Hiring in 2026: What the Research Shows About What Predicts Job Performance

Hiring decisions are among the most consequential a manager makes — the people you bring into your organization shape its output, culture, and trajectory in ways that are difficult to reverse once made. They are also, based on the research, made using methods that have surprisingly poor predictive validity for actual job performance. After years covering organizational psychology and business, here is the honest guide to what the evidence shows about which hiring methods actually predict who will perform well.

The Interview Problem: Why Your Gut Is Often Wrong

The unstructured interview — a free-flowing conversation where the interviewer asks whatever seems relevant and forms an overall impression — is by far the most common hiring method and has consistently poor predictive validity for job performance. The research across decades of industrial-organizational psychology is consistent: unstructured interviews have correlations with subsequent job performance in the range of 0.2-0.3 (where 1.0 is perfect prediction and 0.0 is no better than chance). The reasons for this underperformance: interviewers are susceptible to halo effects (overall favorable impression from one characteristic influencing evaluation of others), affinity bias (preferring candidates similar to themselves), and overweighting factors like verbal fluency and physical attractiveness that correlate poorly with job performance in most roles.

The specific interview content that interviewers typically focus on — rapport, personality impression, hypothetical answers to questions like where do you see yourself in five years — predicts job performance poorly. The interview content that predicts better: specific behavioral evidence about what the candidate has actually done in past situations relevant to the job.

What Actually Predicts Job Performance

The industrial-organizational psychology meta-analyses on selection validity are among the most replicated findings in applied psychology. The methods with the strongest predictive validity: work sample tests (having the candidate actually perform a representative sample of the job — coding test for a developer, writing sample for a writer, case study for a consultant) have correlations with job performance in the range of 0.54. Structured interviews — asking every candidate the same predetermined questions with predefined scoring criteria, particularly behavioral questions (Tell me about a time when...) — have validity in the range of 0.51. Cognitive ability tests (general intelligence measures) have validity in the range of 0.51 for complex jobs. Job knowledge tests have validity of approximately 0.48. Conscientiousness personality assessment (carefully validated instruments, not impressionistic) has validity around 0.31.

The methods with poor predictive validity that remain widely used: years of experience (0.18 correlation with performance — weak), reference checks as typically conducted (0.26 — moderate, but rarely done well), education credentials for most roles (0.10 — weak), and unstructured interviews (0.20-0.28 — weak to moderate).

The Practical Changes That Improve Hiring

Moving from unstructured to structured interviews is the single highest-leverage change most organizations can make — it requires the same time investment and produces significantly better predictive validity. The structure elements that matter most: predetermined questions that all candidates receive (eliminating the variable of which questions an interviewer happened to ask), behavioral question framing (past behavior is more predictive than hypothetical behavior), and defined scoring criteria applied consistently. Adding a work sample assessment appropriate to the role provides the highest-validity selection data available. Debiasing during resume review — evaluating resume content against job-relevant criteria rather than against overall impression — reduces the impact of name-based, education-based, and photo-based biases that affect unstructured resume evaluation.

Honest Bottom Line: Unstructured interviews (the most common hiring method) have weak predictive validity (0.20-0.28 correlation with performance) due to halo effects, affinity bias, and overweighting verbal fluency and appearance. Methods with strongest evidence: work sample tests (0.54 correlation — have the candidate actually do a job sample), structured behavioral interviews (0.51 — same predetermined questions, defined scoring), and cognitive ability tests for complex roles (0.51). Methods with poor validity still widely used: years of experience (0.18), education credentials for most roles (0.10). The highest-leverage change: moving from unstructured to structured interviews requires no additional time and substantially improves predictive validity. Adding a relevant work sample provides the highest-validity data point available for most roles.