Today’s target: proxy metrics in recruiting — specifically, the resume as a predictor of technical quality.
Resumes: The Original Bad KPI
Recruiting teams love to talk about “data-driven hiring,” but the moment you look closely, the whole process runs on proxies:
- pedigree
- brand names
- buzzwords
- job titles that mean everything and nothing
The resume is treated like a KPI for competence.
In reality, it’s more like a glossy dashboard hiding the fact that the underlying data is garbage.
And we all know what happens when Garbage is in – right?
A resume reliably measures exactly three things:
- How someone wants to be perceived (or paying someone that understands it)
- Their tolerance for corporate jargon
- Whether they understand margins and whitespace
It does not measure:
- technical judgment
- debugging instincts
- business reasoning
- the ability to say, “No, your requirement is impossible — pick one constraint and try again”
But because companies need something — anything — to filter hundreds of applicants, they cling to the resume as if it’s a meaningful signal rather than a noisy reflection of who had access to good storytelling skills.
Jeff Bezos Already Explained This (Accidentally)
On Lex Fridman’s podcast, Jeff Bezos talked about the difference between inputs and outputs — you can’t control outcomes directly; you control the process, the work, the actions that lead to those outcomes.
Recruiting flips this logic on its head.
Instead of measuring the inputs (how candidates think, solve, question, debug, and deliver), companies obsess over output-looking artifacts — the resume, the job title, the brand names — as if these things cause competence rather than merely signal it.
It’s a fundamental confusion of correlation and causation, and worse: a lazy one.
Proxy worship feels analytical.
It feels objective.
But it’s analysis paralysis.
The corporate kind (not the good kind, i.e. me)
Hiding under the comfort blanket of “structured hiring.”
What You Lose When You Hire the PDF
When you optimize for:
- Pseudo polished storytelling
- Inflation-proof adjectives
- LinkedIn theater
…you get hires optimized for exactly those traits.
People who look strong on paper but disappear when faced with a messy dataset or a frustrated stakeholder. People who can answer textbook questions but panic when the numbers don’t make sense.
Or not, I mean, it is possible that the talented people learnt how to speak the language of your ATS, and managed to get by with a great recommndation because they rock – but isn’t that seems all too accidental?
A Better Metric (Hint: It’s Not a Metric)
If you want to hire people who can do the work, test the work:
- a short but realistic SQL/business task
- a messy dataset and a prompt that isn’t perfectly formed
- a scenario where the candidate must reset assumptions
- a debugging exercise
- a prioritization exercise under uncertainty
You’re not looking for a perfect solution.
You’re looking for how they think.
Resumes measure storytelling.
Work samples measure capability.
Hiring based on resumes is like evaluating a surgeon based on their handwriting.
Final Thought
If resumes were good predictors of technical competence, every team would already be excellent.
Instead — well, you’ve met teams.
Stop treating the resume as a performance review.
Stop confusing polish with intelligence.
And stop pretending proxy metrics make hiring scientific. They don’t.
Hiring is messy, human, uncertain.
You reduce the uncertainty by watching how people work — not how they format their past.
(This post was edited using CahtGPT, so you can count ~18 em-dashes – do you know what that means? check it here)
