A recruiter screen has a narrow job. Confirm the basics, find out what the candidate wants next, catch anything that could stop the process, and tell them enough about the role to decide whether to keep going. Then hand what you learned to whoever runs the next stage.

That last part is where the stage quietly fails. 31.2% of the 5.2 million captured candidate interviews in this corpus carried at least one attached scorecard, which is the industry's normal state. The pattern holds even when a decision depended on it. 41.9% of the 296,555 advances observed in the same corpus happened with nothing submitted behind them. Submission ran at 50.3% when Metaview generated the first draft of the scorecard, against 28.6% when the interviewer started from a blank template. Those two submission rates are counted on scorecards created, which is a separate base from the 31.2% coverage figure, and none of it was a controlled trial.

So plan the screen backwards, starting from the record it owes the next interviewer. That's the one part of this stage you can specify in advance, check afterwards, and fix when it comes back thin.

Step 1: decide what the screen owes the next interviewer

Before the call, read the application against the role's must-haves and write down two lists.

The first is the three points you need to clarify. Usually one of:

• A gap in experience
• An unclear responsibility
• A location or work-authorization requirement
• A claim that needs a concrete example

The second is the fields your scorecard template expects from this stage. If the scorecard template has a row for relevant experience and nothing in your screen produces evidence for it, that row comes back empty, or it comes back as an opinion. Both are useless to whoever runs the next interview, and you won't spot either one until the handoff has already happened.

Don't build a different interview for every candidate. The core stays the same for everyone applying to the same role. Your three clarification points are where you adapt to the person in front of you.

Metaview Application Review drafts an Ideal Candidate Profile from the job description and the intake call, and the recruiter reviews and approves that profile before anything gets measured against it. Its written fit reasoning is a starting point for your three clarification points, and you still write them yourself. The fraud and application automation indicators are there to make a recruiter look twice. The recruiter decides who moves forward.

Step 2: set the agenda and defend the time

Tell the candidate how the call will run before you start. It makes the time manageable, and it guarantees them a real chance to ask questions.

You don't need to deliver that word for word. Keep the structure and use your own language.

The agenda exists to protect the time. Interviews run long constantly: 54.3% of 174,558 captured interviews ran past their scheduled end, and the ones that did ran a median of 11.9 minutes over. The data records the overrun without recording what got squeezed to pay for it, so read it as an argument for setting an agenda. What the overrun cost, the data doesn't say. In practice the minutes come out of the end of the call, which is where the candidate's questions and the next steps sit.

The agenda also gives you a polite way to redirect a long answer. You can point back at the time you set aside for their questions instead of cutting someone off with no explanation.

Step 3: ask a consistent core and know what each answer is for

Six areas cover what most recruiter screens have to establish: interest, relevant evidence, expectations, possible blockers, compensation, and timing.

They're areas to cover, and they aren't a scoring system. Adapt the wording to the role, the country, and your process. Ask the same core of everyone applying to the same role, then follow up on what each person actually said.

The right-hand column is the part that usually goes missing: what each answer should leave behind.

Your follow-up depends on what the candidate has already said. When an answer is vague, ask for the action, the reasoning, and what happened afterwards. Stop once the field it feeds is full.

Leave the rest of the assessment to the stages built for it. Technical depth, work samples, and detailed role-specific evaluation belong in later stages with the person qualified to run them. Trying to cover everything is how a screen runs long and still leaves the next interviewer with nothing specific to read.

Step 4: give the candidate the information they came for

The screen is also the candidate's first proper look at the job. They arrive with a short list of things they want answered, and how you handle that list is a large part of the candidate experience at this stage.

Compensation came up in 72.4% of the captured conversations where candidate questions were detected, growth and progression in 62.2%, team and culture in 45.8%, and remote or flexible working in 43.2%. Those shares come from topic detection on the conversations where at least one of these themes was found, so they describe what gets discussed. They don't rank what every candidate cares about, and the mix shifts by seniority.

Answer as you go instead of saving a pitch for the last five minutes. When something the candidate says connects to the role, say so out loud:

"What you said about owning a product area from discovery through launch is relevant here. That would be a large part of this role."

