# The Duties-to-Evidence Resume Rewrite, One Draft Carried Through

*You can take one duties-based resume and rewrite each bullet into a defensible achievement, choosing a number, a labeled estimate, or plain scope, and cutting the lines that fail.*

- Canonical URL: https://www.refolk.ai/candidates/guides/duties-to-evidence-resume-rewrite
- Pillar: Positioning and materials
- Format: Teardown
- Published: 2026-08-05
- Last reviewed: 2026-08-05
- Reading time: 16 min
- Keywords: turn job duties into accomplishments, quantify resume bullets without numbers, how to rewrite resume bullet points, resume reads like a job description, defensible resume metrics

## Key takeaways

- The recruiter tell is not the number, it is the mismatch: about one in three recruiters claim to spot an AI resume in under 20 seconds, mostly through metric-free vagueness and resume/interview gaps.
- Scope beats a forced percentage because it is self-verifying: a baseline-free 20 percent has no audit trail, while 12-location brand, 5 posts per week carries its own.
- The present-tense test is a near-mechanical duty detector: if a bullet reads Manage rather than Managed, you probably listed a responsibility, not an accomplishment.
- Volume is capped at roughly 3 to 5 bullets for a recent role, so every forced weak metric crowds out a real, defensible one.
- In Refolk's index, roughly 109,692 US recruiters face about 5,169 US career coaches and resume writers, a 21-to-1 ratio, so most job seekers rewrite without expert review.
- UC San Diego sanctions labeled estimates like approximately 50 and over 75, so you can quantify scope without inventing precision you cannot defend.

Your resume reads like a job description because it was written from one. This guide is for anyone whose bullets list responsibilities instead of results, and who wants to fix that without inventing numbers that fall apart in the interview. I take one same-field resume from duties to evidence, bullet by bullet, showing the real intermediate drafts and the wrong turns, so you can follow along on your own page rather than copy someone else's numbers.

The core move is a per-bullet fork. Each surviving line gets a hard number, a labeled estimate, or plain scope, and the ones that carry no answer to "so what?" get cut. That decision, made line by line, is the whole job.

## Why a duties resume fails, and what actually breaks in the room

A duties resume fails because it lists what you were told to do, not what changed because you did it. The specific thing that breaks later is not the missing metric. It is defensibility: a number you cannot reconstruct out loud.

Here is the distinction the whole rewrite turns on. The Muse states it cleanly: "A duty describes what you did and an accomplishment describes how well you did it." Their example contrasts "planned events" with "raised $100,000 by selling out tickets to a 200-person charity event." The first is a responsibility any holder of the title would share. The second is yours.

The trap people fall into next is over-correcting into fiction. AI tools now generate a convincing "reduced reporting cycle time by 35%" on demand, which is why the documented recruiter tell is a mismatch, not a missing digit. About one in three recruiters say they can spot a ChatGPT resume in under 20 seconds by looking for "generic buzzwords like leverage and spearhead, vague accomplishments without specific metrics, uniform sentence structures, and a mismatch between resume polish and interview performance." The number on the page reads fine. It collapses when someone asks you to walk through it.

> The defensibility gap, not the metric gap, is what breaks a resume in the interview room.

So the standard this guide holds every bullet to is not "does it have a number?" It is "can you give the one-sentence backstory aloud?" That reframing changes which lane each bullet belongs in.

## The three evidence lanes, and what each proves

Every surviving bullet lands in one of three lanes: a hard number, a labeled estimate, or plain scope. Each is endorsed by a named career office, and each proves something slightly different.

| Evidence lane | Named endorser | Example phrasing |
|---|---|---|
| Hard number | MIT / Harvard | "$500k increase in annual revenue" |
| Labeled estimate / range | UCSD / The Muse | "approximately 50", "7-12 each year" |
| Plain scope | UCSD / UC Davis | "weekly", "audience of 50+" |

A hard number proves magnitude, but only if you can source it. It lies when it is a baseline-free percentage: "increased sales 20%" looks quantified and is uncheckable, because there is no "from what" to test it against.

