How to attract and select graduate talent : an employer's guide

How to attract and select graduate talent

Attracting graduate talent is no longer a volume problem. On Handshake, the largest American early-career platform, the average number of applications per job rose 26% in the year to August 2025 while the number of jobs posted fell by more than 16% (Handshake, 2026 Workforce Outlook). Almost nobody is short of applicants. What has become difficult is telling which of them can do the job, and the filters that used to answer that question have quietly been withdrawn.

So the work has moved. Attraction is now about precision rather than reach, and the decisive part of graduate recruitment has shifted to selection.

The numbers you are actually working with

The National Association of Colleges and Employers benchmarks its employer members each year on hiring, pay and method. Read alongside the platform data, the picture is this:

• Applications per job up 26% in a year on Handshake, against a fall of more than 16% in the number of jobs posted.
• Employers expect to hire 5.6% more from the Class of 2026, having projected a 2.4% cut in the autumn. The gain sits with organisations of more than 5,000 employees, which are up 8.7%.
• The projected average starting salary for a business degree is $68,873, up 5.5% on the previous class, a real increase.
• Seventy-nine per cent of entry-level jobs still require a bachelor's degree, and finance, accounting and business administration sit among the five most in-demand majors.

The longer trend runs through the platform rather than the survey. Postings for entry-level roles have fallen across categories supposedly exposed to automation and categories that are not, at broadly similar rates: technology roles down about 15% in a year, healthcare roles down about 12% (Handshake, 2026 Workforce Outlook). Whatever is shrinking the entry-level market, it is not doing it selectively.

The change that matters most for selection is quieter. Seventy per cent of employers now use skills-based hiring, up from 65% a year earlier, and 71% of them use it at least half the time (NACE, Job Outlook 2026 Spring Update). When employers choose between candidates for an internship, the top three factors are the key skills the position requires, a previous internship with the company, and a previous internship in the industry. Major and grade point average sit below all three (NACE, 2025 Internship & Co-op Report). The credential filter that used to sort the pile has largely gone, and for most employers nothing systematic has taken its place.

Why more applications have not helped

Forty-four per cent of employers say they have detected AI-generated applications, and a further 35% do not know whether they have. Only 20% use any tool to look, and 64% use none at all (NACE, Job Outlook 2026 Spring Update, 159 respondents). The uncertainty is the finding: a third of the market cannot say whether the documents it is reading were written by the candidate.

The early stages have therefore stopped measuring the candidate. A written application, a take-home exercise and an unsupervised online test now largely record which tools someone had to hand.

The instinct here is to filter harder and faster, usually by automating the sift. That is the wrong end of the problem, and the evidence on what predicts performance is worth reading before you rebuild a process around it.

Attract fewer, better

If 140 applications is the baseline, broadcasting more widely adds noise that you then pay someone to remove. Worth asking instead: which channels deliver candidates you can already tell something about?

Be accurate about what this cohort responds to. Salary has returned to first place among their professional priorities, cited by 74% of respondents and up 17 points on the previous year, and 79% want to be trained to use AI at work (SKEMA x EY Talent Barometer 2026, conducted by Ipsos BVA among 1,609 students and recent graduates, February 2026). American graduating seniors put career growth and advancement first and a high starting salary in the same top five (NACE, 2026 Student Survey). A campaign built on culture and purpose that avoids the pay question is answering a question this audience is not asking first.

Channels where you meet people before the paperwork do two jobs at once, because they attract and they qualify. Campus events are the clearest case: SKEMA runs career fairs and sector days across its campuses, including Finance Day, Audit & Consulting Day and Luxury Day, where the audience is already filtered by discipline before anyone arrives.

What actually predicts graduate performance

Most graduate processes are built on evidence that has since been corrected. For twenty years, selection design rested on Schmidt and Hunter's 1998 meta-analysis, which put general cognitive ability at the top of the table with a validity of .51 and treated the structured interview as a useful supplement. That ranking has been substantially revised. Sackett, Zhang, Berry and Lievens showed in 2022 that the standard corrections for range restriction had systematically overstated validity across the field, and re-estimated cognitive ability tests at .31 (Journal of Applied Psychology, "Revisiting meta-analytic estimates of validity in personnel selection"). The structured interview now stands at least comparable to the aptitude test, rather than behind it.

