# The Resume Skill Trajectory Score: Feature, Keep, or Cut

*You can take any single skill, tool, or credential on your resume and score it on four dimensions to decide whether to feature it, keep it quietly, or cut it.*

- Canonical URL: https://www.refolk.ai/candidates/guides/resume-skill-trajectory-score
- Pillar: Reading the market
- Format: Framework
- Published: 2026-09-17
- Last reviewed: 2026-09-17
- Reading time: 16 min
- Keywords: which skills to put on resume, outdated skills to remove from resume, how to prioritize resume skills, in-demand skills for resume, is this skill worth listing

## Key takeaways

- Scarcity, not trend direction, is the resume signal: Microsoft Office sits at 8,529,737 US profiles in Refolk's index versus 60,665 for Prompt Engineering, a ~141x gap that makes the common skill worthless to feature.
- The declining line keeps moving up: WEF now lists graphic designers among the fastest-declining roles as generative AI reshapes work, and workers can expect 39% of their skills to be transformed or become outdated over 2025 to 2030.
- The skills section is parsed twice and scores 10 to 15 points higher than other sections, and tailoring it per posting adds about 31 points, so cutting a dead line and adding an exact match moves the resume more than editing prose.
- Score each line on four axes - demand direction, posting scarcity, automation exposure, and verifiability - because no public body has published cutoffs that would make two readers grade a skill identically.
- Geography flips the verdict: Prompt Engineering has 60,665 US holders versus 6,050 in Germany, so a differentiator in Berlin can be closer to table stakes in a larger market.
- Cap the finished section at 15 to 25 skills grouped by function in plain text, since fewer than 10 reads as under-qualified and more than 30 reads as keyword stuffing.

You have a skills section and tool names scattered through your bullets, and you need to decide, line by line, which to lead with, which to leave quietly in place, and which are dead weight. This guide is for a job seeker who is done copying someone else's "best skills" list and wants a method for grading the actual lines on the page in front of them. It gives you a repeatable per-line score built on four dimensions, so you finish with a verdict for every item: feature, keep, or cut.

The reason a generic list fails you is simple. A list hands you a verdict; it cannot tell you whether the specific skill on your resume is scarce in your target market, whether you can defend it in an interview, or whether it survives an applicant tracking system parse. Those answers depend on your lines and your postings. The score below produces them.

## Why score each line instead of copying a list

A published "in-demand skills" list gives you someone else's answer to a different question. The question that matters is whether a specific line on your resume earns its place, and that depends on how rare the skill is in your target postings, whether you can prove it, and whether an ATS reads it at all.

The core mistake the listicles make is treating demand direction as the whole story. It is not. A skill can be rising in demand and still be worthless on your resume because everyone already holds it. In Refolk's index of professional profiles, Microsoft Office sits at 8,529,737 holders in the United States. Prompt Engineering sits at 60,665. That is a roughly 141x gap. This is why every "remove these skills" article tells you to cut Office - not because office software is declining, but because ubiquity reads as zero differentiation. A recruiter scanning your section spends that line's attention learning nothing new about you.

**141x - How much more common Microsoft Office is than Prompt Engineering among US profiles**

Microsoft Office holders 8,529,737 versus Prompt Engineering 60,665 in Refolk's index. Ubiquity, not decline, is why the common skill gets cut.

Public data stops at the occupation. The BLS Employment Projections program does not produce skills data at the state or local level, and the WEF survey works at the role level. Neither tells you how rare your specific tool is in your specific market. That is exactly the gap holder counts fill, and it is the private denominator that makes per-line scoring possible rather than guesswork.

## The four dimensions and what each one proves

Score every line on four axes: demand direction, posting scarcity, automation exposure, and verifiability. Each proves something different, and each has a way of lying to you that you need to watch for.

