Screening 200 resumes for one role without missing the right profile

The problem is not volume, it is drift

Nobody reads the hundred-and-fiftieth resume the way they read the first. Criteria move along the way: you grow stricter after a run of good applications, more lenient after a run of poor ones. The final ranking reflects the order of the pile as much as the quality of the candidates.

That is the first thing a systematic approach fixes: not reading faster, but reading every file against the same demands as the one before.

Set the requirements before opening the first resume

Writing down what you want after reading a few applications describes the candidates who applied rather than the role. The reverse order is the right one: list the requirements, then rank them in three levels.

Three are enough, and the ranking matters more than the list itself:

• critical: without it the person cannot hold the role. Rarely more than four or five lines;
• important: what separates two viable profiles;
• nice to have: what can be learned in a few weeks, and therefore has no business rejecting anyone.

Compare on the same criteria, not on impressions

Once the requirements are set, every resume meets the same grid and the result reads flat: which requirements are covered, which are not, on what evidence.

The point is not the ranking — that is up for discussion — but comparability. Two recruiters looking at the same grid are finally talking about the same thing, and a disagreement lands on a named criterion instead of a preference.

The keyword trap

Filtering on the presence of a term rejects two populations at once: those who lack the skill, and those who have it but call it something else. Absence of evidence is not evidence of absence.

A useful screen therefore looks for what backs the skill — an assignment, an achievement, a date — rather than the word itself. It is slower, and that is what separates a screen from a filter.

Keep a record of decisions

Writing down why a file was set aside costs ten seconds and pays three times: to answer the candidate who asks, to justify the choice six months later, and to check afterwards whether the stated criteria were the right ones.

It is also the only serious defence if a decision is challenged. A reason written at the moment of the decision is worth infinitely more than a reconstruction from memory, and it forces you to state a criterion rather than an impression — which is often enough to reveal how thin it was.

What you gain

Time, yes — but above all the certainty that the twentieth file was read with the same demands as the second, and that the chosen profile was chosen for a reason you can still state a month later.