A labeled estimate proves you know your own scale without pretending to precision you lack. UC San Diego sanctions this directly: "It's okay not to have an exact number - think phrases like 'weekly', 'hour-long', 'approximately 50', 'over 75'... Numbers enhance credibility and give an idea of scope." The Muse uses the range form: "Supervised 7-12 undergraduate research students each year." An estimate lies when you strip the label and write "~50" as a flat "50", which invites an interview challenge you cannot win.

Plain scope proves ownership through concreteness, and it has a quiet advantage: it is self-verifying. Consider the documented rewrite of a weak line. "Managed social media accounts" becomes "Managed LinkedIn, Instagram, and X accounts for a 12-location retail brand, writing 5 posts per week" - "strong without a single number." That bullet contains its own audit trail. A baseline-free "20%" does not.

> **Rule:** Every number needs a spoken backstory
>
> Before a figure goes on the page, say its one-sentence origin aloud: where it came from and how you would reconstruct it. If you cannot, drop the bullet to scope or cut it.

## The worked example: one duties resume, carried through

Here is the draft I will carry through the whole guide. It is a marketing coordinator's resume, written entirely as duties, which is the most common form of the problem.

**The starting draft - four duties-based bullets**

```
- Responsible for managing the company's social media accounts.
- Manage a team of 5 marketing assistants.
- Assisted with quarterly reporting for the marketing department.
- Helped increase sales through various campaigns.
```

*These are the "before" lines. Every section below edits these same four.*

Run the first tag on each. The duty-versus-accomplishment test asks whether the line says how well, not just what. The present-tense check backs it up: "if you use the present tense, there's a pretty good chance you've just listed a responsibility - it's hard for something you have accomplished to be written in the present tense (e.g. use 'Managed', not 'Manage' nor 'Managing')."

- Bullet 1 starts with "Responsible for." Pure duty.
- Bullet 2 reads "Manage." Present tense, so a responsibility, not an accomplishment.
- Bullet 3 opens with "Assisted with." No ownership, no result. Duty.
- Bullet 4 says "Helped increase." Vague verb, no scope, no anchor. Duty.

Four for four are duties. That is the honest starting count, and it is normal. Now apply the "So what?" test: what changed because of the action?

- Bullet 1: managed accounts, and... nothing stated. Fails. Rewrite to scope.
- Bullet 2: managed people, and... the team did something. Recoverable if I can name the outcome the team produced.
- Bullet 3: assisted with reporting, and... this is the classic vague survivor. It fails "so what?" outright.
- Bullet 4: increased sales, and... this is where a forced percentage will tempt me.

**21x - US recruiters per career coach in Refolk's index**

Roughly 109,692 recruiters against 5,169 career coaches and resume writers, so most job seekers rewrite without expert review.

That ratio is why the rule has to be self-applicable. In Refolk's index there are about 109,692 US profiles with recruiter or technical-recruiter titles - top employers include Experis and Robert Half - against roughly 5,169 US profiles carrying career-coach or resume-writer titles, with Google among their prior employers. You are outnumbered by screeners about 21 to 1, and almost none of them is on your side of the table. The per-bullet rule below is the realistic intervention.

## The per-bullet decision: which lane, which rewrite

For each surviving bullet, decide the lane before you write. The judgement call has two axes: how defensible your figure is, and whether a real outcome actually exists to point at.

#### Choosing an evidence lane per bullet

Horizontal axis runs from No defensible figure to Defensible figure in hand. Vertical axis runs from No real outcome existed to A real outcome existed.

| Quadrant | What it means |
| --- | --- |
| Plain scope | Name volume, frequency, audience, or tools; do not force a number |
| Labeled estimate | Use approximately, a tilde, or a range, and keep the label visible |
| Cut or demote | If nothing changed and nothing is countable, the line is a duty; cut it |
| Hard number | State the exact figure with its baseline; this is your anchor |

*Pick the lane by whether a real outcome exists and whether you can defend a figure for it.*

Now walk the four bullets through this matrix, with the intermediate drafts and the wrong turns kept in.

