A Technical Interview Scorecard Template That Interviewers Actually Fill In

By Pooja Goenka · Published 2026-10-02 · Updated 2026-10-22

A technical interview scorecard that works has four or five competencies, a written description of what each score means, one line of evidence per score and a forced yes or no at the end. The template below has all of that, including an AI judgement row for interviews where candidates may use an AI assistant, and you can paste it into any spreadsheet or download it as a free file that opens in Excel and Google Sheets.

A disclosure: we build an interview platform with scorecards in it. Everything below works in a spreadsheet, and the template is written so you can copy it into one today.

The scorecards I see most often have eight criteria, each scored 1 to 5, and no one has written down what a 3 means. So every interviewer invents their own scale, writes "good communication, solid fundamentals" in the notes box, and gives a 4. The debrief then compares four private scales as if they were one.

I judge a scorecard by one test: would two interviewers who saw the same answer give it the same score? Passing it takes fewer criteria and much more specific wording than most teams use.

Do structured interviews actually predict performance better?

Yes. Structured interviews, where every candidate gets the same questions and is scored against the same written criteria, predict job performance better than unstructured conversations. That finding has held for decades. Schmidt and Hunter's 1998 meta-analysis put structured interviews well ahead of unstructured ones, and a 2022 re-analysis by Sackett and colleagues found structured interviews were the strongest single predictor in their review. Google's guide to structured interviewing makes the same practical point: use standardised rubrics so every reviewer shares one understanding of each score.

Most of that structure is written down in the scorecard. If everyone asks the same questions but scores from general impression, you have a shared list of questions and nothing written to score the answers against.

The template

These four competencies are for a mid-level software engineer, with a fifth, AI judgement, to add when the interview allows an AI assistant. Change them for the role, but keep the count low: four is enough, and six is the most an interviewer will fill in properly.

Below the table, every scorecard gets three more fields:

• Evidence. For each score, one sentence describing what the candidate said or did. "Wrote a brute-force solution, then reduced it to O(n) using a hash map when asked about scale" is evidence. "Strong problem solver" is not.
• Overall recommendation. Strong no, no, yes or strong yes, with no "maybe" option. A 2.5 average means you have not decided, and the debrief should hear that.
• What would change my mind. One line. It is the most useful field in the debrief, because it tells the panel what to ask the other interviewers.

A copy-and-paste version

Prefer a file? Download the scorecard (CSV). It has the same rows plus empty Score and Evidence columns. No sign-up.

To start a sheet, copy the block below into the first cell of an empty tab. Google Sheets and Excel both split it into columns on paste, or use Data, then Split text to columns, with comma as the separator.

Competency,1: Concern,2: Below bar,3: Meets bar,4: Strong,Score,Evidence
Problem solving,"Could not get to a working approach, even with hints",Reached an approach only after the interviewer gave the key idea,"Found a working approach, with small hints at most, and explained why it works","Compared two approaches and chose one for a stated reason",,
Code quality,"Code would not run, or could not be followed",Ran but with bugs the candidate did not find when asked to test,"Correct, readable, and tested with at least one edge case",Also named the cases it would fail on in production and how to guard against them,,
Technical judgement,Accepted a flawed suggestion without question,Spotted a problem only when pointed at it,Questioned a suggestion and explained what was wrong with it,Explained the trade-off and when the flawed version would be fine,,
Communication,Interviewer could not follow the reasoning,"Explained what they did, not why","Thought out loud, and checked assumptions before coding",Adjusted the explanation when the interviewer looked lost,,
AI judgement (optional),"Pasted AI output in without reading it",Checked the AI's output only after being asked,"Said what they would hand to the AI and what they would keep, and checked the output","Also set up a check that would catch the AI being wrong",,
Overall recommendation,Strong no,No,Yes,Strong yes,,
What would change my mind,,,,,,

Why use a four-point scale instead of 1 to 5?

