Skip to content
Analyst work

Data study · October 2026

Handshake and LinkedIn

Question

How can Handshake win new-grad hiring against LinkedIn and Indeed?

Headline finding

Recent graduates were out of work at 5.7% in June 2026, against 4.1% for all workers. Employers on Ashby’s hiring platform now sift 291 applications per hire. In NACE’s spring survey 43.4% of employers had detected AI-written applications and 20.1% used a tool to detect AI help.

My recommendation

Sell employers a verified applicant channel first. Then add alumni referrals and proof of skill from Handshake AI work on that verified base.

Tools
  • CSV tables with a source URL on every row
  • JavaScript on Node for the derived rates
  • D3.js for the first dashboard
  • React and SVG for the charts on this page

The question

The client

Handshake’s leadership. More than 1,600 schools run their career services on Handshake. By press reports, AI training made up 87% of its gross annualized revenue in April 2026. In June 2026 it opened sign-up to any email address and let search engines index student profiles.

The decision they face

Should Handshake keep opening up into a general network like LinkedIn, or build its recruiting product on the school-verified students it already has?

My starting hunch

I came in with the calls from two earlier studies of mine. The LinkedIn Premium study picked Premium Career as the job seeker entry tier, and the Indeed study backed AI hiring software.

How the hunch held up

The first held in part. LinkedIn’s job match cut low-match applications by 10% for Premium users, so a paid signal does sort the pile. New grads complicate it, because Handshake is free to students and tied to their school, and LinkedIn’s own entry-level hiring fell 6%. The second got harder to hold. 70% of hiring managers say AI speeds up good decisions, yet 41% of job seekers admit to prompt injections and 25% of recruiters have no confidence in their AI systems. Screening software works better when the applicant behind the file is verified.

The data, 4 tables

  • nyfed_unemployment.csv184 rows

    Unemployment rates for recent graduates, all workers, all college graduates and young workers at quarter-end months from March 2015 to June 2026.

    Where it came from
    The New York Fed’s Labor Market for Recent College Graduates data file, which it builds from the Current Population Survey.
    Why I chose it
    It is the study’s one long time series, and it puts recent graduates beside the whole workforce on one definition.
    What it cannot tell us
    Anything about a platform. A recent graduate here is aged 22 to 27 with a bachelor’s degree or more, and small samples make that line jumpy from one reading to the next. The NY Fed also revises the series, and the table holds the values as pulled on October 31, 2026.
  • metrics.csv135 rows

    Every figure that appears once, covering new-grad hiring, application volume, trust in hiring and each platform’s business results.

    Where it came from
    NACE, the NY Fed, BLS, Indeed Hiring Lab, LinkedIn, Handshake, Ashby, Greenhouse and press reports.
    Why I chose it
    Each figure gets an ID, so the driver matrix and the code can point at the exact row.
    What it cannot tell us
    A like-for-like comparison across platforms. The rows come from separate surveys with separate bases, and 32 of the 135 come from a platform reporting on itself. Handshake is private, so its revenue rows are press reports and Sacra estimates.
  • events.csv8 rows

    Dated Handshake moves from the launch of its AI line in January 2025 to September 2026.

    Where it came from
    Handshake product updates and press coverage.
    Why I chose it
    It puts dates on the turn toward AI training and on the change to open sign-up.
    What it cannot tell us
    Why Handshake made each move. 4 of the 8 rows are outside reports about a private company.
  • drivers.csv8 rows

    A matrix of 8 drivers of new-grad hiring, each scored 0 to 2 for Handshake and for its two rivals.

    Where it came from
    My own coding, with the IDs of the rows each score rests on.
    Why I chose it
    It shows where Handshake leads LinkedIn and where it trails, and each option answers one of those gaps.
    What it cannot tell us
    Stand as measurement. The scores are my judgment, and the row IDs are there so a reader can disagree. This table carries row IDs where the other three carry a URL.

The process

  1. One row per figure

    Each figure went into a row of its own with a source name, a URL, the date I retrieved it and a flag for whether a platform reported it about itself. metrics.csv holds 135 of them from 30 named sources.

