A candidate shortlist template that flags where your signals disagree

I’ve never sat in a hiring debate that was actually about the ranking. It’s always about one candidate. The one whose resume says yes and whose interview says something else, sitting at number eleven where the scorecard quietly filed them.

So here’s the shortlisting matrix I’d build instead. Six columns, one row per candidate, and a status column that names a decision rather than a rank. Copy it into Sheets or Excel and fill in whatever signals you actually collect.

The candidate shortlist template

Column	What goes in it
Candidate name	Self-explanatory
Resume signal	Pass, borderline, or fail against your must-haves, plus one line of evidence (“5 years in a similar role” or “no direct experience”)
Interview signal	Same three-value scale, plus one line of evidence from the one-way interview answers
Assessment signal	Same scale, plus one line of evidence, only if the role uses an assessment
Agreement	Do the filled-in signals point the same direction, or not
Status	Advance, pass, or needs a look

The status column is where it earns its keep. Three rules, and they don’t change no matter how many signals you’re running:

• All signals agree, and agree positive. Advance. You don’t need to reread this row before you call them.
• All signals agree, and agree negative. Pass. Same certainty, other direction.
• Signals disagree. Resume up and interview down, or the reverse. Needs a look. This is the row that gets your actual attention.

If you only screen one way, start here

Plenty of owners only read resumes, and this template still beats a ranked list for that. With one signal there’s nothing to disagree with, so every row lands in advance or pass, and what you get is a clean decision list with the evidence written next to each name instead of a column of scores you have to re-interpret at 10pm.

The third pile appears the day you add a second signal. Send a one-way interview to the candidates who cleared the resume, or run a talent assessment on the finalists, and the needs-a-look rows sort themselves out of the pile automatically. You don’t rebuild the sheet. You fill in one more column.

Why most shortlisting matrices rank instead of decide

Open any shortlisting matrix you can download today and the bones are identical. A column per hiring criterion (experience, a required skill, a certification), a weight per column, a formula that collapses all of it into a single sortable number. Sort the sheet. Whoever floats to the top is your shortlist.

That design was built for a single input. One person, or one panel, scoring each candidate against a job spec from a resume and maybe a phone screen folded into the same number. When there’s one stream of evidence, ranking it is a perfectly reasonable thing to do.

But the moment you screen a second way, the matrix has exactly one answer for what to do with two numbers. Average them. And averaging treats disagreement between your signals as noise to smooth over, when disagreement is the most useful thing on the page.

Three candidates, one ranked list

Say you’re hiring a coordinator. Three candidates, scored out of 10 on the resume and again on the interview.

The first is an 8 and an 8. Strong on paper, strong on camera.

The second is an 8 and a 4. The resume checks every box. The interview answers are flat and generic, the kind where you learn nothing you couldn’t have guessed from the resume.

The third is a 5 and an 8. Thin resume, nonlinear path, and then answers that walk you through exactly how they’d handle the job.

Blend those and candidates two and three land within a rounding error of each other, somewhere in the middle of your list, both looking like a maybe. They are not the same kind of maybe. One photographs well on paper and might not be who the paper promised. The other might be exactly who you want, undersold by a resume that can’t show what the interview did.

The ranked list treats them identically. They deserve opposite next steps.

This is the same failure candidate screening software for small staffing agencies runs into at higher volume. The more signals you layer on, the more a single number hides. It’s just louder at agency scale. It’s happening on your 40-candidate pile too, and it’s the same tension you hit when you try to screen candidates without losing your best matches. Speed and accuracy pull against each other until the process can hold more than one signal at a time.

That third status is the entire point of keeping the columns separate. It gives you three or four rows worth a real read, instead of a top-ten list padded with candidates who never needed the read at all.

Why not just add a weighted fourth column?

Because a fourth column that blends the first three brings back the exact problem the template exists to solve. The second resume, interview, and assessment collapse into one weighted number, the disagreement vanishes again. Just one column later than before.

Keep the signals separate. Let the status do the sorting.

Isn’t this more work?

Fair objection, especially at 9:40pm. But the extra columns aren’t extra reading. They’re the opposite.

A ranked shortlist tells you to read the top ten rows in full, because “top ten” is the only information a rank carries. This template tells you to read the “needs a look” rows in full and skim everything else, because the advance and pass rows already told you what to do. On a 40-candidate pile, that’s usually the difference between rereading 10 candidate files and rereading 3.

The one-way interview does part of this work before you touch the sheet. Instead of guessing at an interview signal from memory, you’re filling that column from a recorded answer you actually watched. The “needs a look” pile ends up built on evidence, not on a hunch you scribbled down at 11pm. It’s the same math behind what manual screening actually costs you in hours. Fewer full rereads is fewer hours, whether or not anything else in your process changes.

