Evaluating FDE Project Portfolios: A 20-Minute Protocol
Updated · Tech checked
Review portfolios for delivery arcs, not tech stacks: 20 minutes per candidate across README quality, failure handling evidence, measurement honesty, and operations thinking - with a scoring rubric and reference-check prompts.
The short answer
Portfolios predict delivery better than resumes if reviewed for the right things. 20 focused minutes per candidate, same protocol every time.
The protocol
Minutes 0-5: README only. Can a stranger run it? Are assumptions listed? Is there a "with more time" section? (That last one predicts self-awareness and roadmap discipline.)
Minutes 5-12: Code dive on one file they're proud of. Error handling present? Tests for ugly cases? Naming a stranger could maintain? You're sampling craft, not auditing everything.
Minutes 12-17: Evidence of the arc. Is there a problem statement with a number? A measurement of outcome? A deployment story (docs, rollback mention)? The portfolio format guide defines the shape.
Minutes 17-20: Security scan. Any real customer data (fail), secrets committed (fail), ACL thinking in AI projects (bonus)?
Scoring
| Signal | Weight |
|---|---|
| Working whole (runs end-to-end) | ×3 |
| Failure handling visible | ×3 |
| Measurement honesty (incl. negatives) | ×2 |
| Docs/runbook quality | ×2 |
| Security instincts | ×2 |
| Framework trendiness | ×0 |
Red flags
- "Demo purposes only" everywhere with no production notes.
- Portfolios of identical CRUD clones (depth zero).
- Dashboard-only projects: visualization without pipeline integrity.
Reference-check prompts that pair well
"Walk me through how they handled the last production incident you saw them own" and "Did their estimates hold?" - both test the same dimensions the portfolio claims.
Continue: Hiring playbook · Scorecard