  2. Quarter-end months

    The NY Fed publishes its rates monthly. The table keeps the quarter-end months, which gives 46 readings for each of the four series.

  3. Pivot

    nyfed_unemployment.csv is long, with one row per series per month. I pivoted it to one row per month and subtracted the all-worker rate from the recent-graduate rate.

  4. Ratios by row ID

    Every ratio is computed from named rows. Growth in applications per hire divides row AP06 by row AP04, and the AI share of Handshake’s revenue divides BZ05 by BZ06.

  5. Assumptions

    Rounded statements enter as the number stated, and the notes column keeps the wording. Row BZ05 reads “Reported as nearly $1B” and enters as 1,000 million dollars. Row BZ01 reads “Midpoint of reported $5M-$10M”. Where a source gave two values I kept one and logged the other, as in row MK10, “Release body says 43%, page summary says 44%”.

  6. Check

    The 36.5x offer multiple divides 7.3% by 0.2%, and the second number is printed to one decimal. I rebuilt both offer rates from the stage rates, application to interview times interview to offer. The stage rates are 2021 to 2024 averages and give 6.4% and 0.18%, a multiple of 35.6x. Rounding alone lets the reported multiple sit anywhere from 29.2x to 48.7x. All 24 row IDs cited in the driver matrix resolve to a row.

One month before and after the pivot.

Before, 4 rows of nyfed_unemployment.csv (5 of 10 columns shown)
metric_id,series,value,unit,period
NYF-young_workers-2026-06,young_workers,7.188,percent,2026-06
NYF-all_workers-2026-06,all_workers,4.061,percent,2026-06
NYF-recent_grads-2026-06,recent_grads,5.698,percent,2026-06
NYF-college_grads-2026-06,college_grads,2.859,percent,2026-06
After, one row per month
{
  "period": "2026-06",
  "recent_grads": 5.698,
  "all_workers": 4.061,
  "college_grads": 2.859,
  "young_workers": 7.188,
  "gap": 1.637
}

Each value is a three-month average of a survey estimate. The gap is the recent-graduate rate minus the all-worker rate, in percentage points.

The code for the key steps

The pivot from long rows to one row per month, with the gap. derive.mjs
export function nyfWide(nyfed) {
  const by = {};
  nyfed.forEach((r) => {
    by[r.period] = by[r.period] || { period: r.period };
    by[r.period][r.series] = Number(r.value);
  });
  return Object.keys(by)
    .sort()
    .map((p) => ({ ...by[p], gap: by[p].recent_grads - by[p].all_workers }));
}
The summary of that gap, before 2020 and since 2022. derive.mjs
export function nyf(wide) {
  const mean = (a) => a.reduce((s, x) => s + x, 0) / a.length;
  const last = wide[wide.length - 1];
  const pre = wide.filter((d) => d.period >= "2015-01" && d.period <= "2019-12");
  const since = wide.filter((d) => d.period >= "2022-01");
  return {
    period: last.period,
    recent: last.recent_grads,
    all: last.all_workers,
    gap: last.gap,
    pre_below: -mean(pre.map((d) => d.gap)),
    pre_readings: pre.length,
    now_gap: mean(wide.filter((d) => d.period >= "2025-01").map((d) => d.gap)),
    quarters_above: since.filter((d) => d.gap > 0).length,
    quarters_total: since.length,
  };
}
The check on the offer multiple. derive.mjs
export function offerCheck(metrics) {
  const v = pick(metrics);
  const referred = (v("CH05") * v("CH07")) / 100;
  const inbound = (v("CH06") * v("CH08")) / 100;
  const half = 0.05;
  return {
    referred_stage_pct: referred,
    inbound_stage_pct: inbound,
    stage_multiple: referred / inbound,
    reported_multiple: v("CH23") / v("CH24"),
    multiple_low: v("CH23") / (v("CH24") + half),
    multiple_high: v("CH23") / (v("CH24") - half),
  };
}

Dead ends

  • Source of hire by platform

    No national survey tracks where new graduates get hired by platform. The closest I could get pairs Ashby’s conversion rates across all roles with one college’s first-destination survey, and each has a gap.