- **Demand direction** proves whether the market is moving toward or away from the skill. It lies when you read a national trend as a local verdict: a skill declining across the country can still be required in your niche.
- **Posting scarcity** proves whether the skill differentiates you or is merely expected. It lies when a skill you value personally is one that nearly every candidate holds, so it feels like an asset and signals baseline.
- **Automation exposure** proves whether the skill is drifting into the displaced cohort. It lies by lagging: a skill that was safe in an older report may sit inside the exposed group now.
- **Verifiability** proves whether you can defend the line and whether the ATS can score it. It lies when a soft-skill phrase looks like a skill but exact-matches nothing and proves nothing.

No external body has published cutoffs that would make two readers grade a skill identically. That gap is the honest limit of this method. What the sources do establish is the direction of each axis, so the scale below is one I have constructed to be applied consistently, not one you will find stamped on a government table.

#### What each dimension sits on

1. **Demand direction** - Broad public trend from BLS, WEF, and LinkedIn - is the skill rising, flat, or declining
2. **Automation exposure** - The declining-role categories - clerical, data entry, word processing, generic design
3. **Posting scarcity** - How many of your 8 to 12 target postings name it, backed by holder counts
4. **Verifiability** - Can you defend it in five minutes and does it exact-match an ATS string

*The four axes stack from broad public trend down to the private, market-local, and personal facts that actually decide a line.*

### Where to read demand direction

Three public sources cover demand direction, and they measure different things. Use all three; do not trust one alone.

| Source | Basis | Granularity | Update |
|---|---|---|---|
| BLS Employment Projections | Model plus O*NET | 800+ occupations, 17 skills, national only | Annual |
| WEF Future of Jobs | Employer survey | 1,000+ firms, 55 economies, role-level | Roughly biennial |
| LinkedIn Skills on the Rise | Platform behavior | 12 markets, skill-level | Annual |

BLS gives you the modeled long view: for the 2023-33 cycle it built a framework of 17 skills and mapped O*NET elements to them for more than 800 occupations. WEF gives you what over 1,000 employers say they expect across 55 economies from 2025 to 2030. LinkedIn measures what professionals actually do on the platform, comparing year over year. When the three agree, the direction is solid. When they disagree, weight the one closest to your target role and market.

## How to read scarcity from your own postings

Scarcity is the share of your target postings that name a skill, read against how many people hold it. A skill that appears in nearly every posting and that millions hold is table stakes; a skill that appears in few postings and that few people hold is a differentiator. No published source gives a numeric frequency cutoff, so you read the share qualitatively.

The documented method is to analyze each job description, pull out the hard skills, tools, and industry terms mentioned repeatedly or listed as requirements, and sort them by must-have versus nice-to-have. Do that across 8 to 12 postings for the role you are aiming at, and a pattern emerges: the terms in nearly every posting are the price of entry, and the terms in only a few are where you can stand out.

The holder counts sharpen this. Two skills can appear in the same number of postings but sit at wildly different holder counts, and the rarer one is the stronger line.

| Skill | Country | Holders | Scarcity read |
|---|---|---|---|
| Microsoft Office | US | 8,529,737 | Table stakes |
| Prompt Engineering | US | 60,665 | Differentiator |
| Prompt Engineering | Germany | 6,050 | Differentiator |
| MS Office ÷ Prompt Eng (US) | - | 140.6x | Derived |
| Prompt Eng US ÷ Germany | - | 10.0x | Derived |

Geography flips the verdict. Prompt Engineering has 60,665 holders in the US and 6,050 in Germany - a 10x gap. A rising skill that is a genuine differentiator in a smaller market can be closer to crowded in a larger one, so score scarcity local to the market you are applying in.

> **Rule:** Scarcity, not trend, is the resume signal
>
> A skill can be rising in demand and still be worthless to feature if everyone already holds it. Feature a line only when it is scarce in your target postings, not merely trending in the headlines.

If you want to see how crowded a rising skill already is in your exact target role before you feature it, a people search reads the field faster than counting postings by hand.

Ask me this: `Data analysts in the United States who list Prompt Engineering and dbt but not Microsoft Office` - [run the search](https://www.refolk.ai/start?q=Data%20analysts%20in%20the%20United%20States%20who%20list%20Prompt%20Engineering%20and%20dbt%20but%20not%20Microsoft%20Office).

