How JobPLT's match score works, and how to use it to prioritise applications
A plain-language look at JobPLT's 0–100 job match score: skill overlap, text similarity and title fit. Learn what each part means and how to use the score to decide which jobs to apply for first.
A long list of job results isn't much help if you still have to read every posting to find the good ones. That's why every job in JobPLT gets a match score from 0 to 100. This post explains exactly how that number is calculated, so you can trust it and use it well.
The three parts of the score
The score is a weighted mix of three measurements:
| Part | Weight | What it measures |
|---|---|---|
| Skills | 45% | How many of the job's skills you have, and how many of yours it uses |
| Text similarity | 30% | How closely the job description's wording matches your CV |
| Title fit | 25% | How well the job title matches the roles you're looking for |
1. Skills (45%)
JobPLT recognises around 160 technical skills, from Python and SQL to LangChain, Kubernetes and Power BI, including common aliases. For each job it finds the skills the posting mentions, then looks at two things:
- Coverage: of the skills the job asks for, what share are on your CV?
- Depth: how many of your skills does the job actually use?
Coverage alone would over-reward vague postings that only mention two or three skills, so depth balances it out. Generic terms such as "Git" or "Statistics" don't count as missing skills, because nearly every job mentions them.
2. Text similarity (30%)
Skills aren't the whole story. A job about "building retrieval pipelines for enterprise search" may fit you even if it never names your exact tools. To capture that, JobPLT compares the full text of your CV with each job description using TF-IDF, a standard technique that gives more weight to distinctive words and less to common ones.
Long job descriptions naturally produce small similarity numbers, so scores are scaled within each search: the most similar jobs in your results get full marks, and the rest are scored relative to them.
3. Title fit (25%)
Finally, JobPLT compares each job title with the titles you searched for and the roles on your CV, using fuzzy matching so "Sr. ML Engineer" still matches "Senior Machine Learning Engineer". Titles that name one of your core skills ("AI", "NLP", "Data") get extra credit.
Sensible adjustments
Two adjustments keep the ranking realistic:
- Non-engineering titles are discounted for technical CVs. A "Sales Director, AI Products" posting mentions AI, but it isn't the job you're looking for.
- Seniority is taken into account. Internships and fresher roles rank lower for senior profiles, and very senior titles rank a little lower for early-career CVs.
How to read the score
| Score | Label | What to do |
|---|---|---|
| 80–100 | Great match | Apply soon. These jobs fit your skills and the roles you want. |
| 65–79 | Good match | Worth a look. Check the "Job also mentions" skills to see what's missing. |
| 50–64 | Fair match | Possible stretch roles. Read the description before deciding. |
| Below 50 | Low | Hidden by default. Lower the minimum match if you want to see them. |
Open any job to see its score breakdown, the skills you share with it and the skills it mentions that aren't on your CV.
Getting better matches
The score is only as good as the keywords behind it, and you can tune them:
- Remove keywords you don't want weighted. If you used a tool years ago and don't want jobs built around it, remove the chip.
- Add skills the CV parser missed. Custom keywords count towards matching too.
- Re-score without searching again. After editing keywords, click "Re-score current results" and the existing results update instantly.
- Search for the right titles. Title fit is a quarter of the score, so make sure the job titles you search reflect the roles you actually want.
Why not just use an AI model?
Large language models are good at many things, but a match score you can't explain is hard to trust. JobPLT's score is deterministic: the same CV and job always produce the same number, and every part of it is visible. That said, we're exploring AI for things it does well, such as rewriting CV bullet points, which you can read about in our post on what the CV score checks.
Ready to see your matches? Start a search or read the introduction to JobPLT.