Numbers That Make Recruiters Stop Scrolling

Two resumes, same job, same experience

Picture two candidates applying for the same operations manager role. Both have run warehouse teams for four years. Both know the job cold.

Candidate A writes: "Responsible for managing warehouse staff and improving shipping processes."

Candidate B writes: "Cut average order-fulfillment time from 48 to 31 hours by redesigning the pick-pack workflow, while managing a team of 14 across two shifts."

Same underlying experience. One sentence describes a job description. The other describes a person a hiring manager wants to meet. The recruiter reading forty resumes in an afternoon doesn't know Candidate A didn't do anything — they just can't tell that they did, and in a stack of forty, "can't tell" loses to "obviously did." That's the whole game, and it's decided in about the time it takes to read this paragraph.

Why "responsible for" had to die

"Responsible for" describes an assignment, not an outcome. It tells a recruiter what was on your job description — which is exactly the information they already have, because they wrote the job posting for the next person doing that same job. It answers a question nobody asked ("what were you supposed to do?") and dodges the one that matters ("what happened when you did it?").

It's also the most overused phrase in the genre, which means it does something worse than sound weak — it sounds identical to every other resume in the pile. Recruiters running high-volume searches see it dozens of times a day. It's not a red flag exactly, it's a non-signal: the bullet could describe someone who transformed a department or someone who showed up and existed in a role for two years. The reader has no way to tell which, so they default to assuming the less impressive one.

Eye-tracking research on recruiter behavior (the widely cited Ladders study, later work has landed anywhere from 6 to 11 seconds depending on methodology) consistently finds the same pattern: recruiters make an initial keep/reject decision in single-digit seconds, scanning for a small number of concrete signals — title, company, dates, and whether the bullets say anything. A duty-based bullet gives the eye nothing to catch on. A number does exactly that: it's a different shape on the page, and shape is what gets read first.

The formula: action, then result, then metric

Every strong bullet has three parts, and the order matters:

1. A strong verb that names what you actually did (built, cut, negotiated, migrated, trained) — not a vague verb like "helped," "worked on," or "assisted with."
2. The specific action — what you built, fixed, negotiated, or shipped, described concretely enough that a stranger could picture it.
3. A quantified result — the number that proves it mattered.

You'll see this same shape under a few different names — STAR (Situation, Task, Action, Result), CAR (Challenge, Action, Result), PAR (Problem, Action, Result). They're all the same idea wearing different acronyms: stop narrating the setup and lead with what changed. For a resume bullet specifically, skip the situation/task throat-clearing — you have one line, not a paragraph — and compress straight to verb → action → number.

"Led a cross-functional initiative to improve customer retention" is a CAR bullet with the R missing. "Led a cross-functional initiative that lifted 90-day retention from 61% to 74%" is the same bullet with a spine.

The data backs this up more concretely than most resume advice: quantified achievements are widely reported to make candidates roughly 40% more likely to reach a shortlist compared to resumes built on duty statements alone, and one frequently cited estimate puts the gap in interview invitations at 2.5x. Treat those exact multipliers as directional rather than gospel — they come from recruiting-industry research, not peer-reviewed studies, and the resume industry loves a clean statistic — but the direction is not in dispute anywhere: unquantified resumes blend together, and the ones that don't blend get read.

The credibility line: when does a number sound made up?

Here's the part most advice skips: a bad number is worse than no number. If a bullet reads "increased efficiency by 250%" with nothing else around it, a recruiter's next thought isn't "impressive" — it's "made that up." Numbers only build credibility when they clear three bars:

• Proportion to scope. A 3-person startup team didn't "reduce infrastructure costs by $4M." A number needs to be believable given the size of the team, budget, and company implied by the rest of the bullet.
• A visible denominator. "Grew engagement 400%" from what base? If the honest answer is "5 users to 25 users," the percentage is technically true and completely useless — it's a vanity metric, a number that's real but doesn't mean anything to the person reading it.
• Specificity that only comes from having actually measured it. "31 hours" reads as more credible than "over 40% faster" — not because it's a bigger number, but because oddly specific figures look like they came from a dashboard, not a guess. Round numbers can work too, but stacking several suspiciously round numbers in a row (50%, 100%, 200%) starts to read like fiction.

The manager test is a good gut check: if your old manager read the bullet, would they nod, or would they laugh? If the honest answer is "they'd laugh," the number needs to come down, not the ambition of the bullet.

What to do when you don't have the exact number

Most people don't have a saved spreadsheet from a job they left three years ago. That's fine — estimation is normal and expected, as long as it's a defensible estimate and not an invented one. Two techniques help:

Bracketing. If you think you improved something "by maybe 15%," ask: was it closer to 10% or 20%? Then narrow again: closer to 15% or 20%? You'll usually land on a number you can actually defend in an interview, which is the real test — not whether it sounds good, but whether you can explain where it came from if someone asks.