*Returns how many people in your target role already hold the rising skill you are considering featuring, so you can tell a real differentiator from a crowded one.*

## What the ATS does to your skills section

The skills section is the highest-scoring section on a resume because most applicant tracking systems parse it twice - once as a structured field and once in full-text search. Resumes with a dedicated skills section average 10 to 15 points higher, and tailoring that section per posting adds about 31 points on average. This mechanical over-weighting is why disciplined per-line scoring pays off: cutting a dead line and adding an exact job-description match moves the resume more than rewriting a paragraph.

A modern ATS runs two layers. A parser extracts structured fields - contact, sections, dates, titles, skills, education. A matcher then compares those fields against the job description and the recruiter's filters. The matching is literal. "Cross-functional collaboration" and "stakeholder communication" are different strings, and only the one that appears in the job description scores. That is why soft-skill phrasing earns nothing in the section and why synonym mismatches quietly cost you: "JS" is not "JavaScript" to the matcher.

#### How a skill line becomes a score

1. **Parse** - The system extracts your skills section as a structured field and again in full-text search
2. **Tokenize** - Each skill becomes a string; bars and meters can parse as empty text
3. **Match** - The matcher does exact-match scoring against the job description strings
4. **Score** - Matches add points; the skills section averages 10 to 15 points higher than others

*Two ATS layers stand between your line and a match, and a line can die at either one.*

75% of employers use an ATS, rising to 99% among Fortune 500 companies, so this is not an edge case. Format matters as much as content here: proficiency bars and percentage meters are parsed as empty text by 30% of ATS systems, so a skill you display with a graphic can vanish entirely.

### How many skills to keep

Published caps disagree, so pick a defensible middle and hold it. The safe range is 15 to 25 skills grouped by function.

| Source | Recommended count |
|---|---|
| hireflow.net | 15-25 |
| resumeoptimizerpro.com | cap at 20 |
| resumehog.com | 8-12 |
| writecv.ai (senior) | 20-30 |

Fewer than 10 can read as under-qualified; more than 30 reads as keyword stuffing that dilutes your strongest matches. The spread across sources runs 8 to 30, but 15 to 25 keeps you inside every reasonable range and forces you to cut the weakest lines.

## The scoring procedure

Run these steps in order on the list you inventoried. The first pass takes roughly two hours; every application after that costs 3 to 5 minutes.

#### Score every line, then rebuild the section

1. **Inventory the lines** - List every skill, tool, and credential as its own line, including tool names buried in your experience bullets. You want a flat list where each item can be scored on its own.
2. **Tag demand direction** - Cross-check each line against BLS Employment Projections, the WEF fastest-growing and declining lists, and LinkedIn Skills on the Rise. Mark it rising, flat, or declining and note the source.
3. **Score posting scarcity** - Collect 8 to 12 postings for your target role, count how many mention each skill, and read the share as table stakes if it appears in nearly all or a differentiator if it appears in few.
4. **Score automation exposure** - Flag any line matching WEF or BLS declining categories such as clerical work, data entry, word processing, or generic design. Clear the flag on the rest.
5. **Score verifiability** - Apply the five-minute-conversation test and confirm the line is a searchable exact-match hard skill rather than unprovable fluff. Tag each keep, prove-in-bullets, or cut.
6. **Combine into a verdict** - Merge the four tags per line. Feature when rising, a differentiator, and verifiable; cut when declining or table-stakes-baseline and unverifiable; keep everything in between.
7. **Reformat and cap for ATS** - Rebuild the section as plain text, comma or line-break separated, grouped by function, and capped at 15 to 25 skills. Confirm every line survives a plain-text paste.
8. **Re-tailor per posting** - For each application, spend 3 to 5 minutes reordering and swapping to mirror that posting's exact strings and priorities.

On step 5, the verifiability test is one practitioner rule: can you have a five-minute conversation about every skill listed? If not, remove it. Soft skills score nothing in ATS keyword matching, so list only hard skills in the dedicated section and prove the soft ones in experience bullets with numbers and outcomes.

