Moving the HireQwik Shortlist Threshold After Scores Exist

The HireQwik shortlist threshold is a number I have moved in front of a hiring manager, on purpose, so they could see the list breathe. The scores were already done. At 60, the table showed 46 names. At 75, it showed 11. Nobody was re-read. The files did not change. Only the mark on the row changed. That is the whole feature, and it is also the whole way to misuse it. If you slide the number until the list matches a headcount you promised, you did not shortlist. You trimmed.

This post is about that cut on the Shortlist screen, after a resume upload or a link sheet has been scored. It is not about voice-interview verdicts. The path that produces the scores from loose files is in the upload walkthrough.

What the HireQwik shortlist threshold is doing

Each finished resume has a score from 0 to 100. When the job completes, HireQwik reads one cut. Out of the box, that cut is 60. A completed score greater than or equal to the cut is stored as shortlisted. A completed score under the cut is stored as not shortlisted. Both groups remain in the job. The low scores are hidden only if you filter them out.

The cut does not know why the score is 72. It does not know the candidate. It compares a number to a number. That sounds cold because it is a rule, not a reader. The judgment was earlier, in the scoring against the job description. If that scoring used a matrix on the job, the matrix did the work. If the job had no matrix, HireQwik judged whether the history fit this role, and a keyword pass was only the fallback. How that match is built is the part you must trust before you trust 60. A precise cut on a vague ruler is still vague.

Failed files never enter this comparison. No score, no mark. Lowering the cut will not surface them. They belong to the file-failure list, not to a more generous threshold.

Why I start the HireQwik cut at 60

Sixty is a default, written so a new job has a line instead of a blank. It is not a validated pass mark for every role in India. A fresher drive with a broad JD may put a crowd over 60 because many CVs share the same coursework words. A narrow specialist JD may put almost nobody over 60 because the matrix is strict. Both outcomes can be correct. Neither outcome means you should immediately edit the number to force a prettier count.

I look at the shape first. Sort by score. If the marks are a cliff, with a pack at 80 and a pack at 30, the cut can sit in the gap and the list will be stable. If the marks are a smear from 50 to 70, moving from 60 to 65 changes many names and changes very little substance. That smear is the old compression problem, described in why scores sit in a narrow band. Fix the ruler. Do not negotiate with the smear using the slider.

On a campus morning, the practical read is in the drive-day folder piece. There, 60 is a first filter so the panel is not holding the zip. It is still not a badge that means “hire.”

What changes when you move the HireQwik cut

Move the cut up, and some people lose the shortlist mark. Their score stays. Move it down, and some people gain the mark. Their score stays. The average on the job does not get recomputed from new readings. You are changing a label.

That is useful in a live meeting. A manager says the panel can see 15 people, not 46. You raise the cut until the shortlisted count is near 15, then you read the names that fell off the edge. If the person at 74 and the person at 76 are interchangeable on the reason text, the headcount trim is arbitrary and you should say so. If 76 has the skill the matrix weighted and 74 does not, the trim matches the ruler. The conversation takes ten minutes. Rescoring the folder would pretend the ruler changed when only the guest list did.

I write the cut I used into the note that goes with the export. “Shortlisted at 70 on 5 October against the backend JD” is a sentence a later recruiter can audit. “The AI shortlisted the best” is not.

A separate decision column is the second mark. Approve-above and reject-below are actions you take. They are not the same as the automatic shortlist flag. You can shortlist at 60 and still reject a 90 whose resume, when opened, is for a different profession. You can keep a 58 you want a human to see. The flag is the rule. The decision column is you. Both should be visible when you hand the list to someone else.

Do not mix this cut up with other HireQwik bands

The Needs Review inbox, fed by the sheet trigger, has four resume-match bands: Strong Go at 90 to 100, Go at 75 to 89, Maybe at 60 to 74, and Reject at 0 to 59. Those labels are a filter on a different queue. How those bands are set before a voice invite belongs to that queue.

If you tell a manager that everyone over 60 on a file upload is a “Go,” you borrowed a word from the wrong screen. On the Shortlist screen they are simply shortlisted at your cut. A 61 and a 96 wear the same flag until you sort by score or raise the cut. The flag is binary. The score is not. Sort by the score when you talk about quality. Use the flag when you talk about who is in the working set.

