Recagn
Sign in

Guide · Scoring

Interview scorecard: how to build a weighted rubric, rated 1 to 5

Decide what the role needs before the first interview, write down what each score means, and keep the decision with a person.

What is the difference between a hard gate and a scored criterion?

A hard gate is a pass-or-fail rule a candidate must meet, such as working on-site in Pune; a scored criterion is rated 1 to 5 and weighted. Gates are checked first, so a missing must-have can't be outscored by strong answers elsewhere. You decide which needs are gates, with the client, before anyone is interviewed.

The test for a gate

Ask the client one question about each requirement: would you still meet a candidate who lacked this? If the answer is no, it's a gate. If they would meet the candidate anyway, it's a scored criterion.

Gates usually cover:

  • the work location, or the number of days on-site;
  • a licence or certification the role needs;
  • a stated minimum of experience;
  • notice period, when the client has a limit;
  • CTC, when the client has a budget it won't move.

Keep the list short. Each gate removes people outright, and a gate the client would have traded away costs you candidates they would have wanted to see.

Everything the client would weigh against something else is a scored criterion. Two years of B2C sales may be a gate for one client; how clearly a candidate explains a delay to a customer is a criterion for almost any of them.

A gate needs a question

A gate is only as good as the question that settles it. If the interview never asks about notice period, the answer is missing, and a missing answer is not a pass.

One rule keeps this the same for every candidate: if the answers don't show it, mark it not met.

The questions that sort a client's needs into gates and criteria belong on the intake call. They are set out stage by stage in the candidate screening process for recruitment agencies.

How do you write scorecard criteria a candidate can show evidence for?

Write each criterion as something the candidate did and can describe, such as "handled payment reversals in production", not a topic like "payments experience". An interviewer can only score what an answer can show, so a topic invites a guess and a described piece of work invites evidence. You decide the wording, because it decides who scores well.

Same answer, different score

Take two versions of one criterion for a Java role. "Owned a payments or ledger service" asks for ownership the client may not need. "Handled payment reversals in production" asks for the work the client actually described.

A candidate who fixed reversals while on call, but never owned the service, scores low on the first and well on the second. The answer is the same in both cases. Only the wording moved.

Four checks for each criterion

Run every criterion past these before it goes on the scorecard:

  • Observable. An answer can show it. "Has a growth mindset" can't be heard in an answer; "set up a follow-up routine without being asked" can.
  • One idea. "Communication and leadership" is two criteria. An interviewer will score whichever one the answer happened to show.
  • Tied to the brief. Each criterion traces back to something the client said or wrote.
  • Not a gate. If missing it rules the candidate out, it belongs with the gates.

Leave out "culture fit". It can't be anchored, and it lets the interviewer's own taste into the score. Ask the client which behaviour they mean, such as "says early when a deadline will slip", and write that down instead.

Each criterion needs at least one question that tests it, and each question should name the criteria it tests. A criterion no question asks about can only be guessed.

Three to six criteria is usually enough for a first-round interview. With more, each one rests on a single answer's worth of evidence, or less.

The same criteria work on CVs before anyone is interviewed. That step is covered in how to shortlist candidates from resumes.

What should an interview rating scale of 1 to 5 look like?

A 1 to 5 interview rating scale needs a written anchor for each score, describing what an answer at that level shows. Without anchors, one interviewer's 3 is another's 4; the US Office of Personnel Management's structured-interview training slides ask for example responses at each level, matched against the candidate's answer. You write or approve the anchors before the first interview.

Here is one criterion with all five anchors written out. It is an illustration made for this guide, not taken from a real rubric.

ScoreWhat an answer at this level shows
1Can't describe the problem without jargon.
2Describes the problem, but the client would still need help to follow it.
3Explains the problem and the next step in plain words.
4Also states the cost or the risk in the client's own terms.
5Does all of that, checks the client understood, and gives an example of an explanation that changed a client's decision.

Source: illustration written for this guide · criterion "Explains a technical problem to a client who isn't technical"

How to write the anchors

Write anchors 1, 3 and 5 first. They are the easiest to picture: no evidence, enough evidence, and the strongest answer you expect to hear. Anchors 2 and 4 sit between them.

