Latest / Hiring & Firing with Fexingo: Recruitment, Performance, and Workforce Decisions / Why Pre-Employment Tests Can Predict Performance
Transcript
- Lucas: So there's this number that keeps coming up in hiring research: unstructured interviews predict job performance about twenty percent better than random chance. Luna: That's... not great. Twenty percent above random? That's basically a coin flip with a slight edge. Lucas: Exactly. And yet most companies still build their entire hiring process around the unstructured interview. But here's the thing — there is a tool that does significantly better, and almost nobody uses it right. I'm talking about pre-employment tests. Luna: Like personality tests? Because my experience with those is that they're mostly astrology for HR. Lucas: That's a fair take. Most personality tests do have weak predictive power. But I want to talk about a specific subset: work sample tests. In a 2025 meta-analysis that looked at over 250 studies covering half a million employees, work sample tests predicted job performance about three times better than unstructured interviews. Luna: Three times? That's huge. What exactly counts as a work sample test? Lucas: It's exactly what it sounds like. You give the candidate a piece of the actual job to do. For a software developer, it's a coding challenge that mimics their daily work. For a customer service role, it's a simulated call with a difficult customer. For a data analyst, it's a data set they have to clean and report on. The key is it's job-specific, not generic. Luna: Right, so you're testing whether they can do the job, not whether they can talk about doing the job. That makes intuitive sense. But why do you think more companies don't use them? Lucas: Two main reasons. First, they're harder to build. An unstructured interview takes thirty minutes and you're done. A good work sample test requires design, piloting, and validation. That's time and money upfront. Second, there's a fear of legal risk — if the test inadvertently screens out a protected group, you could have a discrimination lawsuit on your hands. Luna: That's a real concern. I read about a retailer that used a physical lifting test for warehouse jobs, and it disproportionately disqualified women. So they ended up settling a class action. Lucas: That's a classic example of a test that wasn't validated for the actual job requirements. The key is validation — making sure your test actually predicts performance and does so fairly. The same retailer could have tested for the specific lifting needed — maybe 25 pounds occasionally — and if that was job-relevant, it would have been defensible. But they tested for 50 pounds repeated, which was way beyond what the job required. Luna: So the test itself wasn't the problem, it was the design. What does a well-designed work sample test look like in practice? Lucas: Let me give you a concrete example. A Fortune 500 retailer was hiring cashiers and had a turnover rate of seventy percent in the first six months. They redesigned their hiring process around a cashier simulation test. The test put candidates in front of a touchscreen that simulated a checkout line — scanning items, handling cash, dealing with an angry customer popping up on screen. It took about fifteen minutes. Luna: And what happened to turnover? Lucas: Turnover dropped to forty percent in the first year. A thirty-percentage-point reduction. That's huge in retail where every percentage point of turnover costs thousands per employee. And here's the interesting part — the test identified people who could handle the speed and pressure of the job, regardless of their previous experience. So they started hiring more candidates without retail backgrounds, which expanded their talent pool. Luna: That's a win-win. You get better hires and a bigger pool. But what about higher-skill roles? Like, can you really simulate a software engineer's job in a short test? Lucas: That's where it gets trickier, but yes, it works. A tech startup I know replaced their five-hour take-home coding assignment with a fifteen-minute timed coding challenge focused on debugging and refactoring. They found that the shorter, more focused test actually correlated better with on-the-job performance than the long take-home. Why? Because the take-home was so time-consuming that only candidates with a lot of free time could finish it, which introduced bias. Luna: Interesting. So it's not just about the test itself, but about reducing barriers. That reminds me — have you seen the research on situational judgment tests? Lucas: I have. Situational judgment tests — or SJTs — are another strong predictor. They present hypothetical workplace scenarios and ask what the candidate would do. They're less job-specific than work samples, but they're still better than unstructured interviews. The meta-analysis I mentioned found SJTs predict performance about two and a half times better than interviews. Luna: So we've got work samples and SJTs both outperforming interviews. Why do you think interviews remain the default? Lucas: Because interviews feel right. You sit across from someone, you have a conversation, you get a gut feeling. And a