Shortlist rate. 3,848 applications produced 20 shortlists and one hire. Of 1,038 applications to hands-on engineering roles, exactly one was shortlisted.
We reviewed 3,848 job applications. We shortlisted twenty.
Over sixteen months, one deep-tech employer posted 34 roles and kept the complete record — every application, every screening decision, every rupee of promotion spend. Publishing it exposes both sides: a candidate market that has not kept up, and employers (ourselves included) who waste most of what they pay for.
Bear Systems · Hiring Research · August 2026 · India
Never reviewed. The four biggest postings drew 1,968 applications and received 34 human verdicts between them. The waste is on the employer's side of the table.
The skills cliff. 41% of early-career engineering resumes name CAD/CAE tooling; 4.5% of 20-year resumes do — controlling for resume length. AI shows the same slope.
The AI ceiling. Even in the youngest, most digital cohort, fewer than three in ten resumes mention any AI/ML capability. The "AI-native generation" did not show up in this pipeline.
In seat under a year. Of currently-employed applicants, 38.5% have held their present job less than twelve months. Among those between jobs, half the just-ended stints lasted under a year.
Cost per applicant. Total promotion spend was $9,520. Volume was never the problem — a tenfold cost spread by role, and quality nowhere at any price.
01 — The funnel
Twenty shortlists from 3,848 applications — and half were never read
The funnel indicts both sides at once. Candidates: of 1,235 applications that received a human verdict, 56% were rejected outright and only 1.6% reached a shortlist. Employer: 2,613 applications — including 96% of everything sent to the four largest postings — received no verdict at all.
Applications received
Given a human verdict
Rated favourably
Shortlisted
Hired
09621,9242,8863,848
"Favourable" combines GOOD_FIT and SHORTLISTED verdicts. One hire resulted — an engineering lead — plus two design-engineer shortlists that progressed. Bar lengths are to scale; that is the point.
| Posting | Apps | Median exp (yrs) | Postgrad | Reviewed | Reviewed % | Shortlisted |
|---|---|---|---|---|---|---|
| Head of Supply Chain Management | 584 | 16.6 | 57% | 10 | 2% | 0 |
| Plant Operations Manager | 549 | 14.5 | 35% | 15 | 3% | 0 |
| Quality Assurance Specialist | 442 | 15.0 | 35% | 3 | <1% | 0 |
| Aerospace Engineer | 393 | 4.4 | 31% | 6 | 2% | 0 |
| Head of Operations (UAV & Tactical) | 358 | 19.4 | 56% | 357 | 100% | 12 |
| IoT & Embedded Systems Engineer | 209 | 3.5 | 22% | 209 | 100% | 0 |
| Business Consultant | 194 | 4.9 | 43% | 46 | 24% | 0 |
| Design Engineer (Drone & CFD) | 178 | 4.0 | 39% | 177 | 99% | 1 |
| UAV Test Pilot | 131 | 4.2 | 12% | 131 | 100% | 0 |
| UAV Technician — Inventory & Procurement | 127 | 4.5 | 24% | 127 | 100% | 0 |
| CTO — UAV & Autonomous Systems | 97 | 17.6 | 52% | 84 | 87% | 7 |
| Business Development (BD) Director | 89 | 21.5 | 61% | 1 | 1% | 0 |
| Service Training Manager | 85 | 11.9 | 45% | 1 | 1% | 0 |
| Business Associate | 73 | 4.5 | 32% | 4 | 5% | 0 |
| Shopify Store Manager | 59 | 6.0 | 44% | 0 | 0% | 0 |
| Business Associate II | 48 | 6.1 | 56% | 0 | 0% | 0 |
| Project Team Leader / Survey Director | 46 | 15.3 | 45% | 3 | 7% | 0 |
| Lead GIS, Remote Sensing & Hydrology | 44 | 5.8 | 79% | 1 | 2% | 0 |
| Technology Officer / Engineering Staff | 36 | 5.7 | 36% | 0 | 0% | 0 |
| Senior Environmental Expert (EIA) | 32 | 5.1 | 71% | 1 | 3% | 0 |
| B2B Sales (Geospatial Analytics) | 26 | 3.6 | 46% | 26 | 100% | 0 |
| BD Executive — Defense & GovTech | 25 | 6.8 | 44% | 25 | 100% | 0 |
Median experience is computed from dated employment history, not self-description. Rows in bold are the postings where fewer than 5% of applications were ever reviewed. Postings under 25 applications are omitted (they add 49 applications and 8 reviews).
So what
An employer that pays to advertise 584-applicant postings and reviews ten of them has not run out of candidates; it has run out of process. Before concluding "there is no talent," check what fraction of your own funnel anyone actually read. Ours was 32%.
02 — The skills vacuum
Experience and modern tooling move in opposite directions
We scanned all 3,714 machine-readable resumes for five families of technical capability. The pattern is monotonic: every additional band of experience makes a resume less likely to mention the tools the roles actually require.