A five-point scale gives everyone a safe middle. Interviewers who are unsure give a 3, and the 3s say nothing. With four levels you have to choose a side of the bar. The two middle levels are labelled "below bar" and "meets bar" on purpose, because that is the decision you are actually making.

What is the technical judgement row for?

Most templates do not have this row, and it is the one I would least want to drop. A lot of engineering work now starts from code or an answer someone else wrote, often an AI assistant, so I want to know whether the candidate notices what is wrong with it.

You can test that directly. Give the candidate a short, confident answer or code snippet with one real problem in it, and ask them to review it. Score what they catch, and score what they wrongly flag too, because a candidate who calls everything a bug is no more useful than one who misses everything.

How is AI judgement different from technical judgement?

Technical judgement is about reviewing an answer someone hands you. AI judgement is about how the candidate works when they have an assistant open themselves: what they choose to hand over, what they keep, and how they check what comes back. If your interview allows an AI tool, score it as its own row, so a candidate who is careful with the tool is not mixed up with one who is good at code review.

Keep the row optional. If candidates are not allowed an AI tool in the round, leave it out rather than guessing. For question ideas and a 0 to 3 rubric aimed at this skill, see AI fluency interview questions. Candidates preparing for this round may find how to answer "Do you use AI at work?" useful.

What makes interviewers fill in the scorecard properly?

First, fill it in before you talk to anyone. The scorecard is submitted before the debrief and before reading anyone else's. Once someone hears a colleague say "I thought they were great", their own memory of the interview shifts. We wrote about how to run the debrief itself separately.

Second, write the anchors before the first candidate. If you calibrate the anchors after seeing candidates, you calibrate them to the candidates you happened to like.

Third, give each interviewer one or two competencies. An interviewer who owns problem solving and code quality goes deep on those. An interviewer asked to score all four in 45 minutes guesses at two of them.

How do you run the scorecard in a spreadsheet?

Make one tab per role with the anchors at the top, and one row per candidate per interviewer holding the competency scores, evidence, recommendation and the "change my mind" line. Hide each interviewer's row from the others until everyone has submitted. Google Sheets protected ranges or a simple form can do that, and a form that writes to a sheet is enough.

Before the debrief, sort by competency and look for the rows where two interviewers are two or more levels apart. Start the discussion there.

Where software helps

The spreadsheet version breaks down in two places: keeping scores hidden until everyone has submitted, and finding disagreements quickly when you have ten candidates in a week. That is what LogicWiz Interviews does for hiring teams. Interviewers score privately against your rubric, nothing is revealed until all scores are in, and the debrief view puts the disagreements side by side. If you would rather stay in a sheet, the template above works there too.

Common questions

How many competencies should an interview scorecard have?

Four is a good default for one role, and six is the practical maximum. With more, interviewers stop reading the anchors and score from a general impression.

Should a scorecard use a 1 to 5 scale?

A four-point scale usually works better for hiring decisions, because it removes the neutral middle and forces each interviewer to say which side of the bar the candidate is on.

What should go in the evidence field?

One sentence describing what the candidate said or did, specific enough that another interviewer could picture it. Adjectives like "strong" or "solid" are not evidence.

When should interviewers fill in the scorecard?

Immediately after the interview, and before speaking to anyone else on the panel. Scores written after the debrief starts are shaped by the first opinion spoken out loud.

What is an AI judgement row on an interview scorecard?

It scores how a candidate works with an AI assistant during the interview: what they hand to the tool, what they keep for themselves, and whether they check the output before using it. Add it only when the round allows AI tools.

Is there a free technical interview scorecard template?

Yes. The table, the copy-and-paste block and the downloadable CSV file above are free to use in any spreadsheet. There is a short version of this answer on our quick answers page.

Can I use this template for non-engineering roles?

Yes. Keep the structure of four competencies, written anchors, an evidence field, a forced recommendation and the "what would change my mind" line, and replace the competencies with ones that matter for the role.