Cover the responsibilities, the reporting line, why the role is open, the working model, the compensation range if you're able to give it, and anything else that would change the candidate's decision. Then leave time for the questions your answers didn't already cover.

Step 5: close with next steps you can meet

Before you end the call, say:

• what the next stage is;
• who the candidate will meet;
• what that stage assesses;
• when they should expect to hear back;
• roughly how long the whole process takes.

Only promise a date you can hold. If the process changes, tell the candidate instead of letting the date pass in silence.

Write the commitment into the notes. It's the first thing a teammate covering the role goes looking for, and nobody can reconstruct it later. Your interview notes template should have a line for it.

Step 6: write the record while the conversation is still there

The screen shouldn't end with a recruiter rebuilding it from memory that evening. Recall is the weakest link in the whole stage, and it's the part every later decision reads from.

Metaview captures the screen and generates a first draft of the structured notes and the objective scorecard fields against your team's template. It leaves the vague and subjective fields blank, because judging those is the interviewer's job.

The recruiter then:

1. checks the draft against what the candidate said;
2. corrects anything wrong or missing;
3. adds their own assessment against the team's rubric;
4. submits or syncs the approved feedback to the ATS.

The usage data shows how much work that review does. Recruiters accepted 23.9% of 728,375 AI-suggested scorecard fields as written, edited 12.0% of them first, and dismissed 64.1% empty. Most of what the draft offers gets thrown out by the person who was actually in the room, which is the arrangement working as designed. Metaview doesn't score the candidate and doesn't make the hiring decision.

The submission gap from the top of this article belongs here too, with its caveats attached. Both rates count scorecards created and neither counts interviews held, the samples are query-capped at 93,502 and 26,498, and nobody randomized who received a generated draft. The comparison is observational and does not show why the gap exists.

Step 7: measure the handoff rather than the verdict

The obvious way to grade a recruiter screen is to count how many of its candidates the hiring manager agreed to see. That's a real measure of whether the two sides are working to the same standard, and it's the first number most recruiting leaders reach for.

That's a practitioner's view, and a sensible one. It's also easy to over-read, in two specific ways.

The base rate is low everywhere. 20.8% of 1,425,761 candidate and job pairs with a captured first interview show up in a later round at all. The rate climbs after that, to 35.9% from round two to round three and 43.7% from round three to round four. A screen-stage conversion figure that looks alarming next to a later-round number can be unremarkable next to the first one, so compare a stage to itself.

Agreement measures alignment, and only alignment. Among 139,336 candidates with both an early-round and a later-round scorecard, the early recommendation matched the final one 54.4% of the time. That doesn't mean early recommendations are wrong half the time. Later rounds see different evidence, ask different questions, and are supposed to change some minds. What it does mean is that an early verdict is a weak guide to the later one. A verdict is the most perishable thing a screen can produce. The evidence underneath it survives into every round that follows.

None of it tells you whether the screen was right. The corpus holds interviews, scorecards, and movement between stages. It holds no data about anything that happened after a hire, so nothing here can be read as a statement about how well someone did the job. Every measure in this section is the process describing itself.

That limit is real and worth stating plainly. What's left is still the useful half, because you can check it: did the screen produce a record, and did that record hold what the next interviewer needed?

So calibrate on the handoff. Review conversion by role, over enough candidates to mean anything, and read it next to:

• the hiring manager's rejection reasons, in their own words;
• which competencies came back missing or unclear;
• candidate withdrawals after the screen;
• the gap between the screen and the next decision;
• differences by role, source, recruiter, and hiring manager.

A low conversion rate is a question you still have to go and answer. Maybe the two sides are working to different criteria. It might equally be a hard market, a role that changed after kickoff, thin sourcing, or a sample too small to say anything at all. Open the interviews before you decide what happened.

One caveat covers every figure in this article. The corpus doesn't classify recruiter screens cleanly: only 7,811 captured conversations carry a recruiter-screen label, so these rates are measured across captured interviews of all kinds. They describe the environment a screen sits in, and no figure here is a measurement of screens on their own.

Metaview Reports can query your own interview data by topic, so you can pull the screens where compensation or remote working came up and read what was actually said in them.