The thing most hiring teams call "preparing for the follow-up interview" is ten minutes of scrolling the submission on your phone while walking from the kitchen back to your desk. Maybe fifteen minutes if you're organised. A meeting starts, the candidate walks in polished and prepped, and the interviewer runs on whatever happens to be fresh in their head from the scroll.

That asymmetry is most of why follow-up interviews produce weaker signal than they should. The candidate has prepared for days. The interviewer has prepared for eleven minutes. Whichever way the conversation goes, it's moving at the pace the candidate is setting, not the pace the hiring team intended.

A real follow-up interview preparation is 25 minutes of uninterrupted reading and writing. That's the whole investment. It almost always pays for itself in the interview.

What the prep actually contains

Here's a workflow that's worked for me, roughly in the order the steps happen.

Minute 0–10: read the submission cold

Open the candidate's submission. Read everything — written answers, code, transcript. Don't take notes yet. Just read.

The point of reading it cold first is that you want to experience the submission the way the candidate expects you to, before you start looking for holes in it. First-read impressions are useful data. They're the data most hiring teams never capture because they read in note-taking mode from the start.

When you finish, write down one sentence: what's my honest overall read on this candidate right now? That sentence is your baseline. Everything else you do is either confirming or revising it.

Minute 10–20: reread with a pen

Now read again, more slowly, marking things up. What you're looking for:

  • Moments where the reasoning is sharp. A trade-off named unprompted. A constraint caught that wasn't in the prompt. A reason the candidate rejected an obvious path.
  • Moments where the reasoning is thin. A decision made without a stated justification. A trade-off the candidate didn't notice. An assumption that doesn't hold.
  • Specific lines in the transcript where the candidate pushed back on the AI, or where they didn't push back when they should have.
  • At least one place where you genuinely disagree with the choice the candidate made.

Don't try to write full questions yet. Just mark the spots.

Minute 20–25: turn three spots into openings

From your marked-up submission, pick three specific moments and write them into opening questions. Not generic ones — anchored ones. The format that tends to work is [reference the specific thing] → [ask for the reasoning behind it].

"In your write-up you argued for splitting the worker pool by tenant rather than by job type. Tell me what would make you change your mind."

"Your transcript shows you rejected the AI's suggestion to use a shared queue. What were you worried about?"

"You didn't name a failure mode in the document. Was that a scope call, or did it not feel load-bearing?"

Write them down. Bring them into the interview. That's the whole prep.

What to hold in your head, not on the doc

Two things worth keeping out of the written prep.

Your hiring recommendation. Don't write "leaning pass" or "probably hire" at the top of the prep doc. The act of writing it makes you confirm it in the interview. Keep your read in your head where it can update.

The questions you're hoping the candidate can't answer. Gotchas corrupt the follow-up. If you walk in wanting to prove a thesis, you'll ignore any evidence that contradicts it. The three openings you wrote down are anchored in the candidate's own work, not in the things you hope to catch them out on.

What not to bother preparing

Re-reading the CV in detail is rarely worth the time. By follow-up stage, the CV is already priced in. Going back to it tends to pull the conversation toward surface-level biography and away from the actual submission.

Prepping a "trick" problem for the last ten minutes is the other common waste. Most trick problems test whether the candidate has seen that trick before. Use the time for a live constraint introduction instead, where you modify the problem they already worked on and watch them update their model.

Why this much prep is worth it

A 45-minute follow-up interview that produced weak signal usually costs the company a second round, a longer debrief, or a bad hire that surfaces six months in. The cost of any of those dwarfs the 25 minutes you skipped. Prep isn't a luxury move — it's the cheapest way to raise the signal-to-noise ratio of the whole process.

At CriticCode the submission is designed to make this prep fast: the answers, the transcript, and the paste events all land on the same page, which means the 25-minute workflow compresses to closer to 15 once you've done it a few times. The workflow isn't tool-specific, though. It works on any submission that includes the candidate's reasoning alongside their artefact.

The short version

If you remember nothing else: read the submission twice, mark the moments, write down three specific opening questions, and don't write down your recommendation before the interview. That's the whole thing. It's almost embarrassing how much better follow-up interviews get when this much prep is actually done.