Structuring Interview Evaluations with Claude
Most interview debriefs turn into a five-minute conversation where whoever speaks loudest decides the outcome. The notes each interviewer took rarely get compared side by side, so the same candidate can look strong or weak depending on who's in the room and how confidently they argue their impression.
Warning
Unstructured debriefs are where bias creeps in easiest. If your evaluation criteria only exist in someone's head, they aren't being applied consistently, even when everyone in the room has good intentions.
A structured evaluation doesn't remove judgment from hiring. It just makes sure every candidate is judged against the same criteria.
Why side-by-side notes change the conversation
The core problem with an unstructured debrief isn't that people disagree. Disagreement is normal and often useful. The problem is that disagreement rarely gets surfaced explicitly. Two interviewers can walk away with opposite impressions of the same answer and neither one realizes it, because the conversation moves on before anyone compares notes closely.
Before: three sets of notes, each interviewer summarizing their own take out loud, no direct comparison.
After: one document, organized by criterion, where every interviewer's note on "technical depth" sits next to every other interviewer's note on the same thing.
That single change, comparing by criterion instead of by person, is what turns a debrief from a vibe check into an actual evaluation.
Turning notes into a decision
- 1
Collect every interviewer's raw notes
Don't summarize yet. Paste in the actual notes from each interviewer, even if they're messy or incomplete. Summarizing too early loses the specific details that make disagreement visible later.
- 2
Ask Claude to organize by your criteria, not by interviewer
Give Claude your actual evaluation criteria, the specific ones your team uses, not "communication skills" in the abstract, and ask it to sort every note underneath the right one.
- 3
Ask for disagreement, not just consensus
Explicitly ask Claude to flag anywhere interviewers rated the same thing differently. That's the part a five-minute debrief usually skips, and it's often the most useful part of the whole evaluation.
A structured evaluation should make it easy to see:
How the candidate scored on each specific criterion
- Where interviewers agreed and where they didn't
Direct quotes backing up each rating, not just a number
Here's a prompt that puts this into practice:
“
Here are notes from three interviewers on the same candidate for our backend engineer role. Organize them under our four evaluation criteria: technical depth, communication, ownership, and collaboration. Flag anywhere the interviewers rated the same thing differently, and include one direct quote per criterion.
”What to do when interviewers actually disagree
Finding disagreement is only useful if you do something with it. The instinct is often to average it out, call it a 3 out of 5 and move on, but that treats a real signal like noise.
Tip
When interviewers disagree on the same criterion, ask Claude to lay out both interpretations side by side with the supporting quote from each. That turns a vague "we didn't agree" into something the hiring team can actually discuss and resolve.
“
Two interviewers rated this candidate differently on "ownership." Show me both interpretations side by side, with the specific quote each interviewer is basing their rating on, so the hiring team can discuss which reading is more accurate.
”Keeping the process fair across a whole hiring round
The same structure that helps one evaluation also makes it possible to compare candidates fairly against each other, not just against an abstract bar. Once every candidate's notes go through the same criteria and format, ranking them becomes a matter of reading four consistent documents side by side, instead of trying to remember which candidate said what three interviews ago.
If your evaluation format changes candidate to candidate, you're not really comparing candidates. You're comparing whoever happened to write clearer notes.
Common mistake
Using a structured evaluation for some candidates in a hiring round and informal notes for others. The comparison only works if every candidate went through the same process.
Inside Claude Tutorial
Spotting when a request needs structure is transferable.
This kind of structuring shows up far beyond interviews. The app has a full lesson on it, with practice that applies to any request that needs more than one instruction to get right.
Writing the final recommendation
Once the structured evaluation exists, the hiring decision still needs a clear written recommendation, something a hiring manager can act on without re-reading every interviewer's raw notes. This is a separate step from organizing the notes themselves.
“
Based on the structured evaluation above, write a one-paragraph hiring recommendation. State the overall recommendation clearly in the first sentence, then reference the two or three pieces of evidence that most influenced it, including any unresolved disagreement between interviewers.
”Tip
Ask for the recommendation to lead with the actual decision, not a summary of the process. "Recommend to hire, contingent on a stronger reference check around ownership" is more useful on a Friday afternoon than three paragraphs restating what everyone already said in the room.
Setting criteria before the interviews happen
Everything above assumes the evaluation criteria already exist. If they don't, the best time to define them is before the first interview, not while organizing notes afterward. Criteria written after the fact tend to bend toward whichever candidate the room already liked.
“
We're hiring a backend engineer. Help me turn this job description into four specific interview evaluation criteria, each with one or two concrete signals an interviewer could actually look for, not vague traits like "strong communicator."
”Common mistake
Reusing the same generic criteria (communication, culture fit, technical skill) for every role, regardless of what the job actually requires. Criteria that don't reflect the specific role tend to produce ratings that don't predict anything useful about on-the-job performance.
Continue reading
- Claude for Everyday HR Workflows: the broader habits this structured approach comes from, if you want to see it applied beyond interviews.
- Getting Started with Claude: the underlying Claude fundamentals, context, iteration, structure, that this entire approach is built on.