Hiring, and the filters · 4.1

What a CV actually selects for

What a CV actually selects for. What is actually the case, and how it compares with what is repeated.

A practical software reference for this part of the discussion is Monitask's overview of remote employee tracking.

A curriculum vitae is an unverified self-report, read for about seven seconds, and the strongest published findings about it concern things it was not meant to convey.

The name experiments

The standard method is to send matched applications differing in one detail. A study published in 2004 sent identical resumes to American advertisements with names chosen to signal race, and found that applications with white-sounding names received about fifty per cent more callbacks.

The design has been repeated in many countries with many characteristics. A meta-analysis covering twenty-five years of American field experiments found no decline in the rate of discrimination against Black applicants over that period, which is a more uncomfortable finding than the original.

The employment gap

Callback rates fall with the length of time an applicant has been out of work, and the fall is steep in the first months. The mechanism is inference: employers read duration as a signal about quality, and every employer reading it the same way makes it self-confirming.

The consequence is that a period of unemployment is a penalty applied by the labour market rather than by any individual employer, which makes it very difficult for any individual employer to stop applying.

What else leaks

Age, through graduation years. Address, which in several studies carries an effect of its own. Gaps for caring responsibilities. Schools and universities, which carry class information in most countries.

None of these is what the document is for and all of them are legible in seven seconds.

Anonymous applications, and the complication

Stripping names and addresses is the obvious remedy and the evidence is genuinely mixed. Some trials find improved callback rates for minority applicants. A large French trial found the opposite, and the explanation offered was that recruiters who had been actively compensating could no longer do so.

That result is worth sitting with, because it says the observed outcome was the product of two opposing forces and removing one of them made things worse.

What a CV cannot tell you

Whether any of it is true, mostly. Employment dates are checkable and frequently checked; achievements are neither. The document is a claim, and its predictive value for job performance is modest at best.

What predicts better

A work sample: a short piece of the actual job, performed and scored against a rubric written in advance. It outperforms the document, it outperforms an unstructured conversation, and it is more work to administer, which is why it is rare.

For roles where a sample is impractical, a structured interview scored against defined criteria is the next best thing, and the entry on interviews sets out what makes one structured.

The practical advice, such as it is

Plain formatting without tables or columns, because of parsing. Language mirroring the advertisement, because of keyword matching. Explicit dates. And a stated reason for any substantial gap, since the inference made in its absence is worse than most true explanations.

All of that is advice about surviving a filter rather than about demonstrating competence, which is a fair description of what the document has become.

What an employer can do

Score against criteria written before reading anything. Have more than one person score independently. Remove the fields not being scored. And audit outcomes by group afterwards, because the studies above describe employers who did not believe they were doing any of this.

Seven seconds

Eye-tracking studies of recruiters reviewing applications report initial screening times measured in single-digit seconds, spent mostly on name, current title, current employer, dates and education.

Everything else on the page is read only if those five survive, which is worth knowing before spending an evening on the wording of a bullet.

What the studies do not show

They do not show that any individual recruiter is prejudiced. A callback gap measured across thousands of applications is a property of a system, and it is compatible with most participants believing sincerely that they judge on merit.

That is why the useful responses are procedural rather than attitudinal: scoring criteria fixed in advance, independent scorers, and an audit of outcomes afterwards. Each works whether or not anybody's beliefs change.

The one change worth making

Decide the scoring criteria before opening the applications. It takes twenty minutes, it is the step everybody skips, and every other improvement described here depends on it existing.

For broader context, consult Europass CV guidance.