Latest / Hiring & Firing with Fexingo: Recruitment, Performance, and Workforce Decisions / Why Your Interview Should Include a Hands-On Work Sample
Transcript
- Lucas: We've all been in an interview that felt like a well-rehearsed performance. The candidate nails every behavioral question — 'Tell me about a time you dealt with conflict' — and you walk out thinking they're perfect. Six weeks later, they're struggling with the actual work. Luna: Right, because answering a question about conflict is not the same as actually resolving a tense client call. There's a gap between what people say they'd do and what they actually do. Lucas: That gap is exactly what work-sample tests are designed to close. Instead of asking someone how they'd perform a task, you give them a realistic version of that task and watch them do it. For a software engineer, that might be a 45-minute coding exercise. For a marketing hire, it could be writing a short brief based on a brief you give them. Luna: And the research on this is pretty striking. I remember a meta-analysis from Frank Schmidt and John Hunter — the landmark one from 1998 — that found work-sample tests predict future job performance with a correlation of about 0.54. For comparison, traditional unstructured interviews are around 0.14. Lucas: That's a massive difference. A work sample is nearly four times as predictive as a standard conversation. Yet most companies still lean heavily on the chat and resume model. So today I want to look at a specific case — a mid-size B2B software company called CloudBridge that introduced a work-sample test for its technical support engineers back in 2023. Luna: And just before we go deep — we should mention that these conversations are possible because listeners like you choose to support the show directly. We deliberately don't run ads on these. If you want to support that choice, the link is buy me a coffee dot com slash fexingo. Lucas: Exactly. No sponsors, no interruptions, just the substance. So, CloudBridge — they had a problem. Their support engineers were taking nine weeks on average to reach full competency. And first-call resolution was hovering around 65 percent, which is below industry benchmark for B2B. Luna: Nine weeks of ramp-up is expensive. You're paying a salary while the person is essentially in training. And if they wash out after six months, you've lost the investment. Lucas: So their head of talent, a woman named Priya Nair, decided to redesign the hiring process. She replaced the standard behavioral interview with a 30-minute work sample. Candidates were given a simulated support ticket — a customer who had lost access to their account after a software update. They had to diagnose the issue, craft a response, and walk through their reasoning with the interviewer. Luna: Did they look at the candidate's technical knowledge, or also how they communicated the solution? Lucas: Both. The scoring rubric had three dimensions: accuracy of diagnosis, clarity of the written response, and the candidate's ability to explain their thought process. What they found was that candidates who scored in the top quartile on the work sample reached full competency in five weeks — down from nine — and their first-call resolution after six months was 84 percent, compared to 65 percent for the previous cohort. Luna: That's a 30 percent improvement in a core operational metric. And it happened because they were measuring actual job-relevant skill, not interview polish. Lucas: Priya told me something interesting: a few candidates who had great resumes and strong references bombed the work sample. And conversely, some candidates who had less conventional backgrounds — a self-taught candidate, a career-changer — performed really well. The test surfaced talent that the resume screen would have filtered out. Luna: That's a powerful argument for work samples as an equity tool. If you're only looking at pedigree, you miss people who have the skill but not the credential. But I wonder — does this scale to non-technical roles? Can you give a work sample for a salesperson or a manager? Lucas: Yes, but it requires more creativity. For sales, some companies use a 'mock sales call' where the candidate has to pitch a product based on a one-pager they're given 15 minutes to review. For managers, you can use a 'inbox simulation' — give them a pile of emails, some urgent, some strategic, and see how they prioritize. Luna: I've seen those used in assessment centers for decades. But they're often expensive to administer. CloudBridge's test was only 30 minutes and done over video call. What about the candidate experience? Do people push back? Lucas: Some do. Priya said about 5 percent of candidates declined to proceed when they learned about the work sample. But she framed it as a positive filter — if someone is unwilling to demonstrate their skills in a low-stakes setting, that might be a red flag. The vast majority of candidates actually reported feeling more positive about the process because it felt fair and job-relevant. Luna: There's also the risk of bias. If the work sample requires specific tools or knowledge that not everyone has equal access to, you could inadvertently screen out good candidates. For example, a coding test that assumes familiarity with a particular framework might favor candidates who've used it recently. Lucas: That's a real concern. The fix is to design the test so it relies on transferable skills, not proprietary tools. CloudBridge's support ticket, for instance, used a generic scenario that any tech-savvy person could tackle. They also allowed candidates to use any resources they normally would on the job — Google, documentation, whatever. That makes it a more authentic simulation. Luna: That's smart. You're testing how they solve problems in the real world, not under artificial constraints. So the key is to make the work sample representative, not restrictive. Lucas: Exactly. And there's another angle: work samples can also reduce the number of interviews you need. CloudBridge cut their interview process from four rounds to two — a phone screen and the work sample. That saved hiring managers about six hours per candidate. And their offer acceptance rate actually went up, because candidates appreciated the efficiency. Luna: So it's a win for the company and for the candidate. But I'm curious about the reliability of a single 30-minute test. What if someone just has a bad day? Lucas: That's a fair point. The research suggests that work samples are reliable enough to be a strong predictor, but they shouldn't be the sole decision point. CloudBridge still did reference checks and a brief values alignment conversation. But the work sample carried about 60 percent weight in the final decision. Luna: Sixty percent — that's a significant shift from the traditional model where the interview usually dominates. It forces hiring managers to trust the data from the simulation rather than their gut feeling about someone's likability. Lucas: And that's the cultural change that's harder to implement. Priya told me that some managers initially resisted because they felt they could 'just tell' who was good. But after six months of data showing that work-sample hires outperformed, the skeptics came around. Luna: It's hard to argue with results. One last thing — what about roles that are more ambiguous, like strategy or innovation? Can you build a work sample for a role where the day-to-day is unstructured? Lucas: It's tougher, but not impossible. For a strategy role, you might give a case study — a real business problem the company is facing — and ask the candidate to outline an approach. The key is that the task should mirror the cognitive demands of the job, even if the output isn't a perfect replica of the work. The goal is to observe how they think, not just what they produce. Luna: So the principle holds: if you can simulate a core chunk of the job, you'll get better signal than from a conversation about hypotheticals. It seems like such an obvious idea, yet most companies don't do it. Lucas: Partly because it's easier to schedule a chat than to design a valid work sample. It takes upfront effort to create the task, pilot it, and train interviewers to score it consistently. But the payoff — in terms of better hires, faster ramp, and lower turnover — is substantial. Luna: And it forces you, as a company, to really define what success looks like in the role. You can't create a work sample unless you know exactly what skills matter. Lucas: That's a benefit that goes beyond hiring. The process of designing the work sample often reveals gaps in the team's understanding of the role. CloudBridge's support team realized they had been overvaluing technical certifications and undervaluing communication skills. Their work sample flipped that. Luna: So next time you're hiring, maybe skip the 'tell me about your strengths' question and hand them a problem to solve instead. It might tell you more in 30 minutes than an hour of conversation. Lucas: Exactly. And if you want to check out the original research, the Schmidt and Hunter meta-analysis is still a great starting point. It's from 1998 but the findings have been replicated many times since. Thanks for listening — we'll be back next week with another angle on hiring.