Each anchor names evidence: a number, an example, a step the candidate describes. "Strong communication" is a label, not an anchor, because two interviewers will hear it in different answers.

Test them before the first interview

Try the anchors on two or three answers you have heard before, for a similar role. Score them on your own, and ask a colleague to do the same. Where your scores differ, the anchor at that point is being read two ways, so add a sentence to it until you agree.

How to score against them

Score each criterion only from the answers to the questions linked to it, and write the answer, or a short quote from it, beside the score. A score with no quote beside it is an impression.

Use whole numbers. If interviewers keep reaching for a 3.5, the anchors aren't doing their job, and the fix is a clearer anchor, not a half point.

Why five points

More points need more anchors, and each extra anchor is harder to tell from its neighbour. A 4-point scale removes the middle score, which some teams want, because it forces a lean one way or the other.

OPM's same training slides (undated; read 10 Nov 2026) advise one range of levels for all competencies, typically between 3 and 7. Five sits inside that range.

The anchors make two interviewers' scores comparable. They don't make the score a decision.

How do you build a weighted scoring rubric for hiring?

Multiply each 1 to 5 score by its weight, add the results and divide by the total weight, so the result stays between 1 and 5. Set two cut-offs before the first interview, one to move on and one for maybe, so every candidate is sorted the same way. The total is a recommendation; you decide who moves on.

Choosing the weights

Weights say what the client would trade. Put the most weight where the role succeeds or fails, and check the split with the client before you lock it.

OPM's guidance on scoring structured interviews (read 10 Nov 2026) recommends equal weights, and asks you to document the reason when one competency counts for more. An agency usually has that reason, because the client gave it to you. Write it down beside the weight.

Dividing by the total weight is what keeps the result on the 1 to 5 scale, whatever the weights add up to.

A worked illustration

This example is made up for this guide; no real candidate or rubric is behind it. A scorecard has four criteria weighted 40, 30, 20 and 10, and a candidate scores 4, 3, 2 and 5 on them.

(4×40 + 3×30 + 2×20 + 5×10) ÷ 100 = (160 + 90 + 40 + 50) ÷ 100 = 3.40.

Say the cut-offs are 3.7 to move on and 2.5 for maybe. A 3.40 sits between them, so the scorecard recommends maybe.

Had the client set the move-on bar at 3.0, the same scores would move on. The bar decides as much as the scores do, which is why it is set in advance.

Read the scores behind the total

Two candidates can reach the same total by different routes. A 3.40 made of 5s and 1s is a different candidate from a 3.40 made of 3s and 4s, so read each criterion score before you read the total. A 1 on the criterion the client cares most about is worth raising with them, whatever the total says.

Cut-offs and gates

Set cut-offs per role, before any interview. A senior role may need a higher bar than a junior one. Changing a cut-off after you've seen the candidates means choosing the candidate first and the rule second.

Gates come before any of this. A candidate who fails one is out of the running whatever their total, so the arithmetic only sorts candidates who passed every gate.

The full arithmetic, gates and cut-offs are set out in how AI interview scoring works.

What goes on an interview scorecard template and evaluation form?

An interview scorecard template holds a role's hard gates, weighted criteria, anchors and cut-offs; an interview evaluation form is one candidate's copy, filled in. Write the template once per role, before any interview, and fill in a fresh form with scores and evidence for every candidate. The form ends with the recommendation and your decision, kept apart.

A blank interview scorecard template, with each field explained and a copy filled in from a worked example, is on the interview scorecard template page.

A blank interview evaluation form, with each field explained and one candidate's copy filled in from a worked example, is on the interview evaluation form page.

Here is a layout for the evaluation form you can copy into a document or spreadsheet.

SectionFields
Candidate and roleName, role, client, interview date, interviewer
Hard gatesEach gate: met or not met, with the answer that shows it. If the answers don't show it, mark it not met.
Scored criteriaEach criterion: its weight, a score from 1 to 5, and the quote or note the score rests on
TotalThe weighted total to two decimals, and the cut-offs used
RecommendationMove on, maybe or no, from the gates and the total
DecisionYour decision and, if it differs from the recommendation, why

Source: layout written for this guide

Fill the form in straight after the interview, before you discuss the candidate with anyone. Once the panel has talked, the form records the group's view.

