Gem’s 2026 Recruiting Benchmarks Report, built on more than 165 million applications and 1.2 million hires between June 2021 and May 2025, found that interviews per hire are up 33% since 2021 while only 8% of applicants get past the initial screen. Ashby’s State of Startup Hiring 2026, covering 1,200 venture-backed startups and 32,000 hires, puts the bill for a startup under 25 people at roughly 21 interviewer hours per technical hire. If you are about to run a six-week internship, that is the process you are about to point at it — and it costs more than the internship does.
Key takeaways
- The industry-standard loop grew to de-risk permanent hires. An internship is already reversible, so you are paying for that certainty twice.
- Under 25 employees, Ashby’s data shows ~18 candidates interviewed and ~21 interviewer hours for one technical hire — at a 12-person company that is most of an engineer’s week.
- Startups without a recruiter take 62 days to hire against 42 with one. Nobody is coming to fix that; a shorter loop is the fix.
- Ten days is enough: a walkthrough of real code, a scoped paid task, and a decision made against criteria written down in advance.
The loop you inherited was built for a different decision
Nobody designed the modern five-stage interview; it accreted. Each round was added after a specific bad hire, and it stayed because removing a stage feels like lowering a standard. Gem’s benchmarks capture the drift: interviews per hire up a third since 2021, recruiters carrying 93% more applications with 14% fewer people, and a hire rate of roughly one per 200 applications. Under that pressure, adding a filter is the obvious move and subtracting one never is.
But every one of those stages is priced against the cost of getting a permanent hire wrong: notice periods, ramp time, the months of ambiguity before anyone admits it is not working. An internship has none of that structure. It is short, scoped, paid, and ends on a date you already agreed. The question at the top of the funnel is not “is this person right for the next four years” — it is “is this person worth six weeks of a defined project.” Running the expensive instrument on the cheap question costs you both ways: your engineers lose the hours, and the candidate spends three weeks in a queue.
Do the arithmetic before you schedule round three
Ashby’s figures are worth sitting with. For a technical hire, startups under 25 employees interview about 18 candidates and spend around 21 interviewer hours; at 100–300 employees that rises to roughly 29. And the smallest startups receive 298 inbound applications per hire, with remote roles pulling 42% more applications than in-office ones.
Now put that against the thing you are hiring for. Twenty-one interviewer hours is most of a week from one of your engineers — the same engineer who will supervise the intern. Spend it at the front and you arrive at day one already behind, having paid for evaluation rather than work. The alternative is not to evaluate less. It is to move the evaluation inside the internship, where the person is doing real tasks, being paid for them, and the evidence is a merged pull request rather than a whiteboard performance.
Speed is the advantage you actually have
A twelve-person startup cannot outbid a bank, cannot outrank a brand a student’s parents have heard of, and has no campus relationship. What it has is that the person making the decision is in the room. Ashby’s data quantifies the flip side: startups running a role without recruiter involvement take 62 days to hire against 42 days with one. Without a recruiter the loop does not get shorter by default — it gets longer, because scheduling and follow-up fall to whoever has time, which is nobody. CoderPad’s State of Tech Hiring 2026 found 44% of hiring leaders naming a cut in time-to-hire as a priority, which is another way of saying most of them know.
For Indian Tier-1 students the delay compounds in a way US founders tend to miss. Availability is set by an academic calendar, not a notice period: mid-semester exams, end-semester exams and the vacation window are fixed months in advance. Three weeks of silence does not just cost goodwill — it can move the start date past the only window in which the person could have worked at all. The calendar is laid out in plan from the calendar, not the offer letter.
What to cut, and what never to cut
Cut anything that produces an opinion you already hold. The second culture conversation, the round where a third engineer re-asks the first engineer’s questions, the take-home that duplicates a project already sitting in the candidate’s GitHub — hours, no information. Cut the unpaid multi-day assignment in particular: it filters for free time rather than ability.
Never cut two things. The first is a live walkthrough of code the candidate actually wrote, which is the cheapest honest signal available and is now the only reliable way to tell authorship from assembly. The second is written criteria agreed before the first call — speed without a rubric is not efficiency, it is just a faster way to hire whoever you liked. That is the case we made in write the intern’s rubric before week one.
The ten-day loop
- Day 0 — write the rubric and the project. Four or five criteria and the actual first task, before you read a single application. Twenty minutes, and it decides everything that follows.
- Days 1–2 — screen on artefacts, not CVs. Repository, commit history, a deployed link. Ten minutes per candidate.
- Day 3 — one 30-minute walkthrough. They share a screen and explain something they built: why that approach, what broke, what they would change. No algorithm puzzle.
- Day 4 — send one scoped paid task. Real, small, four hours at most, paid at whatever rate you have set. Give a deadline of three days, not three hours — you are testing work, not availability.
- Days 5–7 — review it like a pull request. Leave comments. How they respond to review is data you cannot get any other way, and it predicts the next six weeks better than the code does.
- Day 8 — a 20-minute founder conversation. Hours, timezone overlap, exam dates, what they want out of it. Logistics and fit, not another technical screen.
- Day 9 — score against the rubric in writing. Each interviewer scores independently before anyone talks. Discussion after scores, never before.
- Day 10 — decide and say so. Offer, or a clear no with one sentence of reasoning. A fast no is worth more to a student than a slow maybe, and it is the thing they tell their juniors about.
Where the hours should go instead
Every hour taken out of the front of the process is an hour available for the part that determines the outcome: scoping the first project properly, reviewing the first week’s pull requests quickly, and deciding early enough to make an offer while the person is still deciding. NACE’s 2026 Internship & Co-op Survey has intern conversion at 63.1%, a five-year high, with 88.3% of offers accepted. The teams getting that result are not the ones running the longest interviews; they are the ones who spent the time after the hire rather than before it.
That is the model Tierones runs on: students from IITs, NITs and IIITs matched to startups on commits, shipped projects and a code walkthrough instead of a CV, so the first conversation starts from evidence. The longer argument for running the internship itself as the evaluation is in run a six-week work trial. Stipends stay yours to set — we do not set, suggest or cap them.
FAQ
How long should it take to hire an intern?
About ten days from first contact to decision. The full-time loops most startups copy keep growing — Gem’s benchmarks show interviews per hire up 33% since 2021 across 165 million applications. That cost is defensible for a permanent hire with a notice period. It is not defensible for a short, paid, reversible internship where the work itself is the evidence.
How many interview rounds does an intern really need?
Two conversations and one piece of real work. Ashby’s data has startups under 25 people spending about 21 interviewer hours per technical hire, most of it buying certainty about an irreversible decision. An internship is already reversible and already time-boxed, so keep the walkthrough, keep the paid task, and drop the rest.
Is moving fast just a way of lowering the bar?
Only if you skip the rubric. Speed should come from removing duplicate stages, not evidence — a walkthrough of real code and a paid task produce more signal per hour than a fourth culture call. Write the criteria down before the first conversation, and the fast loop is the more rigorous one.
Tierones matches students from IITs, NITs and IIITs with startups that hire on proof of work. Tell us what you need — we’ll show you proof, not CVs. Or write to hire@tierones.io.