Campus recruitment: screening thousands of graduates fairly and fast

A campus drive brings in large applicant pools from many universities, and on paper a lot of them look the same: similar degrees, grades and activities. Three things make the first cut harder than it looks.

• The CV has little to rank on. Graduates have little or no work history, so recruiters fall back on what's left, usually the university's reputation and the GPA.
• Grades don't travel well. A 3.5 at one university isn't the same as a 3.5 at another, and grading standards differ between faculties of the same university.
• The usual tests are easy to prepare for. Answers to the common psikotes are a quick search away, so the results often show how well someone prepared.

Add a deadline, a team that scores results by hand, and candidates who give up halfway through a long test, and it's easy to end up with a shortlist that's quick but hard to defend.

Less than you'd hope: a 2024 meta-analysis in the Journal of Applied Psychology combined 114 studies of academic performance and work performance. The average link between the two was .21. Academic performance predicted training performance better, at .34, and didn't predict turnover at all.

Two of its findings matter for campus screening:

• The link varies a lot. Grades were most useful when the courses were directly relevant to the job, such as practicum classes, where the estimates rose into the .30s and .40s.
• Comparing within a university works better. Grades predicted better in samples from the same university or major. Across many campuses, the same GPA means different things.

The study's authors conclude that organizations should think carefully about whether and how grades are relevant to a particular job. For a drive across many universities, that's a case for using grades as an eligibility rule, if at all, and ranking on something everyone takes in the same way.

Filtering on the name of the university has a cost too: it shrinks the pool before you've measured anyone.

For campus recruitment we pair two DEUS modules. Together they show who each graduate is and how they think.

The same pairing works for any role with a large applicant pool. For a Management Trainee program, where the final round needs a business simulation, see the guide on Management Trainee recruitment.

Keep the online assessment for the first cut and save the time-consuming methods for the shortlist. Structured interviews are worth that time: a 2022 re-analysis of selection research put them first among common methods for predicting job performance, at .42. They take too long per candidate to run on the whole pool.

The logistics are where campus drives lose the most time. A setup that runs like this works for one campus or fifty:

1. One list. Add every applicant from one CSV file, whichever campus they applied through. Branded invitations go out automatically, and candidates who haven't started get a reminder.
2. Their own device. Candidates play in a phone or laptop browser, with nothing to download. They need a stable internet connection. Device checks and anti-cheat logic run before the first game.
3. Live progress. You see who has started, who has finished and who hasn't opened the invitation.
4. A ranked list. Each candidate's report is ready when they finish. The batch report ranks every applicant and flags those below your cut-off score, and the interactive dashboard lets your hiring committee filter by score band and open any candidate's full profile.

A process can look fair and still screen out one group far more than another, and counting is the only way to know.

• Compare pass rates. For each stage, divide the number who passed by the number who took it, separately for each university and region, and for men and women. US hiring rules treat a group's pass rate below four-fifths of the highest group's as a first sign of adverse impact, a simple yardstick you can borrow anywhere.
• Look at each stage. If one university's graduates pass the assessment at the usual rate but drop out at the interview, the interview is where to look.
• Use the right norms. Scores should be compared with the population your candidates come from. DEUS assessments are normed on the Indonesian population, not on Western samples.
• Keep the conditions equal. Everyone takes the same assessment, with the same time limits, on their own device. Nobody gets an advantage because their campus was visited first.

Early research finds game-related assessments have no greater adverse impact than conventional tests, but the studies are still few and results vary with what's being measured, so check your own numbers.