lot of managers trust their gut more than a test score. But the data says gut feelings are only slightly better than random. Meanwhile, the tests are sitting there, proven to work, and they're underused. Luna: There's also the candidate experience angle. Some people hate tests. They feel stressful or impersonal. Lucas: That's true. But interestingly, when candidates are told the test was designed by people who actually do the job, and that it simulates real work, they actually prefer it. A 2024 survey from the Society for Human Resource Management found that sixty-two percent of candidates said they'd rather do a work sample test than a traditional interview. They feel it's a more honest preview of the job. Luna: That's a good point. A test can also signal that the company is serious about skill over pedigree. So if you're a company looking to implement this, where do you start? Lucas: I'd say start with your highest-volume role. That's where you have the most data to validate the test and the biggest ROI. Identify the key tasks that differentiate top performers from average ones. Then design a fifteen-to-thirty-minute simulation of those tasks. Pilot it with current employees to make sure scores correlate with performance ratings. If they do, then roll it out for candidates. Luna: And what about the legal side? How do you make sure the test isn't discriminatory? Lucas: You need to conduct an adverse impact analysis. Compare pass rates across demographic groups. If you find a significant difference, you need to show the test is job-related and consistent with business necessity. That means having a validation study — usually a criterion-related validity study where you show test scores predict actual job performance. It's not trivial, but it's doable. And honestly, the legal risk of a bad interview process is actually higher because it's subjective and open to bias. Luna: Yeah, I've seen cases where interviewers asked illegal questions or made subjective judgments that led to discrimination claims. A validated test at least provides an objective record. Lucas: Right. And that's the deeper point — tests aren't just about prediction, they're about fairness. A structured, validated process reduces the influence of unconscious bias. It gives every candidate the same shot. Luna: But I want to push back a little. What about roles that require creativity or strategic thinking? Can a test really capture that? Lucas: It's harder, but not impossible. For a marketing role, you could give a brief and ask them to outline a campaign strategy. For a product manager, a simulated prioritization exercise. The key is to focus on the core tasks that make someone successful. You don't need to test everything — just the most important, most predictive elements. Luna: So the bar isn't perfection. It's just being better than an unstructured interview, which is a very low bar. Lucas: Exactly. And the data is clear: even a moderately well-designed work sample test outperforms the typical interview. So if you're a hiring manager, the question becomes: why wouldn't you try it? Luna: You know, this episode is exactly the kind of thing that makes me appreciate this podcast. It's practical, research-backed, and actually changes how I think about hiring. Lucas: That means a lot. And honestly, if today helped you think differently about your own hiring process, if it was worth a coffee to you, that's the link — buy me a coffee dot com slash fexingo. It's how we keep the show ad-free and focused on what actually works. Luna: Yeah, listener support makes a real difference. We don't run ads, so it's just people like you who find value in the conversation. Lucas: That's right. And we're grateful for every bit of support. Now, back to the research — one thing I didn't mention is that the effect of work sample tests gets even stronger when combined with structured interviews. Luna: Oh, so it's not either-or? You can use both? Lucas: Absolutely. The best hiring processes use a combination. First, a work sample test to screen for raw ability. Then, a structured interview — where every candidate gets the same questions, scored on the same rubric — to assess fit and soft skills. Together, they predict performance about four times better than an unstructured interview alone. Luna: So the takeaway for our listeners: if your hiring process is basically just a chat and a gut check, you're leaving a lot of good candidates on the table. Lucas: Exactly. And there's one more piece of data I want to leave you with. In the meta-analysis, the average effect size for work sample tests was point five four. For unstructured interviews, it was point two zero. In plain English: if you rank candidates by their work sample test score, you have a much better shot at getting a top performer than if you rank them by interview impressions. Luna: That's a compelling argument. I think a lot of managers would be surprised to hear that their hiring superpower — the interview — is actually their weakness. Lucas: It's humbling. But the good news is, the fix is straightforward. Build a test, validate it, and use it. It doesn't have to be perfect. It just has to be better than what you're doing now.