Note one genuine tension between steps 3 and 5. Scarcity scoring will sometimes flag a soft skill as a differentiator because few postings name it, while the ATS reality is that soft skills earn nothing as section strings. Resolve it in favor of the ATS: a soft skill never goes in the section, no matter how its scarcity reads. It goes in a bullet with an outcome attached.

## Reading the verdict: the feature, keep, cut map

The verdict comes from crossing two axes: demand direction against posting scarcity, with verifiability as the gate. A line is only ever featured if it clears the verifiability gate first.

#### Feature, keep, or cut

Horizontal axis runs from Table stakes (everyone holds it) to Differentiator (few hold it). Vertical axis runs from Declining demand to Rising demand.

| Quadrant | What it means |
| --- | --- |
| Declining and common | Cut - it dates you and differentiates nothing |
| Declining but scarce | Keep quietly if your niche still requires it; check your postings, not the trend |
| Rising but common | Keep, list only when the posting names it, never feature |
| Rising and scarce | Feature if verifiable - this is your strongest line |

*Cross demand direction against scarcity, then gate every "feature" on whether you can defend it in five minutes.*

Read it as a decision, not a description. Rising and scarce, and you can defend it, is the top-left of your section. Rising but common - the trending skill everyone now holds - stays in the section only when a posting names the exact string, and never leads. Declining and common is dead weight; cut it. Declining but scarce is the one that needs judgment: COBOL is declining nationally but still required in banking, so the verdict depends on your target postings, not the macro trend.

> A skill can be rising and worthless on your resume if everyone already holds it.

## Where this goes wrong

The score fails in predictable ways, and each failure has a false positive - a line that looks right and is not. These checks are the most valuable part of the method because they catch the errors a clean-looking section hides.

- **Table stakes mistaken for a differentiator.** A skill 8.5 million US profiles hold feels valuable to you and signals baseline to a recruiter. Check: if it appears in nearly every target posting and has a huge holder count, it is table stakes. Keep it only when a posting names it, and never feature it.
- **The rising-skill halo without verifiability.** Adding a hot keyword like Prompt Engineering because it trends is a false positive if you cannot pass the five-minute test. Check: run the conversation test before you feature anything trending.
- **Cutting a legacy skill that is still required.** Declining at the national level does not mean dead in your niche. Check: score scarcity in your actual target postings, not the BLS trend line.
- **Soft-skill fluff scored as a hard skill.** "Team player" counted as a match is a false positive; soft skills score nothing in ATS matching. Check: does the line exact-match a hard-skill string in the job description?
- **Format kills a good skill.** A proficiency bar looks listed but is parsed as empty text by 30% of ATS systems. Check: paste the section into a plain-text editor and confirm every skill survives as words.
- **Synonym mismatch.** "JS" versus "JavaScript," hyphenated versus not. Check: mirror the job description's exact string, character for character.
- **Over-stuffing.** More than 30 lines reads as keyword dumping and dilutes your real matches. Check: cap to 15 to 25 and cut the weakest.

> **Watch out:** The declining line keeps moving up the ladder
>
> WEF now lists graphic designers among the fastest-declining roles as generative AI reshapes work, alongside cashiers, administrative assistants, bank tellers, and data entry clerks. A skill that felt safe in an older report may be inside the current 39% skill-churn cohort, so re-score rather than trust a past verdict.

The scale of that churn is worth holding in view while you cut. On average, workers can expect that 39% of their existing skill sets will be transformed or become outdated over 2025 to 2030, up from the 35% employers expected when the report first ran in 2016.

## Rebuilding the section and keeping it current

Once every line has a verdict, rebuild the section as plain text, grouped by function, capped at 15 to 25 skills, with the featured lines first. Here is the shape to copy.