  • A current table of recruiting methods

    The full table of methods employers use and rate comes from the NACE 2019 Internship & Co-op Survey, the latest edition with open figures. Later NACE reports confirm the ranking in words only.

The 3 findings that change the decision

  1. Finding 1

    1.6 points

    Recent graduates 5.7% against all workers 4.1%, June 2026

    Recent graduates have been out of work at a higher rate than all workers in all 18 quarter-end readings since 2022. From 2015 to 2019 they averaged 0.3 points below.

    Dataset nyfed_unemployment.csv
    Query derive.mjs, nyf()

    Recent graduates now sit 1.6 points above all workers

    Unemployment rate at quarter-end months, as three-month averages. Recent graduates are aged 22 to 27 with a bachelor’s degree or more, and all workers are aged 16 to 65. Hover, tap or use the arrow keys for each reading.

    • All workers
    • All college graduates
    • Recent graduates

    Source: nyfed_unemployment.csv, derive.mjs, nyfWide(), derive.mjs, nyf().

    Show the numbers as a table
    Quarter-end monthAll workersAll college graduatesRecent graduates
    Mar 20155.42.85.1
    Jun 20155.32.74.4
    Sep 20155.02.64.6
    Dec 20155.02.64.4
    Mar 20164.82.63.7
    Jun 20164.82.64.0
    Sep 20164.72.64.1
    Dec 20164.72.63.8
    Mar 20174.42.54.2
    Jun 20174.22.54.1
    Sep 20174.22.43.8
    Dec 20174.02.23.5
    Mar 20183.92.33.8
    Jun 20183.82.23.6
    Sep 20183.62.13.6
    Dec 20183.72.34.1
    Mar 20193.72.23.6
    Jun 20193.52.24.3
    Sep 20193.62.34.0
    Dec 20193.52.13.6
    Mar 20203.72.23.9
    Jun 202012.88.113.4
    Sep 20208.45.39.0
    Dec 20206.74.36.7
    Mar 20216.13.96.5
    Jun 20216.03.66.2
    Sep 20215.02.95.2
    Dec 20214.22.44.7
    Mar 20223.82.23.9
    Jun 20223.52.14.2
    Sep 20223.42.04.0
    Dec 20223.52.14.1
    Mar 20233.52.14.0
    Jun 20233.52.24.3
    Sep 20233.62.34.4
    Dec 20233.72.34.9
    Mar 20243.72.24.5
    Jun 20243.82.44.7
    Sep 20244.02.65.3
    Dec 20244.02.64.8
    Mar 20254.02.65.7
    Jun 20254.02.74.9
    Sep 20254.12.95.8
    Dec 20254.33.15.6
    Mar 20264.23.15.6
    Jun 20264.12.95.7
  2. Finding 2

    43.4%

    Employers who have caught AI-written applications, against 20.1% with a tool to screen for AI help

    43.4% of employers have detected AI-written applications and 20.1% use a tool to detect AI help, while applications per hire have risen 140% since early 2021.

    Dataset metrics.csv
    Query derive.mjs, ratios()

    43.4% of employers have detected AI-written applications and 20.1% have a detection tool

    Share of employers in NACE’s spring survey, with 159 respondents to each question. The detection-tool question covers the whole hiring process, so it is wider than applications alone.

    Share of employers (%)

    Source: metrics.csv, derive.mjs, ratios().

    Show the numbers as a table
    Employers thatShare of employers (%)SurveyFieldedRow
    Have detected AI-written applications43.4%NACE Job Outlook 2026 Spring Update2026-02/2026-03AP13
    Are unsure whether they received any35.2%NACE Job Outlook 2026 Spring Update2026-02/2026-03AP15
    Use a tool to detect AI help20.1%NACE Job Outlook 2026 Spring Update2026-02/2026-03AP14
  3. Finding 3

    36.5x

    Offer rate per application, 7.3% referred against 0.2% for people who applied on their own

    Referrals are about 1% of applications and about 18% of hires, and a referred applicant gets an offer 36.5x as often as someone who applied on their own.

