Tech Hiring Report

Job-search advice usually treats a posting as a standing invitation. It is not. Across the 2,869 postings we watched from first sighting through to close, the median stayed open 12.3 days. A quarter closed within 5.0 days.

The practical consequence: 36.2% of them were gone within a week of us first seeing them — a share of the postings we watched close, not of everything tracked. If you keep a list of roles to apply to "this weekend," roughly a third of that list expires before you get there.

That is only half the picture. 10,599 of the 19,737 postings we tracked were still open when this snapshot was taken, and a posting that is still open has no lifespan yet. So this median is a lower bound, not the typical wait. The long tail is its own story: in our ghost jobs analysis, nearly a third of postings were still open more than two months after we first saw them.

Cumulative share of 2,869 closed postings, measured from first sighting to last verification.

Hybrid work dominates the discussion but not the listings. Only 4.3% of postings were labelled hybrid, against 24.3% fully remote and 71.3% on-site.

One caveat worth stating plainly: this reflects how employers label a role in their applicant tracking system, not how the job is actually worked. A large share of roles tagged on-site are hybrid in practice; employers simply have no incentive to say so in the posting.

The work-model split matters more for timing than for anything else we measured. Among postings that closed, remote ones closed in a median of 7.2 days against 14.0 days for on-site — and nearly half of those remote roles were gone inside a week.

If you are filtering for remote, the practical read is that a remote posting you find on a Friday is meaningfully less likely to still be open on Monday than an on-site one.

Bar length is the share closed within seven days. Hybrid's sample (79 closed postings) is small enough that we would not lean on it.

60.1% of postings included at least one salary figure, and 38.7% published a full range. Across postings that disclosed a band, the median ran from $150,000 to $190,000.

The counterintuitive part is where disclosure happens. Remote roles disclosed pay in only 46.4% of postings, against 64.9% of on-site ones — a gap of 19 points, in the opposite direction from what remote-work transparency norms would predict.

The likeliest explanation is regulatory rather than cultural. US pay-transparency laws attach to a job's location, so an on-site role in a covered state is unambiguously in scope while a "remote" role is easier to leave unlabelled. We can measure the gap; we cannot prove the cause from this data.

Share of postings in each work model that disclosed at least one salary figure.

Python leads on demand, mentioned in 34.4% of postings — roughly 3× the next-most-common technology. But demand and pay are not the same list.

On advertised pay, the spread across the skills we track is about $39,450. Postings mentioning Machine learning advertise a median floor of $175,000, against $135,550 for SQL. We ran the same comparison inside remote and on-site postings separately before publishing this, and the ordering held in both — so it is not simply a reflection of where the work happens, though that check is not itself in the table below.

Remote availability splits even harder, and along a different line. React postings are remote 43.8% of the time against 11.5% for C++ — front-end and web skills cluster at the top, systems and hardware-adjacent ones at the bottom, against a site-wide baseline of 24.3%.

Read this as a description of postings, not as a return on learning a skill. A posting mentioning machine learning is usually a different job from one mentioning SQL, and we cannot separate the skill from the seniority and role type that travel with it. Mention rate is measured over every scanned posting; pay and remote share are measured within the postings that mention the skill. Skills with fewer than 150 such postings are omitted from this table. A posting can mention several skills, so shares do not sum to 100%.

Ranking cities by job count is misleading, because one large employer's campus outranks an entire metro. So we counted distinct companies hiring instead, which is closer to what a job seeker is actually shopping in — and then looked at how many roles each of those employers posts.

That second number is where it gets interesting. Washington DC and Chicago have almost the same number of employers hiring (79 and 82), but Washington DC carries 4× as many open roles. Its market is a handful of large employers hiring in volume; Chicago's is spread thinly across many.

Neither shape is simply better, and it is worth being clear about that: a low postings-per-company number means demand is distributed rather than dependent on a few employers, but it can equally mean the market is small. San Diego has the lowest concentration on this table and also the fewest openings on it. Read the two columns together, never the ratio alone.

Metro grouping is hand-curated over free-text locations (2,586 distinct spellings), so it is approximate and deliberately conservative — city names that are ambiguous carry their state (a bare "New York" matches the state and drags in Buffalo; a bare "Austin" matches Austintown, Ohio), locations we could not confidently place are left out entirely rather than guessed at, and metros below 25 companies are not ranked. A posting open in several cities counts in each of them, which is right for "is anyone hiring near me" but means these rows overlap and do not sum to the total.

Every figure comes from postings we ingested directly from employer applicant tracking systems, not from a third-party aggregator. We record each posting when we first see it and re-check it on a schedule, so "time open" is measured by observation rather than by a stated closing date.

• Sample: 19,737 postings from 1,680 companies, collected June 20, 2026 to September 3, 2026. 10,599 were still open at the time of writing and 9,138 had closed.
• Lifespan: measured on the 2,869 postings we observed from first sighting through to close. A posting still open cannot have its lifespan measured, so the median is a lower bound on the full population.
• What we exclude from lifespan, and why: a posting's lifespan only means something if we watched it from the moment it appeared. 3 days in this window were bulk imports rather than live sightings (2026-06-20, 2026-06-29, 2026-07-04), covering 11,902 postings whose recorded start date is a database event, not an observation. Including them manufactured a false spike at exactly five days and pulled the reported median up by several days. Every lifespan figure on this page excludes those cohorts. Population figures — work model, pay disclosure, skills — are unaffected and use the full sample, because those are attributes of the posting rather than measurements over time.
• Sources: Greenhouse (10,355), Ashby (5,798), Lever (3,584). These are modern ATS platforms, so the sample skews toward technology employers and venture-backed companies, and away from government, agency and enterprise-legacy hiring.
• Skills: a keyword scan over an AI-extracted skill summary for each posting, so figures describe how often a technology is mentioned, not how often it is strictly required.
• Salary: figures as published by the employer, restricted to a $20,000–$900,000 sanity window to exclude feed parse errors. No currency normalisation is applied.
• Refresh: quarterly, at this same URL. Numbers on this page reflect the September 3, 2026 snapshot.

Free to quote, with a link back. If you need one line with the claim, the sample and the window attached:

InterviewChamp.AI tracked 19,737 tech job postings from 1,680 companies between June 20, 2026 and September 3, 2026. Of the 2,869 postings watched from first sighting through to close, the median stayed open 12.3 days and 36.2% closed within a week.

Source URL: https://interviewchamp.ai/tech-hiring-report

Working on a story or a piece that needs a cut we have not published — by title, by region, or over a longer window? Get in touch and we will run it.