Reconstructing from what you do remember. You may not know the exact percentage, but you probably remember the team size, the before-and-after state, the frequency of a task, or the scale of what you touched ("supported roughly 40 client accounts," "trained about 25 new hires over two years," "handled 15-20 tickets a day"). Approximate ranges, stated as ranges, are more credible than a suspiciously precise number you can't back up.

The source matters too — dashboards, ticket systems, sales reports, survey results, a manager's written feedback, even your own notes from the time all count as legitimate bases for an estimate. "I made this number up because it sounded good" doesn't.

Before and after, across five jobs

Software engineer

• Before: "Responsible for backend API development and bug fixes."
• After: "Rebuilt the checkout API's caching layer, cutting p95 latency from 800ms to 210ms and eliminating the #1 source of customer-reported timeouts."

Product manager

• Before: "Worked with engineering and design to launch new features."
• After: "Shipped a redesigned onboarding flow across 3 teams, lifting activation from 44% to 58% within one quarter."

Operations manager

• Before: "Oversaw daily warehouse operations and staff scheduling."
• After: "Redesigned shift scheduling for a 20-person warehouse team, cutting overtime costs 22% while holding on-time shipment rate above 98%."

Sales representative

• Before: "Responsible for managing client accounts and closing new business."
• After: "Closed $1.2M in new annual contract value across 30 accounts, exceeding quota by 118% in FY25."

Teacher

• Before: "Responsible for classroom instruction and student assessment."
• After: "Redesigned the algebra curriculum for 5 sections of 28 students, raising average end-of-year proficiency scores from 62% to 79%."

Notice none of these bullets invent a new skill — they take the exact same underlying work and describe the outcome of it instead of the assignment.

Nudging toward the top of the plausible range — carefully

If the honest answer to "was it 20% or 30%?" is genuinely "somewhere in there, I'm not sure," it's reasonable to write the higher end you can defend — 30%, not 20%. That's not lying; that's picking the true number that happens to also be the more flattering one, from within a range you'd stand behind under questioning.

The caveat, and it's a real one: this only works up to the edge of plausibility, and that edge is set by scope, not ambition. A number that would require your team, budget, or timeline to have been meaningfully larger than they actually were isn't a nudge, it's a fabrication with better dressing. If you can't picture explaining the number to the hiring manager in the follow-up interview without flinching, it's too high. The goal is the strongest number that survives a five-minute conversation about it — not the strongest number, period.

The mistakes that undo good numbers

• Vanity metrics. A statistic that's real but doesn't map to anything the reader cares about — "increased Twitter followers by 4,000%" from a base of 12 followers is technically accurate and completely worthless. Ask what the number would mean to someone who's never met you.
• The "we" trap. "We increased revenue 30%" tells a recruiter about your team's quarter, not your contribution to it. If you were one of ten people on the initiative, say so and name your slice: "drove the pricing-model redesign that contributed to a team-wide 30% revenue increase." Own your part; don't borrow the whole team's.
• Denominator-free percentages. Any percentage without a sense of scale ("grew the program 300%") invites the reader to imagine the smallest possible base. Give them the real one, or use a raw number instead.
• Stacking implausible round numbers. Three bullets in a row at exactly 50%, 100%, and 200% reads like a pattern, and patterns get noticed. Real work produces oddly specific numbers; sanitize too much and you erase the thing that made them credible in the first place.
• A number for its own sake. Not every bullet needs a percentage bolted onto it if there's nothing real to report — a bullet describing a genuine before/after state, scale of ownership, or scope of impact can carry weight without a fabricated metric wedged in.

Where the tool comes in

This is also, not coincidentally, the rule we built Penny Resume's tailoring engine around. When you paste in a job posting and let it draft your resume, the model isn't allowed to write a "responsible for" bullet — full stop. We told it: every bullet earns its line with an action verb, a real action, and a defensible outcome, and if there's a genuine number anywhere in your background that fits, use it or nudge it to the honest top of its range; if there isn't one, describe the scope and stakes of the work instead of inventing a metric to fill the space. It's one way to get there — plenty of people do this rewrite by hand, and the exercise above works fine with a blank page and twenty minutes. If you'd rather skip straight to a tailored draft, the browser extension is the fastest path: it reads the job posting you're already looking at and drafts results-focused bullets against it before you've finished your coffee.

Try the rewrite on your own resume

Pick your three weakest bullets right now — the ones that still say "responsible for" or "helped with." For each one, ask: what changed because I did this, and by how much? If you know the number, write it. If you don't, bracket toward a defensible estimate. You'll probably find you already know more than you thought — you just never wrote it down.

If you want a faster pass at it, paste a job posting into Penny Resume and see what a results-focused rewrite of your own background looks like. Either way, the resume that gets read is the one that tells the recruiter what happened, not just what you were asked to do.