    Dataset metrics.csv
    Query derive.mjs, ratios()

    Referred applicants get an offer 36.5x as often as those who apply on their own

    Offers per 100 applications across Ashby customers, all roles, Q4 2024. The stage rates in the detail are 2021 to 2024 averages. The rate for people who applied on their own is printed to one decimal, so the multiple could sit between 29.2x and 48.7x. Rebuilt from the stage rates it is 35.6x.

    Offer rate per application (%)

    Source: metrics.csv, derive.mjs, ratios(), derive.mjs, offerCheck().

    Show the numbers as a table
    Applicant sourceOffer rate per application (%)Application to interviewInterview to offer
    Referred applicants7.3%40%16%
    Inbound applicants0.2%3%6%

My recommendation

Sell employers verified applicants first, then build referrals and proof of skill on top.

  1. A

    Sell a verified applicant channel

    Require a school link before any job application, which reverses the 2026 sign-up change for that one step. Cap applications per student each week, and charge employers for a premium tier built on those applicants.

    Evidence
    Finding 2. 43.4% of employers have detected AI-written applications and 20.1% have a tool for it, and Handshake drew 26% more applications per job in a year. On my 0-to-2 scale, Handshake scores 2 on verified identity and 0 on limits to application volume.
    Main risk
    It reverses part of the 2026 sign-up change, and the case rests on one NACE survey with 159 respondents. Nothing in the data tests whether employers would pay for verified applicants.
  2. B

    Build referrals into the student graph

    Match students with alumni at hiring companies and let those alumni refer them inside Handshake.

    Evidence
    Finding 3. Referred applicants get offers 36.5x as often as people who applied on their own, and 44% of Gen Z name a missing network as their biggest barrier. On my 0-to-2 scale for referral networks, Handshake scores 0 and LinkedIn scores 2.
    Main risk
    The Ashby rates cover experienced hires too. Referral rates for new grads are likely lower, since few have a network yet.
  3. C

    Turn Handshake AI work into proof of skill

    Let students publish verified work from Handshake AI projects and AI Skills Studio missions to employer profiles.

    Evidence
    Postings that mention AI nearly doubled to 4.2% of full-time roles, and Class of 2026 resumes mention AI skills more than 9x as often as Class of 2022 resumes. It ties the business behind 87% of gross revenue back to recruiting.
    Main risk
    Both figures in the case are Handshake’s own, and nothing in the data shows an employer hiring on AI project work.
  • School-verified students go first because they’re the one thing LinkedIn and Indeed would find hardest to copy.
  • Employers need this now. 43.4% have caught AI-written applications, and only 20.1% have a tool to screen for them.
  • Referrals and proof of skill both build on those verified applicants.

Limits and next steps

What the data cannot support

  • No dataset tracks source of hire for new grads nationally. The referral rates cover all roles, and the one direct reading is a single college where Handshake accounted for 10% of post-grad jobs.
  • Handshake publishes no financials. Its revenue rows are press reports and Sacra estimates, and the gross figures include contractor pay.
  • 32 of the 135 figures in metrics.csv come from a platform reporting on itself.
  • The sources define a new grad differently. The NY Fed counts bachelor’s graduates aged 22 to 27. Handshake counts early-talent roles, and Indeed’s entry-level class includes hourly work.
  • The driver scores are my judgment on a 0 to 2 scale.
  • The NY Fed revises its series. The table holds my October 31, 2026 pull, and the live file has since moved by enough to change the first decimal of the latest recent-graduate rate.

What I would test with more access

  • With Handshake’s own data I would compare offer rates for school-verified applicants against applicants who signed up with any email.
  • I would pull Google Trends search interest for each platform.
  • I would build an independent recent-graduate series from the BLS Current Population Survey microdata.

TL;DR

  1. Studied

    How Handshake can win new-grad hiring against LinkedIn and Indeed.

  2. Found

    In June 2026, 5.7% of recent graduates were out of work, against 4.1% of all workers.

  3. By the numbers

    Recent graduates over all workers
    1.6 points
    Applications per hire on Ashby
    291
    Offer rate, referred against applied on their own
    36.5x
  4. The catch

    No dataset tracks source of hire for new grads nationally. The referral rates cover all roles.

  5. Next

    With Handshake’s own data, compare offer rates for school-verified applicants against sign-ups with any email.