13%). Tech / unicorns / scale-ups skew toward Greenhouse (~19%), Lever (~17%), Workday (~16%), and iCIMS (~15%). Early-stage startups trend toward Greenhouse, Lever, or Ashby.

The widely repeated '75% of resumes are auto-rejected' figure is not well-sourced. What Jobscan does document: more than 90% of employers filter or rank candidates pre-recruiter, and 76.4% of recruiters search and rank by skills pulled verbatim from the job description. Practical implication: skills that mirror the JD measurably increase your odds of being surfaced.

The often-cited '6 seconds' figure traces to a 2012 Ladders eye-tracking study that's hard to retrieve today. The more rigorous recent data point is interviewing.io's controlled resume-evaluation study: median resume evaluation took 31 seconds, and spending an additional 15 seconds increased accuracy by ~34%.

What are they looking at? Despite recruiters stating 'missing skills' as their main rejection reason, the actual drivers were: a recognizable brand-name employer (+35% interview rate), and clarity of contribution descriptions — bullets that say what you built and why it mattered, not jargon stacks.

Practical takeaway: in the first scan, recruiters are pattern-matching on logos and a clear, results-oriented top third of the resume. The most readable predictor of an offer in Lerner's data wasn't pedigree — it was the absence of typos (87% of offer recipients had ≤2 mistakes).

Software engineering compensation at scale is a step function, not a slope — and the step you land on at offer time often persists for years. A single level difference at FAANG can mean $50K–$150K+ in total comp annually depending on stock refresh and band.

Interview difficulty also scales: an L4-targeted loop and an L5-targeted loop at Google use the same question types but evaluate scope, ambiguity-handling, and design depth very differently. Get the level question right before you walk in — recruiters can and do downlevel during debrief, and that's the single largest comp lever most candidates never explicitly negotiate.

Different companies want different behavioral structures. Amazon is the most explicit: their interview-prep page directs candidates to study the 16 Leadership Principles and answer behavioral questions using the STAR method. Every behavioral interviewer in an Amazon loop is assigned 1-2 LPs to evaluate; your stories need to map to specific LPs.

Google uses structured behavioral interviewing across the board — same question set, calibrated rubrics — though no single named framework. STAR works well; Google rewards concrete impact (numbers, scope, ambiguity).

Meta, Stripe, and most unicorns accept STAR or CAR interchangeably. Early-stage startups often skip a formal behavioral round and substitute a founder culture-fit chat.

Practical rule: prepare 8–12 STAR stories that each hit 2–3 different competencies (conflict, ambiguity, technical depth, mentorship, prioritization). At Amazon specifically, label each story with the LPs it demonstrates before you walk in.

Resume screen: the loudest stated reason is 'missing skills,' but the loudest actual drivers are the absence of a brand-name employer (−35% interview rate) and visible signal noise like typos and jargon. Recruiter disagreement on the same resume averages 41 percentage points — meaning a non-trivial share of rejections is reviewer-specific.

Recruiter screen: most common kill is a mismatch between stated comp expectations and the level/band the recruiter has open. Second-most common is communication clarity — recruiters explicitly screen for whether you can describe your last project in 90 seconds.

Online assessment: the dominant failure mode is under-practice volume, not lack of CS fundamentals — candidates who hadn't approached ~500 problems showed measurably lower interview scores.

Technical phone screen: the top failure mode is not communicating while coding — interviewers cannot assess silent candidates and default to lower scores.

Onsite loop: uneven performance — strong in 3 interviews, weak in 1-2 — combined with weak behavioral/LP answers. At Amazon specifically, LP performance is the explicit tiebreaker on borderline technical candidates.

Debrief / committee: calibration — does the panel agree on level? — kills more offers at FAANG than raw technical performance. A unanimous-pass at L3 but split at L4 typically becomes a downlevel or no-hire.

Reliable, specific stats on tech-candidate negotiation rates are surprisingly hard to source. What is well-grounded: the financial leverage is enormous. Patrick McKenzie's classic essay calculates that a one-time $5,000 salary increase compounds to roughly $100,000 of gross lifetime value over a decade.

Companies expect negotiation and are not punished for it. Haseeb Qureshi documents that companies spend on the order of $24,000+ to recruit a single candidate — they're motivated to close, not to punish. His core rules: have a real BATNA, never give the first number, give reasons for every ask, and stay 'winnable' so the recruiter wants to fight for you internally.

What to negotiate beyond base: sign-on, RSU grant, refresh cadence, level, start date, remote flexibility. At FAANG specifically, level is usually the biggest dollar-per-effort lever. Nothing in the negotiation literature suggests asking is risky in tech — rescinded-for-asking is essentially never reported in the published guides.