Google has more hiring data than almost anyone, so it's worth listening when they say the things most companies screen for are useless. Their former head of People Operations, Laszlo Bock, put it bluntly: GPAs are worthless for hiring, test scores are worthless, and brainteasers are a complete waste of time. None of them predict who does the job well. The thing that does, learning ability, is the thing most interviews barely test.
Key Takeaways
- Google found GPAs and test scores don't predict job performance, except a slight signal for brand-new grads (Laszlo Bock/Google).
- Brainteasers had essentially zero correlation with performance; they "make the interviewer feel smart" (Bock).
- Google's #1 criterion became general cognitive ability, the ability to learn on the fly (Bock).
- Screen for how someone thinks and learns, not where they went to school.
What Google Learned
Laszlo Bock, who ran People Operations at Google, said their data was unambiguous: "GPAs are worthless as a criteria for hiring, and test scores are worthless, no correlation at all except for brand-new college grads, where there's a slight correlation" (Bock/Google). Google had required transcripts and GPAs, then stopped, because they didn't predict anything. Brainteasers, the "how many golf balls fit in a bus" genre, fared no better: Bock called them a complete waste of time that mostly serve to make the interviewer feel clever (more on Google's data).
So what did predict performance? Google's number-one criterion became general cognitive ability, specifically the ability to learn on the fly and pull together disparate information, assessed through structured behavioral interviews they actually validated for predictiveness.
Why Pedigree Fails
Credentials measure where someone has been, not what they can do now. A degree from a famous school, a high GPA, a name-brand former employer, these are proxies, and weak ones, for the thing you actually care about: can this person solve your problems and grow into harder ones? Proxies are seductive because they're easy to screen on. They're also easy to have without the underlying ability, and easy to lack while having it in abundance. Plenty of excellent engineers have unremarkable resumes, and plenty of pedigreed candidates coast.
| Pedigree signal | What it actually predicts |
|---|---|
| GPA / test scores | Little to nothing (Google's data) |
| Brainteaser performance | Nothing; interviewer ego |
| Brand-name school/employer | Weak proxy, easy to have or lack |
| Learning ability + real problem-solving | Actual job performance |
A Concrete Version
Two candidates. One has a CS degree from a top school and a big-tech logo on the resume; the other is self-taught with a scrappy job history but a portfolio of real shipped systems. The pedigree screen fast-tracks the first and filters out the second. Then you actually test them on a realistic problem and on how they learn an unfamiliar tool, and the self-taught one runs circles around the pedigreed one, who leaned on the resume and stopped growing. Google's data says this happens constantly, which is why they stopped screening on the resume signals.
The Honest Counterpoint
Pedigree isn't zero information, and pretending a great engineer never comes from a top school is as silly as assuming they always do. A strong track record of shipping real systems is genuinely predictive, that's experience, not pedigree. And credentials can be a reasonable tiebreaker at the very margin. The mistake is using pedigree as a filter, screening people out on GPA or school before you've tested what they can do, because that's exactly where Google found you lose great people and keep coasters. Use pedigree as weak context, never as the gate.
How We Think About It
This is core to how we vet engineers: we test how people reason through real, unfamiliar problems rather than filtering on where they studied, which is also how we surface strong engineers from a much wider LATAM talent pool that pedigree-first screening would miss. It's the same logic behind structured interviews: measure the ability, not the proxy. See available engineers.
Frequently Asked Questions
Do GPAs predict job performance?
Not according to Google's data. Laszlo Bock said GPAs and test scores are essentially worthless for hiring, with only a slight signal for brand-new graduates.
Are brainteaser interview questions useful?
No. Google found brainteasers have essentially zero correlation with job performance. Bock said they mostly make the interviewer feel smart.
What should I screen for instead of pedigree?
The ability to learn on the fly and solve real, unfamiliar problems, assessed through structured, validated interviews. That predicts performance far better than credentials.
Does this mean pedigree is worthless?
Pedigree is weak context, not a filter. A track record of shipping real systems is genuinely predictive, but screening people out on GPA or school before testing ability costs you great candidates.
The Bottom Line
The company with the most hiring data concluded that GPAs, test scores, and brainteasers predict nothing about who does the job well. What predicts it is the ability to learn and solve real problems. Screen for that, use credentials as weak context at most, and you'll find strong engineers everyone else filtered out.
Roberto Espinoza is CEO of Ruzora, which helps US startups hire pre-vetted senior LATAM engineers in 72 hours. See available engineers.
