How to Shortlist Candidates for an Executive Role

Every shortlist is built by throwing people away. Say two hundred names become fifty, fifty become twenty, twenty become six. The work is not finding people. The work is deciding who goes.

Which means the honest test of a shortlisting process is not the quality of the names that survive. It is whether you can say, for each person you cut, which requirement they failed and what evidence you looked at. Most processes cannot answer that, because most cuts happen on a title, a company name, or a scan that took four seconds.

Here is how to run it so the cuts are defensible.

Start with criteria you can actually evaluate

Shortlisting fails upstream more often than it fails in the shortlisting. If the requirements cannot be checked against evidence, no filtering discipline downstream will save you.

Compare these two versions of the same requirement:

• “Strong operational leadership in a complex manufacturing environment.”
• “Has held P&L responsibility for two or more manufacturing sites simultaneously, at least one of them in a regulated industry.”

The first cannot be confirmed or denied. Every candidate passes it if you like them and fails it if you do not, which means it is not doing a criterion’s job. Your judgment will land there later, unrecorded.

The second produces a result: yes, no, or ambiguous with a reason. Before you filter anyone, go through the brief and convert every requirement into something with that property. What is left over — the things that genuinely resist definition, like presence in a boardroom — is real, but it belongs to the interview stage, not the research stage. Keep it out of the filter. Writing the brief so it can carry this weight is its own piece of work.

Then separate must-haves from strong-preferences. A must-have failure removes someone. A preference failure changes where they sit, not whether they exist.

Stage the filter — wide first, deep last

The instinct under time pressure is to narrow early. That instinct is the most expensive mistake in the process, because a candidate cut in the first pass is never seen again. There is no later stage that recovers them.

Run it in three stages instead.

Stage one — the wide pass. Assemble the pool before you evaluate anyone. Cast wider than feels comfortable. The cost of an over-inclusive pool is a few wasted minutes at the next stage; the cost of an under-inclusive one is a candidate nobody will ever know was missing. You are not judging yet. You are making sure the people who should be considered are in the room. How that discovery pass is actually assembled is a separate discipline.

Stage two — the broad filter. Now apply the small number of criteria that can be checked quickly and reliably from what is already visible: function, seniority band, sector, geography, obvious disqualifiers. Nothing that requires interpretation. This is where most of the pool goes, and it should be fast — you are removing people who could not plausibly qualify no matter what the deep research found.

Stage three — the deep pass. Whoever survives stage two gets researched properly against the full rubric, one criterion at a time, with sources attached. This is where the hours go. Running it on a filtered pool rather than the raw one is the whole reason stage two exists.

The sequencing matters more than it sounds. Deep research on the wrong fifty people produces a beautiful document about the wrong fifty people.

Filtering feels like progress because you can count what you removed. Nothing counts what you removed by mistake.

The order of the questions inside the deep pass

Within stage three, quick-to-check disqualifiers go first. If someone is geographically immovable and the role cannot relocate, you find that out before spending an hour reconstructing their operating scope.

After that, work through the must-haves in the order the client cares about. For each one, you want a result and a source: confirmed with a citation, unconfirmed because nothing was found, or flagged because the evidence is ambiguous. Three states, not two. Collapsing “unconfirmed” into “no” quietly cuts people for the sin of having a thin public record, and that says nothing about whether they can do the job.

What that per-criterion output looks like on the page is worth getting right, because it is what your client will read.

Do not throw away the near-misses

The pile of people who failed exactly one criterion is the most valuable byproduct of the whole process, and most searches discard it without looking.

Go back through it once, deliberately, asking a different question: who misses something stated but has the substance the role actually requires?

The plant director who has never carried the VP title but has run the same footprint. The operator from an adjacent regulated sector where the constraint set is nearly identical. The person whose company is a size band below the one specified but whose scope was larger than their title implies.

These are laterals, and they go into the shortlist as a separate section with the miss named explicitly and the reasoning attached. You decide whether the trade is worth making — that call requires context about the client that the research does not have. But the candidate should reach you, with the trade-off stated, rather than being silently removed by a rule.

A shortlist that only contains people who cleared every stated box is a shortlist that a database query could have produced.

Five ways shortlisting goes wrong

Filtering on title. Titles are inconsistent across company sizes and sectors. A “Director of Operations” at a €2B manufacturer may hold more scope than a “VP Operations” at a €200M one. Filter on scope; use title as a hint.

Cutting for a thin public record. Some strong operators barely exist online. Absence of evidence, recorded as failure, systematically removes them.

Single-source dependence. If every claim traces to one profile page, you are shortlisting on what people wrote about themselves.

Mistaken identity. Two people with the same name in the same industry is not rare. A merged record produces a candidate who looks extraordinary and is not real. This is the failure mode that does the most damage, because it survives every downstream check that is not looking for it.

Padding to a number. If six people confirm and the client expects ten, four entries in that document exist to fill space. You will find out which four in the meeting.

What “done” looks like

Shortlisting is finished when you can take any name in the document — included or excluded — and say which criteria they met, which they did not, and what you looked at to decide. Not “we reviewed them.” The actual source.

If you can do that for every name, the shortlist will hold up under questioning, and the cuts you made were the cuts you meant to make. If you cannot, the list may still be correct. You just have no way to know, and neither does your client.

Related reading: candidate research as a whole practice, executive search best practices for candidate research, how to screen executive candidates, and the research bottleneck in boutique executive search.