You send eight CVs. Eight come back.

"Not quite right."

You send six more. Six come back. Still not quite right.

By round three you've decided your hiring manager is impossible. Your hiring manager has decided you don't understand the team.

You're both wrong.

This was never a taste problem. It's a scoring problem. You're grading against the JD. Your hiring manager is grading against a picture in their head — one that has never been written down anywhere.

And that picture updates every time they open another CV.

Why every shortlist round pushes you further apart

Intake happens once. The criteria keep living, quietly, inside someone else's head.

The bar moves and nobody announces it

Day one: "Senior Backend Java, five years, microservices."

Week two, after twelve CVs: "Fintech background would help."

Week three: "Honestly, the real question is whether they survive on-call."

Nobody's lying. Real criteria only surface when someone looks at real people.

Meanwhile you're still sourcing against the week-one JD.

The must-haves live in your notes and nowhere else

You took excellent intake notes. In your Notion.

Your hiring manager kept a copy too. In their head.

Those two versions have never sat side by side. So when you disagree, there's nothing to check against except memory. And memory always sides with whoever is more annoyed.

"Not quite right" isn't feedback

• Failed on which skill?
• Level, or domain?
• Comp, or profile?

No answer means no correction. You just send more CVs and watch for a reaction.

You're not screening. You're playing Wordle with someone else's word, and the only clue you ever get back is "nope".

Every interviewer builds a private rubric

Tech lead grades architecture. Team lead grades communication. Hiring manager grades grit.

Three people, three exams.

One candidate, three verdicts.

You get the blame for "not finding anyone".

Three rounds and fourteen days later, the strongest person you sent has signed elsewhere.

The 3 Pins: get both of you grading the same test

You don't need another meeting. You need a pinned sheet — one page both sides can see and both sides can edit.

A Google Sheet does it. Three steps.

Pin 1 — Pin the criteria

Five must-haves. Hard cap.

Each one has to be written as something you can verify from a CV or from a single interview question.

• Weak: "strong systems thinking"
• Pinned: "designed a system spanning 3+ services, can name the trade-offs"

If a criterion can't be verified, it isn't a must-have. It's a feeling.

Pin 2 — Pin the weights

Split 100 points across those five. The hiring manager splits them, not you.

Forcing the split is where the truth shows up. A surprising number of "must-haves" quietly demote themselves to nice-to-have the moment someone has to give them 5 points.

You stop arguing about taste. You start arguing about numbers. Much easier argument.

Pin 3 — Pin the evidence

One rule: every rejection names the criterion that failed, and where it failed.

If the reason isn't in the pinned five, it's a new criterion. Pin it. Re-split the weights.

This is the part that does the work. The sheet isn't a document that dies after intake. It gets updated every round, and every update narrows the gap between the two of you.

By round three, neither side is guessing anymore.

The Knoot Angle

Recruiter–hiring manager drift is rarely a relationship problem. It's an evidence problem: two people holding two versions of the criteria, neither in a form you can put side by side.

That's the job JD Analyzer does inside Knoot. Drop the JD in, and it breaks the text into discrete criteria, sorts them into must-have, nice-to-have and bonus, and attaches a weight to each. You keep full editing rights: pull a weight up, demote a must-have, add the criterion that only appeared after the first interview.

That same criteria set feeds your next screening round, so you're never filtering blind.

What you end up with isn't an analysis to read. It's a pinned sheet two people can argue over productively.

A JD is a wish list. Pin it down and it becomes a scorecard.