Keep the recommendation and the decision in separate fields. Later, you can see how often you overrode the total and why, which tells you whether the weights or the cut-offs need changing.

Ready-made scorecards for six real roles are at the end of this guide. They are role templates, with the gates, weights, anchors and cut-offs already filled in; the evaluation form is still yours to fill for each candidate.

More than one interviewer

When two people interview the same candidate, each fills in a form of their own, from the same template. Compare the forms only after both are in. Where two scores on one criterion are far apart, read the quotes side by side: usually one anchor is being read two ways, and it needs another sentence.

What are the most common interview scorecard mistakes?

The most common scorecard mistakes are criteria nobody can show evidence for, scores without anchors, and one strong impression lifting every score: the halo effect. Each one makes the total look precise while it records the interviewer's general feeling, and each has a fix you can make before the first interview.

  1. Criteria nobody can show evidence for. "Passionate" and "culture fit" can't be heard in an answer. Rewrite them as work the candidate did, as in the four checks above.
  2. No anchors. Without them, interviewers drift apart and each one's 3 means something different. Write anchors 1, 3 and 5 first.
  3. The halo effect. One good answer raises every score. Score one criterion at a time, only from its linked questions, and quote the answer beside each score.
  4. Too many gates. The pile empties before the interview starts. Put every gate through the "would the client still meet them?" test.
  5. Weights set after interviewing. Weights and cut-offs chosen once you've met the candidates can be bent to fit a favourite. Set both before the first interview, and divide by the total weight.
  6. A scorecard filled in after the debrief. It records what the panel agreed, not what this interviewer heard. Fill it in first, then talk.
  7. The total treated as the decision. It is a recommendation. Record your decision in its own field, with a reason when it differs.

When is an interview scorecard the wrong tool?

A fixed scorecard fits poorly when the client can't yet say what the role needs, or when only one or two candidates are in play. Then a conversation with the client, or an open interview with notes, tells you more than a weighted total does. You decide when to set the scorecard aside, and say so to the client.

  • The brief keeps changing. Each change means rescoring everyone or comparing candidates on different scales. Lock the scorecard only when the client agrees it reflects the role.
  • The client is choosing between shapes of role. In some executive searches the client wants to meet different profiles before deciding what the job is. A scorecard would fix an answer they haven't reached.
  • One or two candidates. With nobody to sort, a weighted total adds a step and tells you little that your notes don't.
  • Skills an interview can't show. A designer's portfolio, a work sample or a reference check carries that evidence. Score those separately, or keep them off the interview scorecard.

How Recagn does this

Recagn is our product. This section shows the scorecards, which Recagn calls rubrics, that it drafted from six real job postings, and how it scored one question's answers in a real run on 8 Nov 2026. Each posting is adapted from a real public job posting, with company details removed. The candidates are invented, and their answers were written for the example.

How Recagn drafts a rubric

Recagn drafts the rubric from the client brief: hard gates, weighted criteria with an anchor for each of the five points, two cut-offs, and interview questions linked to the criterion or gate each one tests. Its drafting instructions ask for anchors that describe observable evidence and prefer three to six criteria, and they treat weights as relative.

You edit the rubric and lock it before anyone is invited. Every candidate is then scored against that locked version. For these six examples, each rubric was locked unedited, to show Recagn's draft as it came.

Turning a client's job description into a scored rubric comes off your plate. What the role needs stays yours, and so do the questions and the hard gates: you edit and approve the rubric before anyone is invited.

