Candidate comparison matrix: stop hiring the best talker. A fair way to score your shortlist

A candidate comparison matrix can save you from the most expensive hiring mistake there is: offering the job to the person who interviewed best instead of the person who can do the work. When three or four shortlisted candidates all seem "pretty good", memory and gut feel fill the gaps. The last person you spoke to often wins, and a strong candidate who was nervous on a Monday morning gets forgotten.
A simple scoring matrix fixes that. You decide what matters before you meet anyone, score every candidate against the same points, and keep a written reason for each score. This article shows you how to build one in an afternoon, with a sample matrix for a trade role, a scoring scale you can copy and the mistakes that make a matrix unfair.
What a candidate comparison matrix is
A comparison matrix is a table. The rows are the things the job needs. The columns are your shortlisted candidates. Each cell holds a score and a short note on the evidence behind it.
It does three jobs at once:
- It makes you decide what matters first, before a confident talker or a familiar name shifts your view.
- It puts every candidate through the same test, so you compare evidence with evidence.
- It leaves a record. If a candidate asks why they missed out, or a colleague questions the choice, you can show the reasons.
A matrix does not make the decision for you. It makes the decision easier to explain and harder to get wrong.
Step by step: build the matrix before the first interview
- Write down the five to seven criteria the job really needs. Take them from the job ad and the duties, not from a wish list. For a mechanic: diagnostic skill, brands and systems, licence, safety habits, reliability, communication with customers.
- Split must-haves from scored criteria. A licence the law requires is pass or fail, not a score out of five. Anyone who fails a must-have does not go into the matrix.
- Give each scored criterion a weight. Use 1 to 3. The skill that decides whether the work gets done well is a 3; nice-to-haves are a 1.
- Agree a scoring scale and what each score means. Write one sentence for each level so two people scoring the same candidate land close together.
- Decide what evidence counts for each criterion. A practical test, a reference answer, a work video, a verified role or a licence check, not "seemed keen".
- Score straight after each step, not at the end of the week, and write the evidence in the cell.
- Multiply each score by its weight and add up the totals, then look at the notes before you look at the totals.
A scoring scale you can copy
Keep the scale short. A five-point scale is enough, and each point needs a plain meaning.
- 5: Strong evidence. Shown in a test or confirmed by a referee or document, and above what the job needs.
- 4: Good evidence. Meets the need, with at least one solid piece of proof.
- 3: Some evidence. Probably meets the need, but the proof is thin or only from the interview.
- 2: Weak. Partly meets the need, or claims that could not be backed up yet.
- 1: No evidence. Not shown at all.
Notice the scale rewards evidence, not confidence. A quiet candidate with a confirmed referee scores higher than a smooth talker with nothing to show.
Example: comparing three welders
Here is a sample matrix for a fabrication workshop hiring one welder. The names are fictional. Must-haves (the right to work and the tickets the job requires) were checked first; all three passed.
| Criterion (weight) | Candidate A: Ravi | Candidate B: Grace | Candidate C: Tomasz |
|---|---|---|---|
| MIG and TIG on stainless (3) | 5: clean test piece, both processes | 3: good MIG, TIG not tested | 4: good test piece, slower |
| Reading drawings (3) | 4: answered the drawing task | 5: spotted an error in the drawing | 3: needed help with one symbol |
| Safety habits (2) | 4: checked gas and earth before starting | 4: same | 2: started without checking ventilation |
| Past work confirmed (2) | 3: one referee reached | 5: two roles confirmed | 4: referee and work video |
| Communication (1) | 3: brief but clear | 4: explained her steps | 4: clear |
| Weighted total | 44 | 46 | 37 |
The totals put Grace ahead, but the notes matter more. Ravi is the stronger welder on stainless. If stainless TIG work is most of the job, test Grace on TIG before you decide. The matrix shows you exactly which gap to close, instead of leaving you with a vague feeling.
The practical test behind these scores matters. The article on running a practical trade test fairly covers how to set one up, and the guide to reference check questions that test the work helps you fill the "past work confirmed" row with real answers.
Keep the matrix fair
A matrix can make bias look official if the criteria are wrong. Check yours against these rules.
Score only what the job needs
Never add a row for age, gender, marital status, family plans, religion, nationality or anything else unrelated to the work. In most countries that is unlawful discrimination, not just bad practice. The UK government's page on preventing discrimination as an employer explains the basics; other countries have similar rules, so check your own government's guidance.
Watch out for "culture fit"
"Fits the team" often means "is similar to the people already here". If teamwork matters, write what it looks like in the job: shares tools, hands over jobs clearly, follows site rules. Then score that.
Score each candidate on the evidence you have, not the evidence you assume
An overseas candidate whose employer is slow to reply is not a liar. Score the row as "some evidence" and ask for other proof, such as a service letter, payslips or a work video. Treat an odd date as a question to ask, not as proof of a problem.
Use more than one scorer when you can
If two people interview, each fills in a copy alone, then you compare. Big differences on one row are worth a conversation before the offer.
Common mistakes that wreck a comparison
- Writing the criteria after the interviews. You end up describing the person you already like.
- Too many rows. Twelve criteria all weighted the same hide the two that matter.
- Scores without notes. A 4 with no reason is useless a week later.
- Trusting job titles. "Supervisor" can mean very different things. The article on questions that test inflated job titles shows what to ask.
- Letting the total decide alone. Two points between candidates is within the margin of a bad morning. Read the notes.
Comparing shortlisted candidates on Talent Tabloid
If your candidates have Talent Tabloid profiles, some of the evidence for your matrix is already in one place. Each profile can show work history with verified badges where a role has been confirmed, a short video intro and work videos, photos or links that show the work. In the employer dashboard, Compare Candidates puts two to four shortlisted candidates side by side, with an AI-written comparison summary and a PDF report you can download. Each employer account can run 10 comparisons a month.
Treat that summary as one more input to your own matrix, not as the decision. Your criteria, weights and test results stay yours. You can look up a candidate who shares their username on People Search, and the employers page explains the plans and contact unlocks.
To start comparing your next shortlist, create an employer account.
Frequently asked questions
How many criteria should a comparison matrix have?
Five to seven scored criteria is usually enough, plus any pass-or-fail must-haves. More rows tend to blur the ones that decide the job.
Should every criterion have the same weight?
No. Give the skills that decide whether the work gets done well a higher weight, and keep nice-to-haves low.
What if two candidates end up with almost the same score?
Read the notes and find the row where they differ most. Then gather one more piece of evidence on that row, such as a second test task or another reference.
Can I share the matrix with an unsuccessful candidate?
You can explain the criteria and where they fell short. Keep other candidates' names and scores private.


