Take-home assignments are broken. Here's the data.

Take-home assignments were meant to fix the whiteboard-interview problem — replace artificial pressure with realistic, unhurried work. In practice, they've created three new problems.

Problem one: time

A well-designed take-home takes candidates 4–8 hours. Strong candidates — the ones with other offers on the table — increasingly decline to spend an unpaid workday on a maybe. You lose them before you ever see their work.

Problem two: grading debt

Someone on your team has to read every submission carefully enough to score it fairly. That review backlog is where most take-home pipelines quietly die — assignments sit ungraded for a week while the candidate accepts somewhere else.

Problem three: authorship

With AI coding assistants now mainstream, a take-home no longer reliably tells you who wrote the code, or how much help they had. That uncertainty erodes trust in the whole exercise for both the panel and the candidate.

What actually works

A live, proctored session — 60 to 90 minutes, scored the moment it ends — solves all three. Candidates get a fast, respectful process. You get a graded result before your next panel call. And you can see how someone works, not just what they eventually submit.