Six rubrics, side by side

Role (worked example)Hard gatesScored criteriaLargest weightQualified fromMaybe from
Inside sales / admissions counsellor, Pune3525% (two criteria)3.62.8
Digital marketer, Bhubaneswar3625% (paid performance marketing)3.83.0
Senior Java full stack and platform engineer, remote4625% (Java full-stack engineering)3.83.0
Account manager, AI software, remote (India)4525% (two criteria)3.83.0
Founding GTM engineer, remote (India)3525% (two criteria)3.83.0
Head of engineering, manufacturing software, Kochi3625% (managing managers and org design)4.03.2

Source: Recagn's draft rubrics, run 8 Nov 2026 · locked unedited · cut-offs on the 1 to 5 weighted score

No criterion carries more than 25% of the score in any of the six. The cut-offs differ by role: the head of engineering rubric asks for 4.0 to qualify, and the inside sales rubric for 3.6.

None of the six has a gate for notice period or pay. None of the six postings mentions a notice period, and only the digital marketer posting states pay, 3 to 4 lakhs per annum, which its draft did not turn into a gate. Adding either is your edit.

Each rubric below is shown as cards, then as a free Word document. The document holds the hard gates, the scored criteria with their weights and five anchors, and the two cut-offs. It doesn't hold the interview questions.

Inside sales / admissions counsellor, Pune

1–8 of 8
inside-sales-edtech-pune-rubric.docxWord document · 12 KB · free, no sign-up

Three hard gates, including on-site availability in Pune, and five scored criteria. B2C sales and conversion, and counselling and objection handling, carry 25% each. It qualifies from 3.6 and gives maybe from 2.8. Recagn drafted it and used it for four invented candidates' interviews.

Digital marketer, Bhubaneswar

1–9 of 9
digital-marketer-fitness-bhubaneswar-rubric.docxWord document · 13 KB · free, no sign-up

Three hard gates, including on-site availability in Bhubaneswar, and six scored criteria, with paid performance marketing on Google and Meta at 25%. It qualifies from 3.8 and gives maybe from 3.0. Recagn drafted it and used it for four invented candidates' interviews.

Senior Java full stack and platform engineer, remote

1–10 of 10
senior-java-full-stack-rubric.docxWord document · 13 KB · free, no sign-up

Four hard gates, including 5+ years of relevant experience, and six scored criteria, with Java full-stack engineering at 25%. It qualifies from 3.8 and gives maybe from 3.0. Recagn drafted it and used it for four invented candidates' interviews.

Account manager, AI software, remote (India)

1–9 of 9
account-manager-ai-software-rubric.docxWord document · 12 KB · free, no sign-up

Four hard gates, including being based in India, and five scored criteria. Expansion and revenue ownership, and technical fluency, carry 25% each. It qualifies from 3.8 and gives maybe from 3.0. Recagn drafted this rubric only; no candidate was interviewed.

Founding GTM engineer, remote (India)

1–8 of 8
founding-gtm-engineer-rubric.docxWord document · 13 KB · free, no sign-up

Three hard gates, including fully remote work with flexible hours, and five scored criteria. Commercial research, and pipeline ownership, carry 25% each. It qualifies from 3.8 and gives maybe from 3.0. Recagn drafted this rubric only; no candidate was interviewed.

Head of engineering, manufacturing software, Kochi

1–9 of 9
head-of-engineering-manufacturing-software-rubric.docxWord document · 12 KB · free, no sign-up

Three hard gates, including being based in Kochi or willing to relocate there, and six scored criteria, with managing managers and org design at 25%. It has the highest bar of the six: it qualifies from 4.0 and gives maybe from 3.2. Recagn drafted this rubric only; no candidate was interviewed.

One question, scored at four levels

In the inside sales run, four invented candidates answered the same question: "How do you organise your follow-ups so no qualified lead slips through, especially when you have many leads at once? Describe your system or cadence." Recagn scored each on follow-up discipline.

Candidate (invented)Their answerRecagn's score on follow-up discipline
Debashish M."I would make a list in a notebook or Excel and call them back when I remember."1
Karan B."I note down who to call back and when, and I try to call them the next day. When I have too many I call the newest ones first."2
Tanvi J."I keep a daily list in the CRM and call high-intent leads first. I follow up every two days until they decide."3
Nisha P."Every lead gets a next step logged the same day. I follow a five-touch sequence over ten days…"5

Source: worked example, inside sales / admissions counsellor, Pune · scored by Recagn on 8 Nov 2026 · invented candidates; answers written for this example

The rubric's anchor 1 for this criterion reads "Follows up only when reminded; no system; lets leads go cold." Recagn's reason for Debashish M.'s 1 says his answer relies on memory, with no system or cadence, which is that anchor. Anchor 5 asks for follow-ups segmented by intent and timed on purpose, and Recagn's reason for Nisha P.'s 5 says her full answer, a five-touch sequence with high-intent leads called first each morning, matches it. No candidate scored a 4.