The obvious objection — senior resumes are terser — is wrong in this corpus, and wrong in the reassuring direction. Twenty-year resumes are 2.6× longerthan early-career ones (median 8,434 characters against 3,234). They have more room and still say less. Normalised per 10,000 characters, CAD/CAE mentions fall from 7.7 to 0.14 — a 55-fold decline; restricted to resumes of similar length, the cliff is unchanged.
| Capability | 0–3 yrs | 3–7 | 7–12 | 12–20 | 20+ yrs |
|---|---|---|---|---|---|
| AI / machine learning | 20.5% | 15.6% | 10.6% | 9.4% | 10.5% |
| CAD / CAE / simulation | 39.3% | 39.7% | 25.7% | 10.9% | 7.4% |
| Embedded / firmware | 24.8% | 17.5% | 9.4% | 5.4% | 6.7% |
| Programming languages | 52.4% | 43.1% | 27.4% | 18.0% | 10.5% |
| UAV / drone specific | 26.7% | 24.6% | 11.0% | 5.2% | 6.4% |
Cell shading is proportional to the rate. Band sizes: 420 / 789 / 756 / 884 / 703. A mention is evidence of self-presentation, not tested competence — but a senior engineer who does not think CAD or Python worth naming is telling you how they expect to work.
The uncomfortable half
The story is not "juniors are AI-native." Nobody is. In the youngest cohort, 79.5% of resumes show no AI/ML exposure at all — against ~90% of senior ones. The juniors are not modern; they are less obsolete, and cheaper. For a country supplying 1.5 million engineering graduates a year into a global AI build-out, a one-in-five mention rate at the entry level is the single most alarming number in this report.
So what
For candidates this is the cheapest arbitrage on the page: demonstrated AI capability — a repo, a deployed model, a documented workflow — currently puts you ahead of ~80% of your own cohort and ~90% of the seniors you will compete against. For employers: the senior hire you are waiting for, pre-trained on modern tooling, statistically does not exist. Budget to upskill or to grow one.
03 — The self-assessment gap
People do not know what level they are
Two symmetrical failures. One hundred applicants with under five years of experience applied for Head-of, Director or Chief-titled roles. Meanwhile 581 applications to hands-on individual-contributor roles — 24% of that pool — came from people with fifteen-plus years.
Among the 238 people who applied to more than one of our postings,27% applied to both a senior-titled and an IC-titled role — the same person presenting as director and as junior engineer in the same pipeline. Title inflation reads as candidate delusion; it is equally evidence that the market gives people no reliable way to price themselves.
Senior-titled postings (n = 1,237) IC-titled postings (n = 2,435)
Under 5 yrs
5–10 yrs
10–15 yrs
15+ yrs
0%15%30%45%60%
"Senior-titled" = Head of / Director / Chief / Lead / Principal / Senior in the posting title. Percentages within each posting group.
So what
A quarter of the senior-role pool cannot clear its own bar, and a quarter of the junior pool is over-qualified people bidding down. Both are the same signal: applicants are pricing themselves by hope, not by market feedback — because rejection, at a 0.5% shortlist rate, carries no information.
04 — Tenure
The eighteen-month employee is now the norm
Median time in the current seat: 1.33 years. Median length of the just-ended job among between-jobs applicants: 0.92 years. These are not outliers churning at the margin; they are the centre of the distribution.
Current role, ongoing (n = 2,838) Last role, ended (n = 803)
Under 1 yr
1–2 yrs
2–3 yrs
3–5 yrs
5–10 yrs
10+ yrs
0%15%30%45%60%
Consecutive roles at the same employer are merged before measuring, so internal promotions do not inflate the churn. 74% of applicants applied while employed.
Read it cynically in both directions, because both are true. Candidates treat employers as eighteen-month stepping stones and would call it ambition. Employers built the stepping stones: if half your leavers exit inside a year, that is a management and progression failure before it is a loyalty failure. "No one stays anymore" and "no one gives anyone a reason to stay" are the same statistic, attributed to whichever side is speaking.
Selection warning
This is a pool of active job-seekers. It over-represents movers by construction and says nothing about the tenure of people who are not applying anywhere. It measures who is in the market — which, for a hiring employer, is the population that matters.
05 — Who is in the market
Credentialled, metropolitan, thirty-two, and interchangeable
87.3% of applicants hold a bachelor's degree or higher; 42% hold a postgraduate degree. Degrees are table stakes — which is precisely why they discriminate nothing. The market's response to credential saturation has been more credentials.
Bachelor's
Master's
Doctorate
Diploma
Other / school
None listed
06001,2001,800
Under 25
25–29
30–34
35–39
40–44
45–49
50+
02505007501,000
Age is not collected in applications and is estimated from career start; a second, independent estimate from bachelor's start year agrees to a median gap of one year. Estimated means estimated — this exhibit describes the pool, and played no part in any screening decision.
Tier-1 metro
Tier-3 and below
Tier-2 city
Unspecified
Outside India
06001,2001,8002,400
Tier-1: Hyderabad, Delhi NCR, Bengaluru, Mumbai, Chennai, Pune, Kolkata. Locations are bucketed so no small group is identifiable.