### Bullet 1: "Responsible for managing the company's social media accounts."

No outcome was stated and I have no defensible engagement figure, so this lands in plain scope. The wrong turn here is inventing "grew followers by 40%" - a baseline-free number I could not source. Instead I name the platforms, the brand size, and the cadence.

Rewrite: "Managed LinkedIn, Instagram, and X accounts for a 12-location retail brand, publishing 5 posts per week." Zero percentages, full self-verification, and it opens with a past-tense action verb.

### Bullet 2: "Manage a team of 5 marketing assistants."

A real outcome existed - the team produced work - and I can defend the headcount and the deliverable, so this can reach a hard number if the outcome is countable, or a labeled estimate if it is not. The wrong turn is leaving it in the present tense as a standing responsibility. First pass: "Led a team of 5 marketing assistants." Better tense, still a duty, because it stops at scope of responsibility with no result. Second pass adds the outcome I can actually defend: "Led a team of 5 marketing assistants delivering the brand's weekly content calendar, cutting turnaround from 4 days to 2." The 4-to-2 is a real before/after I can walk through, so it is a hard number with its baseline attached.

### Bullet 3: "Assisted with quarterly reporting for the marketing department."

This is the vague survivor, and it is exactly the line the documented inflation example warps: "'Assisted with quarterly reporting' becomes 'Led a data analytics initiative that reduced reporting cycle time by 35%.'" The employer and dates stay real while the scope and impact are inflated. That 35% is the trap - AI makes it "more convincing, more specific, and harder to detect." If I did not lead an analytics initiative and cannot reconstruct 35%, I must not write it.

What I can defend: I built the campaign-performance section of the report each quarter. So the honest rewrite is a labeled estimate on scope: "Compiled the campaign-performance section of quarterly marketing reports covering approximately 20 active campaigns, presented to department leadership." "Approximately" stays visible because 20 is my recollection, not a figure I pulled from a system.

### Bullet 4: "Helped increase sales through various campaigns."

A real outcome may have existed, but I cannot defend a sales-lift figure I did not own, so a hard percentage is out. The wrong turn is "increased sales 20%" - the baseline-free forced percentage that "looks quantified but is uncheckable." I demote to what I can defend: the campaigns and the audience. "Ran 6 email and paid-social campaigns reaching an audience of over 50,000, coordinating creative and scheduling end to end." That is scope with an audience figure I can source, and it survives a challenge.

> **Watch out:** The forced percentage is the most common wrong turn
>
> "Increased sales 20%" with no baseline reads as an achievement but has no audit trail. If you cannot state the "from what" number, drop the line to scope. A defensible "audience of over 50,000" beats an unverifiable "20%" every time.

## The procedure, start to finish

Run these seven steps in order on your own draft. The first three sort and decide; the last four rewrite and verify. Budget roughly two hours for a one-page resume the first time.

#### Duties to evidence, seven steps

1. **Inventory the duties draft** - List every current bullet verbatim, then tag each as duty or accomplishment using the what-versus-how-well test plus the present-tense check. Done when each line has a tag and you have counted the duties.
2. **Apply the So What test per bullet** - Ask what changed because of the action; mark any bullet that produces no answer for cutting or a scope-only rewrite. Done when every bullet either survives with a reason or is flagged.
3. **Choose an evidence lane per bullet** - For each surviving bullet decide hard number, labeled estimate, or plain scope. Done when every bullet is assigned exactly one lane.
4. **Rewrite using a structural formula** - Apply Action Verb plus Task and Context plus Result, following MIT PAR or the UCSD pattern. Done when no bullet starts with Responsible for or Helped with and no pronouns remain.
5. **Attach or label the metric** - Exact figure where defensible, tilde or approximately or range where estimated, scope words where neither fits. Done when every number traces to a real source you could defend.
6. **Verify line by line** - Read every line, confirm the claim is true, and delete any tool or metric you cannot defend. Done when the full read-every-line pass is complete.
7. **Run the fabrication audit** - Read as a skeptical recruiter and flag any bullet that is both vague and suspiciously round, plus any resume/interview gap. Done when each number has a one-sentence backstory you could say aloud.