Each score uses the answers to both questions linked to this criterion: this one, and a question about how the candidate uses a CRM. The table shows one of the two answers each score rests on.

Each score comes with its written reason beside the answers, so every score cites the answer it rests on. You read them and decide whether the score holds. Scoring the answers against the rubric comes off your plate.

From scores to a recommendation

Tanvi J. scored 3 on all five criteria, so her weighted total is 3.00. That sits between the inside sales rubric's cut-offs of 2.8 and 3.6, so Recagn recommends Maybe. Nisha P.'s scores of 4, 4, 5, 5 and 5 give 4.50, so Recagn recommends Qualified.

The client report shows each criterion's score and the recommendation, not the weighted total. The totals here are worked out from the report's scores by the scoring rule: each score times its weight, added up, divided by the total weight.

Recagn checks the hard gates first. It counts a gate as failed when the answers don't show it is met, and a failed gate makes the recommendation Disqualified, whatever the total.

Recagn recommends. Only you can reject a candidate. Recagn runs these first-round interviews by text, voice or video, as its AI interview software page shows.

In these runs, Recagn drafted six rubrics and scored the interviews in three of them. Editing each rubric, setting the gates and every decision stayed with the recruiter.

See a worked example

Questions people ask

Can AI score an interview against a scorecard wrongly?

Yes. Recagn's score can miss what an answer meant, so each criterion's score comes with a written reason beside the candidate's answers. You approve the recommendation or override it with a note, and only then decide. Candidates agree before they start that AI evaluates their answers and that a person reviews them and makes every decision.

Does Zoho Recruit have interview scorecards?

Yes, as interviewer assessments. Zoho Recruit lets interviewers rate with stars from 1 to 5, a thumbs-up or thumbs-down, or a number out of a set maximum for each question, and recruiters can set a qualifying score as a percentage of the total, going by Zoho Recruit's help page, read 10 Nov 2026. Recagn works alongside an ATS like it: it drafts the rubric and scores the first-round interview, and doesn't replace the ATS. You approve or override each recommendation, and you decide who moves on.

Are these interview scorecard templates free?

Yes. The six rubric downloads on this page are free Word documents, with no form to fill in. Scoring an interview on Recagn is part of the credits the interview uses: one for a text or voice interview, three if any answer is video. On Starter, monthly, that is at most ₹50 per text or voice interview (₹49.98; ₹58.98 with 18% GST) and at most ₹150 per video interview (₹149.95; ₹176.94 with 18% GST).

Should the weights on a scorecard add up to 100?

Only if you write them as percentages, so a reader can see each criterion's share at a glance. Any numbers work for the arithmetic, because the weighted sum is divided by their total.

Is a 1 to 10 scale better than 1 to 5?

Not usually. Every extra point needs another written anchor, and an anchor that differs only slightly from its neighbour is hard to score against. Five anchored points are enough for a first-round interview.

Can you change the scorecard after interviews have started?

You can, but candidates already interviewed were scored against the old version, so compare like with like. In Recagn, locking an edited rubric makes a new version, and each candidate keeps the score from the version that was current when they finished. You decide whether anyone should be interviewed again.

What is the difference between an interview scorecard and an interview evaluation form?

They are used at different times by different people. The scorecard is agreed with the client before interviews and doesn't change between candidates. The evaluation form is filled in by one interviewer straight after one interview, and it holds that candidate's scores and quotes.

Should notice period be on an interview scorecard?

Yes, as a hard gate, when the client has a limit: it is pass or fail, never scored. None of the six rubrics on this page has one, because Recagn drafts from the posting and none of the six postings mentions a notice period. Adding it is the recruiter's edit.

See one real brief scored from rubric to shortlist.