We reject 97 of every 100 engineers who apply. That number worries some clients until they hear the other one: 97% of the people who get through are still on the team six months later. The two are the same fact from different ends. A vetting bar that predicts retention is the only kind worth running.
Key Takeaways
- We pass roughly 3% of applicants. The bar is the product.
- Five stages, in order: screen, technical, English, culture, AI tooling.
- The strongest retention signal comes from the culture interview, not the coding test.
- Result: 97% six-month retention against an industry average we estimate near 85%.
Why Most Vetting Fails
Most providers check a résumé, run one coding screen, and call it done. That tells you someone can pass a coding screen. It says nothing about whether they'll still be shipping for you in a year, whether they communicate cleanly across a timezone, or whether they can handle a vague ticket without a manager standing over them. Those are the things that decide whether a hire works, and a single algorithm test misses all of them.
The Five Stages
Each stage cuts the pool. By the end, what's left is small and reliable.
| Stage | What it tests | Roughly who's left |
|---|---|---|
| 1. Screen | Experience, real projects, references | ~40% |
| 2. Technical | Systems design + live coding | ~15% |
| 3. English | C1+ in live conversation | ~9% |
| 4. Culture | Remote habits, communication, judgment | ~4% |
| 5. AI tooling | Copilot, Claude, Cursor in practice | ~3% |
The technical stage is live, not a take-home you can game. We watch how someone reasons through a problem they haven't seen, because that's the day job. The English check is a real conversation, not a written score, because a B2 reader who freezes on a call can't pair with your team.
The Stage That Predicts Retention
Here's the part that surprised us when we ran the numbers. The coding test screens out the people who can't do the work. But the culture interview is what predicts who stays. Communication quality, self-awareness, and how someone handles being wrong in front of others. Those traits track with retention far better than raw algorithm speed. It's the same thing strong engineering leaders already know about hiring, and it's why we built our process around retention rather than résumé keywords.
What This Means for You
When the bar is this high, you stop interviewing to filter and start interviewing to confirm. Most clients run one or two conversations with a shortlisted engineer and move to an offer, because the hard filtering already happened. That's the whole point of paying for vetting: you inherit the rejection work, not just the résumé.
Frequently Asked Questions
What percentage of applicants do you accept?
About 3%. Each of the five stages cuts the pool, and the last two (culture and AI tooling) reject more strong coders than people expect.
Why test AI tooling?
Because the job now includes it. An engineer who uses Copilot, Claude, and Cursor well ships more, and one who leans on them without judgment ships bugs. We test both the with and the without.
Can I see how a candidate did?
Yes. You get the assessment summary for every engineer on your shortlist, so the bar is visible, not a black box.
The Bottom Line
Vetting is only worth paying for if it predicts the thing you care about, which is whether the hire lasts. Test for communication and judgment, not just code, keep the bar high enough to hurt, and the retention number takes care of itself.
Roberto Espinoza is CEO of Ruzora, which helps US startups hire pre-vetted senior LATAM engineers in 72 hours. See available engineers.