Two structural formulas do the same work from different directions. MIT uses PAR - Project, Action, Result - and extends the task "to an outcome to demonstrate the purpose or the 'why?' behind it." UC San Diego frames it as Action Verb plus Specific Task plus Purpose or Result. Pick one and apply it consistently.

#### The rewrite path for a single bullet

1. **Tag** - Mark the line duty or accomplishment by what-versus-how-well and tense
2. **So What gate** - If nothing changed, cut or demote to scope; otherwise continue
3. **Pick lane** - Hard number, labeled estimate, or plain scope
4. **Rewrite** - Action verb plus task plus result, no pronouns, past tense
5. **Source and defend** - Attach the figure with its label and a spoken backstory

*Each bullet moves left to right; a bullet that stalls at the So What gate is cut, not padded.*

There is one order disagreement worth knowing. Sources built on the Google XYZ or QVIR formulas put quantification first. STAR and PAR sources put action and context first and quantify last, "where possible." I follow the PAR order here because deciding the lane before you write stops you reaching for a number the sentence does not support.

## How this goes wrong: the failure modes

The rewrite fails in seven predictable ways, and each has a false positive - a bullet that looks fixed but is not. This section is the one to reread before you send.

| Failure mode | What it looks like | The check |
|---|---|---|
| Forced percentage, no baseline | "increased sales 20%" | Can you state the "from what"? If not, drop to scope |
| The "assisted with" survivor | Humble, safe, no ownership | Does it name a specific action you owned? |
| AI-invented metric | A suspiciously precise "35%" | Give the backstory aloud; if you can't, cut it |
| Present-tense duty in disguise | "Manage a team of 5" | Rewrite in past tense with an outcome; if none follows, it was a duty |
| Estimate presented as exact | "~50" written as "50" | Label with approximately, tilde, or range |
| Over-quantification clutter | Every line stuffed with digits | Keep only metrics that add meaningful context |
| Scope hiding a real result | Tools named to dodge an outcome you have | Was a genuine outcome available and skipped? Prefer it |

The most dangerous of these is the AI-invented metric, because it is the one the modern rejection targets. A 2025 survey of 874 HR professionals found "seventy-two percent of recruiters have already encountered AI-generated fake applications," and a separate 2025 survey found "62% of employers reject AI-generated resumes that lack personalization." The buzzword-plus-fake-metric bullet is now a liability, not a polish. The practitioner rule is blunt, from Mark Horstman of Manager Tools: "Don't BS numbers you can't justify."

The subtler failure is the seventh one, which cuts the other way. Scope is a valid fallback, not a hiding place. If you actually have a defensible outcome and you bury it under a tool list to avoid stating it, you have traded down. Reach for scope when you have no defensible number, not when you have one and feel shy about it.

And do not over-correct into clutter. As one source puts it, "Too many statistics can overwhelm the reader. Balance your bullet points with clear language and only include metrics that add meaningful context." This matters because volume is capped: sources converge on "3-5 strong, quantified bullets for recent roles; 2-3 for older positions." With only that many lines per role, every weak forced number crowds out a real anchor. Pruning is as load-bearing as quantifying.

## Verify before you send

Before you call the rewrite done, run this pass. It is the "read every line" verification plus the fabrication audit, folded into one checkable list. The instruction from the sources is direct: "Read every line of your final resume and confirm that every claim is true and verifiable."

#### Final verification pass

- [ ] No bullet starts with "Responsible for," "Assisted with," or "Helped with"
- [ ] No bullet is written in the present tense, and no personal pronouns remain
- [ ] Every hard number states its baseline or its "from what"
- [ ] Every estimated figure carries "approximately," a tilde, or a range
- [ ] You can say a one-sentence backstory aloud for each number on the page
- [ ] Each recent role has 3 to 5 bullets, not more, and no line is stuffed with digits
- [ ] No tool or metric remains that you could not defend if asked to walk through it
- [ ] Every line that failed "So What?" is either cut or rewritten to defensible scope

The final four bullets from the worked example now read as evidence, not duties:

**The rewritten draft - the same four bullets, carried through**

```
- Managed LinkedIn, Instagram, and X accounts for a 12-location retail brand, publishing 5 posts per week.
- Led a team of 5 marketing assistants delivering the weekly content calendar, cutting turnaround from 4 days to 2.
- Compiled the campaign-performance section of quarterly reports covering approximately 20 active campaigns, presented to leadership.
- Ran 6 email and paid-social campaigns reaching an audience of over 50,000, coordinating creative and scheduling end to end.
```

*Compare against the starting draft; each line is now a lane decision you can defend.*

Two lanes are hard numbers with baselines, one is a labeled estimate, one is scope with a sourced audience figure. None is a forced percentage. Every line survives the question "how did you get that?"

