The short answer
Test the work itself: give each candidate a real task under real conditions, have someone who does the job watch and score it, and keep a recording. Work sample tests are among the strongest predictors of job performance in the hiring research.
Look at what other kinds of test actually check. Multiple choice checks whether you can recognize an answer. An essay checks whether you can write about a thing. Neither checks whether you can do it.
Frank Schmidt and John Hunter pulled together 85 years of hiring research. Work sample tests were among the best predictors of job performance, and years of education were among the weakest. A 2022 re-analysis by Paul Sackett and colleagues lowered many of the numbers. Methods that look straight at the job still came out ahead, and structured interviews, built around real situations and scored the same way for everyone, were the best single predictor in that review.
That points to a simple design. Pick a task from the real job. Tell candidates ahead of time what the task covers. Have a person who does the work watch the candidate do it, and score it against the same list for everyone. Record the session, so a bad day, or a biased one, can be checked later.
There's also the pile problem. One posting can bring hundreds of applicants, and no employer can run a long exam on all of them. A good test works as a second step, after something quick narrows the list. The test shows the manager who is worth an hour.
And you'll learn more than skill. Watch someone take in something new, stay with it when it's hard, take feedback and finish. Those are the things an employer wants to know, and they show up in a few weeks.
Use the five-part checklist. Is it real work? Under real conditions? Watched and recorded? Judged by people who do the job? Checkable by anyone afterward?
Proof can be done badly. A test can be too narrow, and an examiner can have a bad day. That's why recording and appeals matter. No method catches everything about a person, and work samples are one input, not the whole decision.
Good proof of skill is real work, done under real conditions, watched and recorded, judged by people who do the job, and checkable by anyone afterward.
A good interview answers "which of these ten?" Something still has to answer "which ten?", and that is the job the degree has been doing.
The one useful thing a bachelor's degree certifies, that a person can learn, focus, and be trained, can be shown directly in about a month instead of four years.
Bring human judgment back into teaching, hiring, and testing, and hold it accountable with recordings, appeals, public records, and ratings tied to real results.