Most interview processes select for people who are good at interviews, not people who do the job well.
The interview process is the highest-leverage hiring decision most startups get wrong. Bad processes select for people who are good at interviewing (polished, articulate, confident on the spot) rather than people who are good at the actual job (careful, thorough, effective on real work). The fix is not more interviews — it's the right interviews, in the right sequence, with structured evaluation rubrics that measure signal rather than gut feel.
Stage 1 (recruiter screen, 20 min): motivation, basic fit, salary alignment, availability. Reject 40-60% here. Stage 2 (hiring manager screen, 30-45 min): role-specific fit, experience relevance, initial signal on skills. Reject another 30-50%. Stage 3 (work sample or paid trial, 2-8 hours): the candidate does the actual job or something very close. Highest-signal stage — most processes skip or shortcut this. Stage 4 (team interviews, 3-5 hours total): technical depth, collaboration, values fit, executive alignment. Stage 5 (references, 3-5 calls): validate stories and dig into concerns.
A 4-hour work sample predicts on-the-job performance ~3-5x better than any interview. For engineers: a scoped take-home or a paid half-day of pairing on real code. For designers: a portfolio deep-dive plus a 2-hour design exercise on a bounded problem. For sales: a mock discovery call on your real product with your team playing a prospect. For PMs: a written strategy exercise on a real product decision. Companies that skip work samples over-index on presentation skills.
Structured interviews (same questions, same rubric, same evaluation criteria for every candidate) predict performance ~2x better than unstructured. Unstructured interviews are essentially expensive chats. Build a scorecard per role: 4-6 core competencies, each with a 1-4 rubric (1 = fails, 4 = exceeds bar). Each interviewer scores 2-3 competencies, not the whole role. Ban gut-feel scoring ("I liked them") from decision meetings — force interviewers to reference specific evidence tied to specific rubric items.
Interview loops without structured debriefs produce hiring decisions driven by the loudest interviewer. Structure: 30-min debrief within 24 hours of loop completion. Each interviewer shares their score + evidence before hearing others (prevents groupthink). Then discuss disagreements, focused on evidence not opinion. Decision framework: unanimous strong yes = hire. Any strong no with credible evidence = don't hire. Mixed = deeper reference dig or additional trial. Never override a strong no from a competent interviewer to "take a chance."
Most references are perfunctory. Great references dig. Structure: 30 min per reference call. Ask open-ended, non-leading questions: "What kind of work does [candidate] do best?" "Where would they need support?" "How did they handle [specific past situation]?" "Would you hire them again — and for what role specifically?" Back-channel references (people the candidate didn't list, sourced through mutual connections) usually produce the highest signal. This is legal and standard practice.
No work sample stage (relies on interview performance for skill assessment). Unstructured interviews ("tell me about yourself") that produce noise. No scorecards or rubrics (evaluation is vibes). Hiring on "we could always fire them if it doesn't work" (firing costs 3-6 months and damages team morale). Skipping references (misses signal that could have prevented the hire). Loops that drag on 4+ weeks (best candidates take other offers).
Investor directory · Fundraising library · Articles A–Z · Company funding database