**Plain-text skills section skeleton**

```
SKILLS

[Function 1, e.g. Analytics]: [featured differentiator], [featured differentiator], [table-stakes skill named in the posting]
[Function 2, e.g. Tools]: [exact JD string], [exact JD string], [exact JD string]
[Function 3, e.g. Methods]: [featured differentiator], [supporting skill]

Notes:
- Hard skills and named tools only. No soft skills, no bars, no meters.
- Match each string to the posting exactly: "JavaScript" not "JS".
- Total 15 to 25 lines. Cut the weakest match before you exceed 30.
```

*Group by function, put featured (rising and scarce) skills first in each group, and mirror the target posting's exact strings.*

Then verify before you call it done.

#### Before you send the section

- [ ] Every line has one verdict: feature, keep, or cut.
- [ ] No cut line survived, and no soft-skill phrase is in the dedicated section.
- [ ] Featured lines are rising, scarce in your target postings, and defensible in a five-minute conversation.
- [ ] Every skill string matches the target posting's exact wording, including hyphenation.
- [ ] The section pastes into a plain-text editor with every skill intact and no bars or meters.
- [ ] The total count sits between 15 and 25 skills, grouped by function.
- [ ] A per-posting variant reorders the top lines to that posting's must-haves.

Keep it current by re-scoring, not by trusting your last verdict. The declining line moves up the ladder over time, holder counts shift as skills spread, and each new posting emphasizes different strings. Re-run demand direction against the three sources once or twice a year, re-count scarcity whenever you switch target roles or markets, and re-tailor per application. This is where [Refolk](/candidates) removes the manual work: it writes your resume from your own history, tailors the skills and bullets to every posting you apply to, and scores how well you actually fit, so the per-line re-scoring happens on every application instead of once a year.

Using one static skills list for every application is a missed opportunity, because each job description emphasizes different skills and spending a few minutes reordering meaningfully improves your keyword match rate. The score you built is not a one-time verdict. It is the judgment you apply to every posting, every time you decide whether a line earns its place.

## Frequently asked questions

### Which skills should I put on my resume?

List the ones that are rising or stable in demand, scarce in your target postings, and verifiable in a five-minute conversation. A skill earns a spot when it exact-matches a job description string and few candidates hold it. Ubiquitous baselines like Microsoft Office, which sits at 8,529,737 US profiles in Refolk's index, belong in the section only when a posting names them, and never at the top.

### What outdated skills should I remove from my resume?

Cut universal basics such as Word, basic Excel, and email, obsolete tech such as Flash and legacy CRM versions, and vague soft-skill fluff you cannot prove with a number. Also cut anything in the WEF and BLS declining categories - clerical, data entry, word processing, generic design - unless your specific target postings still require it. Declining nationally does not mean dead in a niche, so check your postings, not the headline.

### How do I know if a skill is worth listing?

Score it on four axes: demand direction from BLS, WEF, and LinkedIn; posting scarcity from your own count of 8 to 12 postings; automation exposure from the declining-role lists; and verifiability from the five-minute test. Feature it only if it is rising, scarce in your postings, and defensible. If it fails scarcity or verifiability, keep it quiet or cut it.

### How many skills should a resume list?

Aim for 15 to 25 skills grouped by function. Fewer than 10 can read as under-qualified and more than 30 reads as keyword stuffing that dilutes your signal. Published ranges disagree, spanning 8 to 30 across sources, but the safe middle is 15 to 25, capped hard and re-tailored per posting.

### Does listing a trending skill like prompt engineering help?

Only if you can defend it. A rising, scarce skill is the strongest thing you can feature, but the halo fails the moment you cannot pass a five-minute conversation about it. Add it when you have real experience to point at, mirror the job description's exact string, and prove it in an experience bullet with an outcome. A hot keyword you cannot back up is a false positive that costs you in the room.

### Do soft skills belong in the skills section?

No. Soft skills score nothing in ATS keyword matching because the matcher does exact-match scoring on hard-skill strings. Keep the dedicated section to hard skills and tools, and prove soft skills inside experience bullets with numbers and outcomes. Listing 'team player' as a skill counts as nothing and spends a line you could give to a defensible match.

---

*From the Refolk guide library. I revise these guides rather than replacing them, so the current version is always at https://www.refolk.ai/candidates/guides/resume-skill-trajectory-score*