ATS guides
ATS keywords, without the stuffing
The keyword advice industry rests on an assumption worth examining: that an applicant tracking system counts words and ranks you by the count. Appliora's model has no keyword check in it at all. Terms are scored where they appear - inside Concrete skills (8 points) and Bullet quality (13 points) - and matching against one specific advert is a separate question with its own separate model.
Where a keyword is actually scored
Two checks carry it, and they measure different things about the same word.
Concrete skills (8 points) asks whether the skills on your CV are concrete and complete: whether there are enough of them, whether they arrive as whole words rather than as fragments, whether a language is listed with a level, and whether your hard skills exist anywhere other than buried in the bullets. It is a question about the skills section as an object.
Bullet quality (13 points) - the heaviest check after reading order - asks whether your experience proves anything. It measures quantification, action verbs, passive voice and bullet length. A term inside a bullet that names the scale and the outcome is evidence; the same term in a list is a claim.
That split is the whole argument of this page. If you want a term to count, the strongest available move is not to repeat it - it is to put it somewhere it is doing work. A recruiter searching the database finds it either way; a recruiter reading the result finds a reason to keep reading in only one of them.
The published thresholds
Appliora publishes the numbers its skills check runs on, so this is checkable rather than asserted.
| Condition | Threshold | What it means |
|---|---|---|
| Too few | Under 5 skills | Reads as an unfinished CV. Five is a floor, not a target. |
| Stuffed | Above 40 skills | Reads as keyword stuffing, which parsers and humans both discount. |
| Orphaned fragment | 4 characters or fewer | The surviving half of a clipped skill - "MS" where a template cut "MS Office". Real short abbreviations are on a curated exception list: SQL, AWS, GCP, CSS, HTML, PHP, SEO, SEM and others. |
| Language with no level | 2 or more languages listed | A language with no level is a claim a recruiter cannot filter on and an interviewer will test in five minutes. Reported as a warning, never fatal. |
| Skills only in the bullets | No skills section | Your hard skills exist but not where a search for them looks. |
Thresholds from the published model, version 4.5.
Note what is absent from that table: any notion of density, frequency or percentage. There is no target number of times to mention a term, because there is no check that counts them.
Choosing the terms that are worth the space
Four sources, in descending order of how much they tell you.
- The advert itself. Its must-have lines, separated from its wish list. Pulling the real requirements out of a job advert is the mechanical version of this and it is worth doing properly.
- Three or four other adverts for the same title. A term that appears in all of them is the profession's vocabulary. A term that appears in one is that company's vocabulary, and belongs in the CV you send that company.
- The tools you actually used. Named versions, named platforms, named methods. These are the terms a recruiter searches on, because they are the ones that are unambiguous.
- The words your own industry uses for the outcome. Retention, utilisation, first-pass yield, churn. Generic verbs describe activity; the outcome word describes what the activity was for.
What does not belong: soft skills as bare nouns. "Team player", "detail-oriented", "excellent communicator" are unfalsifiable, they are on every CV in the pile, and they consume the line that could have carried a tool. The skills section guide goes further into what to keep and what to prove.
Placing them where they count
Same words, two arrangements, different score and different reader.
- Weak
- Skills: SQL, Python, SQL queries, data analysis, analytics, reporting, dashboards, BI, business intelligence, Tableau, data visualisation, visualisation
- Stronger
- Skills: SQL, Python, Tableau, dbt, experiment design
- Why
- The first list says one thing five ways to look longer. Duplicated stems and their synonyms are what the skills check reads as stuffing; the second is shorter, and every item is a different capability.
- Weak
- Responsible for reporting using SQL and Tableau.
- Stronger
- Rebuilt the weekly commercial report in SQL and Tableau, cutting the manual assembly from four hours to twenty minutes and removing two recurring reconciliation errors.
- Why
- Both contain SQL and Tableau. Only the second is scored well by the bullets check, which measures quantification and action verbs - and only the second survives being read by a person. Numbers here have to be your own; see how to quantify an achievement.
- Weak
- Languages: English, Polish, German
- Stronger
- Languages: Polish (native), English (C1), German (B1)
- Why
- The level is the filterable part. Appliora warns about missing levels once 2 or more languages are listed, because that is the point at which the omission looks like a habit rather than an oversight.
- Weak
- A block of forty terms under the heading Core Competencies
- Stronger
- Twelve terms in the skills section, each of them also visible in a bullet that proves it
- Why
- Forty is above the point where the list reads as coverage. Twelve that recur inside your experience are twelve the recruiter can verify without asking you.
Keywords are a per-advert problem
A general CV has general terms in it, which is the correct compromise for a CV you are not sending anywhere.
The moment you have a specific advert, the question changes from "which terms are good" to "which of this advert's requirements does my CV cover, and where does it say so". That is what Appliora's Job Match does inside the checker: match model version 2.0 returns a verdict per requirement - covered, partial or missing - with the line of your CV that covers it, rather than a percentage of shared words.
The published split is 60% to the advert's must-haves, 15% to the nice-to-haves, 10% to the years of experience it asks for, 8% to languages, 4% to education and 3% to certifications. Must-haves dominate on purpose: a CV covering every nice-to-have and missing two must-haves is not an 80% match by any honest reading.
Tailoring a CV to one advert is the twenty-minute routine that acts on that output. It is the highest-return habit in this cluster and it involves inventing nothing.
Questions people ask
Is there a keyword score in Appliora's model?
There is no check called keywords. Terms are scored where they appear: concrete skills is worth 8 points and bullet quality is worth 13, and a term that proves itself inside a bullet is doing work that the same term in a list cannot do. Matching your CV against one specific advert is a separate tool with its own published model.
How many skills should I list?
Appliora expects at least five and treats a list beyond about forty as stuffing rather than as coverage. The useful number is much closer to the bottom of that range than the top, because a list long enough to include everything is a list that proves nothing about any of it.
Should I put a keywords section in white text at the bottom of the page?
No. Hidden text is still text, so it extracts and it is visible to anyone who selects the page or opens the file in a different viewer. It reads as an attempt to deceive, and the version of you that arrives in the database now claims skills your interview will not support.
Do I need to repeat a keyword for it to count?
Repetition is not what makes a term count. A term used once inside an achievement that shows the scale, the tool and the outcome is stronger evidence than the same term repeated four times, and a dense block of repeated terms is what the skills check reads as stuffing.
What about a keyword I genuinely do not have?
Leave it out. The alternative is a CV that gets you into a conversation you cannot finish. If you half have it - the adjacent tool, the same work at smaller scale - write the true version, because a partial match stated honestly is scored as a partial match rather than as a miss.
See what a parser reads from your CV
Upload your current file and get the extracted text plus a 0-100 score across 14 checks. No signup.