Do you want to learn how to hire well? Then give up the thing most founders trust most. That thing is the feeling in the room. Almost everyone thinks they can read a person. We walk out of an interview sure we know whether someone is sharp, able, and a fit. Then we make big calls on the strength of that hunch.
But decades of hiring research point the other way. That confidence is mostly unearned. The biggest sign of whether a hire works out is not how good the candidate felt. It is not even how smart they are. It is how steadily and how carefully you sized them up in the first place.
That is the idea I want to make concrete. It flips the usual advice. The instinct is to hunt for the perfect candidate. But the real leverage is in fixing your process. A strong process plus ordinary judgment beats brilliant gut feel plus a messy one.
How you decide matters more than how good you think you are at deciding. That is genuinely good news. You can build a process. Gut feel is something you can only hope you have.
Why gut feel fails so reliably
The unstructured interview is the default everywhere. It is the open, free-flowing chat where you "get a sense" of someone and follow your gut. It is also one of the weaker tools we have. The problem is not that interviewers are foolish. It is that the format quietly lets bias in through the front door.
Each candidate gets different questions. So you never compare like with like. First impressions form in seconds. Then they bend everything that follows to confirm them. We warm to people who resemble us. We mistake confidence for skill. And we let one charming moment colour a whole judgment.
None of that feels like bias from the inside. It feels like good judgment. That is exactly what makes it risky. Your faith that you can read a person is strongest right where it is least reliable. A free chat does not measure the candidate much at all. It measures how the candidate makes you feel. Those are very different things. The second tells you whether you will like them. That is not the same as whether they will deliver.
What the research actually shows
The most current and careful review comes from Sackett and colleagues. They revisited decades of meta-analytic estimates. They also fixed math errors that had inflated older figures (Sackett et al., 2022). Two results from that work should reshape how a founder hires.
First, structured interviews are among the strongest signs of job performance we have. Their validity is roughly r = .42. A structured interview is not a personality chat. It uses the same job-relevant questions for every candidate. You ask them in the same order. You score them against a set rubric. That one change moves the interview from one of the weakest tools to one of the strongest.
Second, the picture on raw smarts shifted. This part surprises people. General mental ability was long seen as the king of predictors. But in the new estimates it came in at about r = .31. That is below structured interviews. Intelligence still matters. But how you size someone up matters more than how clever they are. A steady, job-relevant process catches things raw smarts cannot. It shows whether the person can do this work, with this team, under these conditions.
It is worth being precise about the history. A famous older number still gets passed around. An earlier landmark study put general mental ability far higher, at around r = .51. That figure was repeated for years (Schmidt & Hunter, 1998). The 2022 reanalysis showed it was overcorrected. It is now superseded. The honest position is simple. Treat the older number as a historical benchmark, not a current one. Plan your hiring around the updated figures.
What consistency buys you
The reason structure works is not a mystery. It forces you to compare candidates on the same terms. That is the only way a comparison means anything. Every applicant answers the same job-relevant questions. Every one is scored against the same rubric. Now the gaps you see start to reflect the candidates. They no longer reflect the order you met them in or your mood that afternoon.
Structure also slows the call down just enough. That lets evidence catch up with instinct. Define what good looks like before you meet anyone. Ask the questions that reveal it. Score each answer as you go. Do not form one big impression at the end and then reverse-engineer the reasons. Where you can, add a work sample. That is a real task close to the job. Watching someone do the work tells you more than hearing them describe it. None of this removes judgment. It disciplines it. Your judgment then works on like-for-like evidence instead of a fog of impressions.
This is the same rule that makes any part of a business reliable. You do not get steady outcomes from shaky processes. That holds in operations and in hiring. I work with founders on the people and systems side of scaling through Magna Hvati. Hiring is where the leverage is most underrated. A repeatable, structured way to size people up compounds quietly over years. Every good hire raises the bar for the next. And every mis-hire you avoid saves a cost. That cost rarely shows up on a spreadsheet. But the whole team feels it.
How to hire with a process you will actually use
The aim is not a red-tape ordeal. It is a few habits you apply every time. First, decide the handful of skills the role truly needs. Do this before you write a single question. Then ask every candidate the same questions tied to those skills. Score answers against a simple rubric in the moment. And add a work sample where the role allows it. Keep your instinct in the process as one input among several. Do not let it be the judge that overrides the rest.
The founders who hire best do not have the most magical read on people. They are humble enough to doubt the feeling in the room. And they are disciplined enough to build something better in its place. How you decide is the decision.
Key takeaways
- The biggest sign of a good hire is how steadily and how carefully you size people up. It is not how impressive the candidate felt.
- Unstructured interviews reliably let bias in. They use different questions, form fast first impressions, and mistake confidence for skill.
- Structured interviews are among the strongest signs of performance, at roughly r = .42 (Sackett et al., 2022).
- General mental ability predicts at about r = .31 in current estimates, below structured interviews. Intelligence matters, but process matters more (Sackett et al., 2022).
- The older r = .51 figure for general mental ability is a superseded historical benchmark, not a current number (Schmidt & Hunter, 1998).
Frequently asked questions
What is the best predictor of job performance? Current research points to structured interviews, at roughly r = .42. They sit ahead of general mental ability at about r = .31 (Sackett et al., 2022). The consistency of the evaluation does much of the work. Why are unstructured interviews so unreliable? They vary by candidate. They also let first impressions and similarity bias take over. So you end up measuring how the candidate makes you feel. You do not measure how they will perform. How do I make an interview structured? Define the skills the role needs. Ask every candidate the same job-relevant questions in the same order. Score each answer against a rubric as you go. And add a work sample where you can. Are you building a team and want a hiring process that compounds rather than costs you? That is the work I do with founders. You can see how I partner on my work with me page. And when the people and operations layer of scaling needs structure, Magna Hvati is where I help leaders build it.
References
Sackett, P. R., Zhang, C., Berry, C. M., & Lievens, F. (2022). Revisiting meta-analytic estimates of validity in personnel selection. Journal of Applied Psychology, 107(11), 2040-2068.
Schmidt, F. L., & Hunter, J. E. (1998). The validity and utility of selection methods in personnel psychology. Psychological Bulletin, 124(2), 262-274.
This article is for informational and educational purposes only and does not constitute financial, legal, tax, medical, or professional advice. Individual results vary.