Two cautions. The corrected figures are still debated, so nobody should be building a league table of methods to two decimal places. And the finding is not that testing is worthless, only that the whole field was measured with a thumb on the scale.

What survives the revision is practical enough:

• Structure the interview. The same questions in the same order, a defined grid, and more than one assessor. The gain comes from the structure, not from the interviewer's experience.
• The unstructured interview, the free-ranging conversation that most hiring managers believe they are good at, remains the weakest common method.
• Ask candidates to account for work they have actually done, in detail, rather than to describe what they would do. It is a read that a generated application cannot give you.

The gap this addresses is a real one. Asked what they want to see on a resume, 59% of employers want specific skills backed by examples, against 16% who accept a list of specific skills and 2% a list of general ones (NACE, Job Outlook 2026 Spring Update, 161 respondents). The example is what a form cannot carry, and what cannot be sustained under follow-up questions about your own work.

Selecting when the signal has gone

Start with your own rules on AI. Two thirds of employers use no detection tool of any kind, and a third cannot say whether they have seen an AI-written application, which leaves an unwritten rule enforced by suspicion. Decide what is permitted at each stage, say so in the candidate pack, and design the stages you care about so that the question does not arise.

Check what can be checked, rather than weighing it. Degree awards are the obvious case, and SKEMA runs an online diploma verification service for that purpose.

Move the assessment itself to where it cannot be outsourced. Live work samples and structured conversations keep their signal. Take-home tasks have lost most of theirs.

Resist rebuilding the filter that has just been dismantled. Reinstating a 2:1 to cut the pile returns you to a proxy most of the market has abandoned, and screens on the wrong thing.

One point if you are considering handing the first cut to an algorithm: in the United States there is no single federal rule, and the obligations are accumulating state by state. From 1 January 2026 the Illinois Human Rights Act, as amended by House Bill 3773, prohibits the use of artificial intelligence that has a discriminatory effect in recruitment, hiring and promotion, and requires employers to notify applicants when AI is used. New York City has required an annual independent bias audit of automated employment decision tools since July 2023, with notice to candidates before the tool is used. Colorado's Artificial Intelligence Act has been amended and postponed more than once and its obligations substantially scaled back. Federal anti-discrimination law applies throughout, and the accountability stays with you rather than with your supplier. Confirm the position in force in every state you hire in before you buy anything.

The longest work sample you will ever get

The most informative assessment of a graduate available to you is the job itself, watched for long enough to judge.

Across successive editions of its Job Outlook survey, NACE finds internship experience to be the single most influential factor when employers choose between two otherwise equally qualified candidates, and an internship served at the hiring organisation itself outranks one served elsewhere in the industry. No 45-minute interview, however well structured, observes a candidate for as long or under conditions as realistic.

This route is competitive in its own right, which is the point: 63.1% of the 2024-25 intern class were converted into full-time hires, the highest rate in five years, and 88.3% of those offers were accepted (NACE, 2026 Internship & Co-op Report). Structured access to a school's cohort is worth more than an open advert. Each year more than 9,000 SKEMA students undertake an internship or apprenticeship lasting between two and 24 months, and companies can recruit early-career talent through job postings, the professional calendar and campus events. Sitting on juries and student challenges, available through SKEMA's corporate partnership options, puts your managers in front of students doing assessed work, which is a longer look than a stand at a fair allows.

What to measure

Applications received is not a performance indicator. In this market it is closer to a symptom.

Track the sequence that ends in a hire who stays: qualified candidates, offers, acceptances, and retention at 12 months and three years. Set each against a baseline, and give each an owner.

Three reference points are useful:

More than three quarters of new hires who interned with the employer are still there after a year, and more than half after five. For hires with no internship experience it is fewer than half at one year and around a third at five (NACE, 2026 Internship & Co-op Report).

The average gap between a student's first interview and an offer is 27 days; the student then takes 9 days to answer (NACE, 2025 Recruiting Benchmarks Report). The first of those is the one you control.

The median early-careers recruiting budget is about $114,000 a year (NACE, 2024 Recruiting Compensation Report). No published figure isolates the cost of a single channel, so treat this as the scale of the function rather than a cost per hire.

The comparison worth running is between what you spend sorting an open application flow and what you spend sourcing a smaller, pre-qualified one for the same number of viable candidates.