What it looks like on a 40-candidate pile

You post a coordinator role and 40 people apply by the weekend.

Resumes first. 25 clear your must-haves, 15 don’t. The 15 get marked pass and you are done with them tonight.

The 25 who cleared get a short one-way interview. 18 finish it. As answers come in, you fill the interview column next to the resume signal you already recorded.

Most of those 18 land in agreement. Call it 11 agree-positive (strong resume, strong interview) and 4 agree-negative (thin resume, interview confirmed it). Those 15 rows are closed. You won’t reread any of them before deciding who to call.

The three that get your evening

That leaves 3.

One has a two-year resume gap and an interview answer that explains it. A caregiving stretch that taught patience under pressure, which happens to be the job. One has a polished resume and an interview that felt rehearsed and thin on specifics. One is genuinely unreadable either way.

Those 3 rows are your evening. Not the other 15, and not all 40. Three candidates, with the reason each one needs a second look already written next to their name.

That’s a shortlist you can act on by tomorrow, and it cost you fewer full rereads than a top-ten list would have. It’s the same discipline behind a screening workflow built for an owner hiring alone. Fewer steps, and every one of them earns its place.

Where the spreadsheet stops

Here’s the honest limit. The template is free, it works in any spreadsheet, and it’ll sharpen your shortlist the day you start using it.

But something still has to produce the per-signal evidence that goes in each cell. That’s the part a spreadsheet can’t do.

It can hold a resume signal. It can’t read the resume and decide what the signal is.

That’s the real work behind every row. Reading 40 resumes against your criteria, watching however many interview answers came back, and writing an honest one-line summary of each. Most owners end up doing that part manually too, at night, on top of the job that already filled the day.

What Truffle fills in

Truffle is candidate screening software for small businesses that combines resume screening, one-way video interviews, and talent assessments, so the per-signal evidence gets produced for you instead of by you at 10pm.

Upload your requirements and AI Match scores every resume against your own criteria with the reasoning shown next to each score, so the resume column fills itself. If you’re running one-way interviews, the AI summary does the same job for the interview column. A short overview of each candidate’s answers and what stood out, instead of you transcribing your own notes into a cell.

The agreement logic still needs you, and that’s deliberate. AI surfaces the evidence per signal. You’re the one who looks at a strong resume sitting next to a flat interview summary and decides that candidate has earned a real second look instead of an auto-advance. Truffle doesn’t skip that step. It’s the step the whole template was built to protect, spreadsheet or not.

If you’d rather not build the sheet by hand, plans start at $49 a month with a 7-day free trial, 30 credits, and no card required. The dashboard sorts into For Review, Reviewed, and In Progress out of the box. Same three-pile logic, already filled in.

A shortlist was never a leaderboard

Ranking made sense when hiring produced one stream of evidence per person. It doesn’t anymore, and the tooling most owners inherited hasn’t caught up. Every new signal you add makes a ranked list slightly less honest, because every new signal is one more thing the average can quietly bury.

So the shortlist stops being a list of your best candidates. It becomes a list of the decisions that still need a human, with everything else already sorted out of the way.

Build the spreadsheet if that’s where you are today. Just don’t let one blended score talk you out of the row that disagrees with itself. That row is usually the call worth making.

If you’d rather see the per-signal evidence filled in from the first upload, see how Truffle’s resume screening works.

Frequently asked questions about candidate shortlist templates

How many candidates should end up on a shortlist?

There’s no fixed number. It depends on how many “needs a look” and “advance” rows your pile actually produces, not on a target you set beforehand. A role that pulls 40 loosely qualified candidates might shortlist 5. A role with a thin, well-matched pool might shortlist 12. Resist rounding the shortlist to hit a number that isn’t in your data.

What if I only use resume screening, with no interview or assessment?

Use the same six columns and leave the interview and assessment cells blank. Every row will land in advance or pass, which still beats a ranked list, because you’re writing down a decision and the evidence behind it instead of a score you’ll have to re-interpret later. The needs-a-look pile shows up the day you add a second signal.

Do I need special software to build this?

No. It’s six columns in a spreadsheet, the same as a resume screening checklist or a recruitment dashboard. Software like Truffle fills the per-signal columns for you with match scores and summaries, but the structure works by hand in Google Sheets or Excel.

What counts as disagreement between signals?

Anything where one signal is clearly positive and another is clearly negative for the same candidate. Not small variation. A resume at 8 and an interview at 6 isn’t disagreement, that’s normal spread. A resume at 8 and an interview at 3 is. Set your own threshold for the scale you’re using and hold it consistent across the whole pile, so you’re flagging a pattern instead of following a mood.