## Keeping the rewrite honest as you apply

Rewrite once, then re-check against each posting rather than each year. The lane a bullet belongs in does not change, but which bullets you lead with does, and the fabrication audit has to run every time you tailor.

The one habit that keeps this current: when a real number arrives - a campaign result, a headcount change, a cycle-time cut - write its backstory down the day you learn it, while you can still reconstruct it. A figure you captured with its source is defensible forever. A figure you reach for months later, unsourced, is exactly the "35%" that collapses in the room.

If tailoring each posting by hand is the friction, that is the part I built [Refolk](/candidates) to remove: it drafts your resume from your own history, rewrites it against each posting, and scores how well you actually fit, so the lane decisions above carry forward without you re-doing them per application. It does not invent your numbers. It keeps the defensible ones in front.

Ask me this: `Career coaches and resume writers in the New York metro area who previously worked at Google` - [run the search](https://www.refolk.ai/start?q=Career%20coaches%20and%20resume%20writers%20in%20the%20New%20York%20metro%20area%20who%20previously%20worked%20at%20Google).

*Returns named coaches with Google in their history, if you want a human review of your rewrite before you send.*

You now have the full method: tag every bullet, gate it on "so what?", pick a lane, rewrite with a formula, label the metric, and audit as a skeptic. Run it on your own page the way I ran it on this one, and cut anything you cannot say out loud.

## Frequently asked questions

### How do I quantify resume bullets when I never worked with numbers?

Use scope instead of invented metrics. UC San Diego endorses phrases like weekly, hour-long, approximately 50, and over 75, and UC Davis lists timeframe, number of people, budget, and caseload as valid context. A line like managed LinkedIn, Instagram, and X for a 12-location retail brand, writing 5 posts per week is strong without a single percentage, because it carries its own audit trail and cannot be challenged the way a baseline-free 20 percent can.

### What share of my resume bullets should have numbers?

Sources disagree and none is authoritative. One vendor cites unlinked eye-tracking research claiming about 80 percent, while Resume Worded takes the softer line of quantifying most bullets. The more useful ceiling is volume: aim for 3 to 5 strong bullets on recent roles and 2 to 3 on older ones. Because each forced metric crowds out a defensible one, prune first and let real anchors earn their place rather than hitting a percentage target.

### How do recruiters tell a resume metric is fabricated?

They look for a mismatch, not the number itself. About one in three recruiters say they can flag a ChatGPT resume in under 20 seconds, citing generic buzzwords like leverage and spearhead, vague accomplishments without specific metrics, and a gap between resume polish and interview performance. The documented failure is a suspiciously specific figure you cannot reconstruct out loud, so the fix is provenance per number, not more numbers.

### What is the difference between a duty and an accomplishment on a resume?

A duty describes what you did and an accomplishment describes how well you did it. The Muse contrasts planned events with raised 100,000 dollars by selling out tickets to a 200-person charity event. A quick mechanical check is tense: if a bullet reads in the present, like Manage a team of 5, it is probably a responsibility, because things you have completed are naturally written in the past, like Managed.

### How do I label an estimate so it is honest but still strong?

Mark the estimate visibly. UC San Diego sanctions approximately 50 and over 75, a vendor documents the tilde convention such as ~1M for a rounded figure, and The Muse uses ranges like supervised 7-12 undergraduate students each year. Writing an inferred number as if it were exact invites an interview challenge you cannot win, so any figure you did not read off a source should carry approximately, a tilde, or a range.

---

*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/duties-to-evidence-resume-rewrite*