Why CVs Are a Poor Predictor of Job Performance

A CV is written by the candidate, optimised for the purpose of getting an interview, and reviewed under time pressure by someone who may read hundreds of them. At every stage of this process, the goal is impression management rather than accurate capability assessment. This is not a flaw in any individual. It is the nature of what a CV is.

The research on CV screening as a predictor of job performance is not encouraging. Studies on hiring validity consistently find that unstructured CV review has low predictive power for actual job outcomes. The signals that tend to dominate CV screening (prestige of previous employers, years of experience, educational credentials) are correlated with some aspects of performance in some roles, but they are far from reliable proxies.

The keyword optimisation problem

Automated screening has added a further layer of distortion. When CVs are filtered by keyword matching before any human reviews them, the effective selection criterion becomes: does this person know which words to include in their CV?

This selects for candidates who understand how automated screening works, or who have paid for CV optimisation services, or who happened to use the same terminology as the job description. It does not select for candidates who can do the job well.

The practical effect is that the candidate pool that reaches a human reviewer has already been filtered by a criterion that has no direct relationship to job performance. Strong candidates who write plainly, or who came from a context that uses different terminology, or who do not know to keyword-optimise, may not make it through.

What a CV cannot show

There are categories of capability that are genuinely difficult to represent on a CV:

• Transferable skills: the ability to apply expertise across domains often does not map onto a job title or a list of employers
• How someone works: how they approach ambiguous problems, how they collaborate under pressure, how they handle being wrong
• Growth trajectory: a candidate who has developed significantly in the past two years may look less experienced on paper than someone who has been doing the same thing for five
• Potential: particularly relevant for roles where the expectation is that the person will grow into expanded responsibilities

None of these are legible from a CV. They require a different form of evaluation to surface.

What research says about higher-signal alternatives

The hiring research literature has converged on a reasonably clear hierarchy of what actually predicts job performance:

1. Work sample tests: asking candidates to do a task representative of the actual job is the highest-validity method available
2. Structured interviews: interviews where all candidates are asked the same questions and assessed against a consistent rubric substantially outperform unstructured conversations
3. Cognitive and skills-based assessments: well-designed assessments of relevant capabilities predict performance better than most other inputs

Unstructured CV review and unstructured interviews, which are the two most common elements of most hiring processes, sit near the bottom of this hierarchy.

The persistence of the CV

Given all of this, why does CV screening remain the dominant first step in most hiring processes? Several reasons:

• It is familiar, fast to administer at scale, and requires no setup or design
• Most hiring teams have not had visibility into the research on predictive validity
• The consequences of poor screening are diffuse and delayed. It is hard to trace a mediocre hire back to the CV screen that let them through

None of these are good reasons to keep the CV as the primary filter. They are explanations for why a suboptimal practice persists.

This does not mean CVs have no place in the process. They provide useful context and conversation starting points. The problem is treating them as a reliable signal for who can do the job. They are not.