Recruiting interviews are moving towards cases, and the reason is simple: a good recruiter and a mediocre one describe their process in almost identical language. The difference shows up the moment there are numbers on the table.
Below is a realistic brief, worked all the way to a plan. The interesting part is two-thirds of the way down, where the arithmetic says the obvious plan is the wrong one.
The brief
We need six backend engineers hired in two quarters. You're the only recruiter on it. Here's last year's funnel. Walk us through your plan.
Stage Conversion Outreach → reply 18% Reply → recruiter screen 55% Screen → hiring manager interview 40% HM interview → onsite 50% Onsite → offer 30% Offer → accept 70% Onsite loops are five interviewers, one hour each.
Work it backwards
The first move, and the one that separates a plan from an intention: start at six hires and divide upwards.
| Stage | Needed |
|---|---|
| Accepts | 6 |
| Offers | 9 |
| Onsites | 30 |
| HM interviews | 60 |
| Recruiter screens | 150 |
| Replies | 273 |
| Outreach messages | ~1,500 |
Saying it out loud: "So the plan is 1,500 pieces of outreach in 26 weeks. That's about 58 a week, every week, on top of running 150 screens — roughly six screens a week as well. I want to put that number on the table first, because it's the thing that decides whether this plan is real, and it's the number that usually goes unsaid until the quarter is half gone."
That's already a stronger answer than most, and it's just arithmetic. But it's not the answer.
The number that's actually wrong
Look at the onsite line. Thirty onsites produce nine offers — twenty-one loops that end in nothing.
Twenty-one loops at five interviewers an hour each is 105 hours of engineering time spent on candidates who were never going to get an offer. That's most of a working month, taken from the same team that's short-staffed enough to need six hires.
Saying it out loud: "Before I ask for anything, I'd want to say what the current funnel costs. A 30% onsite-to-offer rate means seven out of ten onsites are a rejection that took five engineers an hour each. That's 105 hours over this plan. The team we're hiring for is paying for the inefficiency of the process that's meant to help them."
This is the moment the case turns. Everything up to here is competent. This is the part that sounds like someone who's run a funnel rather than filled one.
The instinct being tested is whether you optimise the stage you control or the stage that's broken. A recruiter's natural lever is the top of the funnel — more outreach, better messages, new channels. It's the part that belongs to them. But the arithmetic says the expensive failure is at the bottom, in a stage owned by the hiring team. Candidates who only propose working harder at the top are demonstrating exactly the boundary that limits them.
What a fix is worth
If onsite-to-offer moved from 30% to 45% — not a heroic number, it's what calibration typically buys — the whole plan changes shape.
| Now | At 45% | |
|---|---|---|
| Onsites | 30 | 20 |
| HM interviews | 60 | 40 |
| Screens | 150 | 100 |
| Outreach | ~1,500 | ~1,010 |
| Outreach per week | 58 | 39 |
A third less work at every stage, for the same six hires. And 50 fewer wasted engineering hours.
"Which is why my first two weeks wouldn't be sourcing. They'd be finding out why seven in ten onsites fail. That's a question I can answer from the data we already have — I'd look at whether the rejections cluster on one interviewer, one stage of the loop, or one profile of candidate. Those are three different problems."
Three hypotheses, named, with what each would imply:
| If the rejections cluster on… | Then the problem is | And the fix is |
|---|---|---|
| One or two interviewers | Calibration, not candidates | Debrief structure, shadowing, sometimes removing someone from loops |
| One stage (say, system design) | The bar for that stage isn't shared | Agreeing what "pass" means before the next loop |
| One profile we keep sending | My screen is miscalibrated | That one's mine to fix, and I'd want it caught in week two rather than month four |
Volunteering the third possibility — that the miscalibration might be yours — is worth a lot. It's also often true.
The plan
Three parts, in the order the hiring manager can act on them.
Weeks 1–2. Diagnose the onsite rate from existing data and run an intake that produces a real differentiator rather than a requirements list. Sourcing starts in parallel but deliberately small — about 20 messages a week — because outreach sent before the profile is right is outreach wasted twice.
Weeks 3–26. Full volume, at whatever the diagnosis makes it: 39 a week if the loop gets fixed, 58 if it doesn't. Three profiles to the hiring manager every week with a standing request to reject at least one, because a week where they like all three means I've pitched too safe and I want to know immediately.
Throughout. One number reported weekly, and it isn't outreach volume — it's onsite-to-offer. Vanity metrics are the ones that measure my effort; this one measures whether the process works.
More at the top
“I'd increase outreach volume, add two new channels, improve the message templates and build a referral push to get the pipeline up.”
Every item is real work and none of it touches the stage that's failing. Six months later the funnel has more people entering it and the same seven-in-ten onsite rejection, and the recruiter is exhausted and behind.
Fix the bottom, then fill the top
“Two weeks on why onsites fail, then volume sized to the answer. If we get to 45%, this becomes a 39-a-week job instead of a 58-a-week one.”
The same six hires, a third less work, and fifty engineering hours given back. It also gives the hiring manager something to do, which turns hiring into a shared problem rather than a request queue.
The honesty moment
Somewhere in this case you should say whether six is achievable. Most candidates avoid it, and avoiding it is itself an answer.
"Can I hit six? At the current conversion, it needs 58 outreach messages and six screens every week for 26 weeks with no gaps — no holiday, no other req landing on me, no quarter where the team is too busy to interview. That's not a plan, that's a plan with no slack in it. I'd forecast four to five at the current funnel and six if the onsite rate moves. I'd rather tell you that in week one than in month five."
Every hiring manager has been told a number that didn't arrive. The recruiter who calibrates early is the one they trust with the next req.
What a weaker plan leaves out
Three things in the answer above are easy to miss, because their absence is silent.
The weekly number. Without it, "six hires in two quarters" is an intention that nobody has costed. With it, everyone in the room now knows what was actually agreed to — and if it's undeliverable, they know that today rather than in month five.
The engineering hours. Recruiters who cost their funnel in their own currency — outreach sent, screens run, submissions made — stay a service function, because those numbers only measure their own effort. Converting the waste into 105 hours of the hiring team's time is what makes fixing the loop a shared problem instead of a favour being asked.
The forecast. "Four to five, and six if the onsite rate moves" is the only sentence in the plan that can cost you something today. It's also the one that earns the right to be believed the next time you give a number.
Practising this
Take a req you've actually run and work it backwards from the hires to a weekly outreach number. Most recruiters have never done this arithmetic explicitly, and the number is usually surprising — which is the point, because it's the number that determines whether the commitment you made was real.
Then find your worst conversion and ask who owns it. If the answer is "not me", you've found the conversation this case is really about.
The behavioural and process side of a recruiting loop is a different set of questions, and it's here: Recruiter Interview Questions. If the role is engineering-focused and the loop includes a technical credibility check, that half is covered here: the technical recruiter interview.
Practical target: work the case above out loud and check one thing — did you say a number that made the target sound harder? Almost every candidate finds the 1,500. Far fewer volunteer that six might be four, and that sentence is the one that predicts how you'd behave in month five, which is what the panel is actually trying to find out.