Voice-call labels are a third language: Strong Go, Go, On Hold, No Go after a conversation. A resume score of 80 does not award those words. Resume match versus a later interview verdict is the gap. Keep it open in the briefing or people will treat the upload list as a hire list.

How I set the HireQwik cut without guessing

I use a short procedure after the first successful run on a role.

1. Leave the default at 60 until you have read five reasons above it and five below it.
2. If the reasons above the line match the JD and the reasons below do not, keep 60 and export.
3. If too many weak reasons sit above the line, raise the cut in steps of 5 and re-read only the new edge.
4. If strong reasons sit just below the line, lower the cut the same way, or fix a matrix that under-weighted that evidence.
5. Record the final number next to the job name in whatever note the panel will see.

Step 4 is the one teams skip. They lower the cut because the list feels small. Sometimes the list is small because the folder was off-role. One opening pulling the wrong resumes is that situation. A lower cut then fills the panel with people the ruler already tried to set aside.

Twelve scores on a HireQwik cut, with no second reading

Here is a toy batch so the arithmetic is visible. These are not a real drive. They are the kind of spread I see when a rubric is doing its job.

At 60, nine people are marked. At 75, four people are marked. The 74 did not become a worse candidate. The 58 did not become a better one. If a manager wants about four conversations, 75 matches that room. I would still open 74 and 76 side by side before I treat the line as wisdom. If their reasons are twins, say that the room size did the cutting. If 76 shows the skill the matrix paid for and 74 does not, the line agrees with the ruler.

Notice what the table cannot do. It cannot show a photo-PDF that never scored. It cannot show a hash-skipped twin. Those names never entered this dozen. A threshold discussion that starts from “we had 200 files and 9 shortlisted” without the failure count will over-praise the cut.

The review, including the cut and the filter, sits on the recruiter home. Changing the number there updates the marks. It does not send interviews. If you want conversations next, that is a separate handoff, described in from shortlist to voice screening.

Ladders measured resume glances at about seven seconds, which is the human version of a crude cut: almost everyone is glanced at, almost no one is judged. A threshold is only better if the score under it is real. SHRM’s 2026 survey puts AI inside HR at 39 percent of organisations. The teams who can explain their cut will be the ones who can defend a name that was left off.

What I say in the room when someone hates the HireQwik line

The fight is rarely about the arithmetic. Someone wanted a name that fell under the line, or they wanted a shorter list than the line produced. I use the same four sentences so the meeting does not wander into “the AI is unfair.”

First sentence: the number on the row did not change when we moved the line. A 72 is still a 72. We only changed who wears the shortlist mark. If they believe 72 is the wrong judgment, that is a complaint about the job’s ruler, and we should open the reason and the CV. Sliding the line will not rewrite the reason.

Second sentence: a score sitting exactly on the line is included. The rule is at or above, not above. I say this because people who grew up with school exams assume 60 means “just failed.” On this screen 60 means “just included,” if 60 is the cut you saved. If you wanted 60 excluded, the cut you save is 61. Do not explain it after the export has already gone out.

Third sentence: unread files are not under the line. They have no number. If the room asks “did we reject the scans,” the answer is no, we could not read them. Putting those names in the same breath as a 40 teaches the panel that every absence is a low score. It is not. Keep a separate count for files that never finished.

Fourth sentence: my approve or my reject is not the line. I can keep someone under the cut because I read the CV and disagree, and I can drop someone over the cut for the same reason. When I do that, I type the reason in the note. A silent exception becomes folklore by the next drive. “We always rescue people just under 60” is how the line stops meaning anything.

Then I stop talking and open two rows: one just over, one just under. If the room cannot tell them apart from the written reasons, I admit the line is a headcount tool today and I say that out loud. Pretending a fuzzy edge is a sharp insight is worse than the fuzzy edge. If the reasons are clearly different, the line can stay and the disappointed manager still has something to read.

I write the four sentences on the agenda, not as a speech. The chair can see them. When the conversation loops back to “just add ten more,” I point at sentence one. Adding ten more is allowed. It is a new line, or it is ten named exceptions. It is not a claim that the old scores were updated.

After the meeting I export once. A second export with a quiet new cut, mailed only to the person who complained, is how two lists enter the panel. One list. One number in the header. Anyone who wants a different number asks before the export, not after the interviews are booked.

If you want help choosing the line for a live role, send the job and a finished export. The useful part is the names sitting on the edge, not a prettier count.