Where you stand — for the candidate reading this
Median applicant: 32 years old (estimated), bachelor's-plus, tier-1 metro, 1.3 years into the current job, no AI on the resume. If that describes you, you are the median — and the median, in this pipeline, was rejected. Every deviation that matters is listed in section 02.
06 — Cost
Applicants are nearly free. That is exactly the problem.
$9,520 of promotion bought 3,848 applications — a blended $2.47 each, ranging from $0.64 for a plant-operations manager to $6.32 for a UAV test pilot. When applying costs the candidate one click and sourcing costs the employer under a dollar, volume is guaranteed and signal is not.
UAV Test Pilot
Service Training Manager
UAV Technician
CTO — UAV & Autonomous
Business Associate
BD Director
Design Engineer
Aerospace Engineer
Head of Operations
Business Consultant
Head of Supply Chain
Quality Assurance Specialist
IoT & Embedded Engineer
Plant Operations Manager
$0$1$2$3$4$5$6$7
Promotion spend divided by applications received, aggregated where a role was posted more than once. Secondary figure: applicants that spend produced.
Spend does not scale
We posted Head of Operations three times: at $61,$57 and $372. The $372 run drew fewer applicants than either cheap one — 110 against 136 and 112. The audience for a role is finite; the platform will happily charge you to saturate it twice.
07 — Our own screening, audited
The reviewers were not consistent either
Honesty cuts inward too. Favourable-rating rates ranged from 0% to 98% across postings — reviewer mood and standards, not hundredfold swings in candidate quality. And the strongest correlate of a favourable verdict was not skill or education. It was the applicant's city.
Applicant in tier-1 city Applicant elsewhere
IoT & Embedded Engineer
36%15%
UAV Test Pilot
46%30%
Head of Operations
35%22%
Design Engineer
24%13%
UAV Technician
38%
CTO — UAV & Autonomous
15%21%
0%10%20%30%40%50%
Postings with 10+ rated applicants in both groups. All roles were Delhi NCR / Gurugram based; proximity plausibly proxies availability and logistics. Compare within rows only — the 0–98% reviewer variance makes cross-posting comparison meaningless.
Observations we cannot measure — recorded as testimony, not data
Two patterns recurred in our screening that the structured data cannot verify, and we flag them as exactly that. First: senior applicants' compensation expectations were consistently far above what their demonstrated (as opposed to claimed) capabilities supported — while junior applicants with more current tooling asked for a fraction of the figure. Second: application communication quality — clarity of written English, coherence of the pitch — was strikingly poor across the youngest cohort. Our compensation field had 67 usable responses in mixed units, so neither claim is chartable. Treat both as one employer's sworn account, and weigh accordingly.
08 — What to do about it
Prescriptions, if anyone wants them
For employers
- Review what you buy. We paid for 3,848 applications and read 32%. Whatever your funnel software promises, unread applications are burned spend and burned goodwill.
- Stop waiting for the pre-trained senior. At a ~10% AI mention rate among 15-year veterans, the modern-toolchain senior hire is a statistical rounding error. Hire the trajectory, budget the training.
- Treat 18-month tenure as your baseline case and design onboarding, knowledge capture and progression around it — or fix the reasons people leave, which your exit data already names.
- Cap posting spend early. Applicant volume saturates; our 6× spend increase bought fewer applicants. Spend the difference on reviewing.
For candidates
- Demonstrated AI beats claimed everything. A public repo or deployed workflow puts you ahead of ~80% of your cohort. It is the only cheap differentiator left on this page.
- Name your tools. Reviewers scanning 300 resumes do not infer your CAD stack from a job title. Seniors: your longer resume that lists no tooling is read as evidence, not modesty.
- Apply at your level. Sub-5-years-to-Director applications were 8% of the senior pool and converted at zero. They cost you nothing and earn you nothing.
- Know the denominator. A 0.5% shortlist rate means rejection is the default outcome of a click-to-apply market — not a verdict on you. Direct evidence of capability is how you exit the denominator.
09 — Method & limits
How this was built, and what it cannot say
Source: complete structured application records for 34 postings by one employer, April 2025 – July 2026, with dated employment and education histories, plus machine-extracted text from 3,714 resumes. Tenure, experience and age proxies are computed from dates, not self-description. Skill rates are keyword-family scans of resume text, cross-checked against resume length two independent ways.
Disclosure controls
Aggregates only. No individual is named or identifiable; no employer of any applicant is named; groups under ten people are suppressed or merged; locations are bucketed to city tiers; qualifications to broad levels. The underlying records stay on our machines.
Limits
- One employer, one platform, sixteen months. This describes our applicant pool, not the Indian labour market; roles skew deep-tech and operations.
- A skill mention is self-presentation, not tested competence — the exhibit measures how people present, which is itself the finding.
- Age is estimated from career/education start (two methods, median disagreement one year) and is reported for pool description only.
- Ratings cover 32% of applications, with 0–98% between-posting variance; every rating-based claim is confined to within-posting comparison.
- Compensation was not reliably captured (67 usable responses, mixed units): no salary claims are made from data anywhere in this report.
- Tenure describes active job-seekers, who move more than incumbents.