80+ resume bullet point examples that survive a back-channel reference call

Most resume-bullet articles are 3,000 words of formulas nobody uses and 300 examples that look identical. This is the version we'd actually keep open while writing. One formula, the verbs to retire in 2026, the plausibility test, and concrete examples for the roles people actually write resumes for.

The XYZ formula (where it comes from, and why it works)

Laszlo Bock, then SVP of People Operations at Google, published the formula in a 2014 LinkedIn post: every bullet should be accomplished [X], as measured by [Y], by doing [Z]. His own example:

• Weak: Studied financial performance of companies and made investment recommendations.
• Strong: Improved portfolio performance by 12 percent ($1.2M) over one year by refining cost of capital calculations.

It works because it forces three things in one sentence: an outcome (X), a measurement (Y), and a method (Z). Each of those is the thing recruiters and ATS systems actually weight. The weak version leaves the reader guessing on all three.

XYZ vs STAR vs PAR

STAR (Situation, Task, Action, Result) was introduced by Development Dimensions International in 1974, for verbal interview answers, not resume bullets. PAR (Problem, Action, Result) is a tighter cousin. Both are useful in interviews. For a resume bullet, XYZ is shorter and lands the metric first.

Simple rule: use XYZ for every bullet, then keep a STAR- shaped version of your three favorite bullets ready for the interview. They're the same story, two formats.

The first four characters of a bullet are the only ones guaranteed to be read

The Ladders 2018 eye-tracking study found recruiters scan a resume in 7.4 seconds on the first pass, with 80 percent of attention on six data points: name, current title, previous titles, dates, education, current company. Bullets get the leftover attention, and bullets are skimmed vertically. The only words guaranteed to be seen during that vertical skim are the first 2 to 3 words of each line. Long paragraphs get skipped entirely.

The operational rule: start every bullet with a verb, ideally a verb that signals ownership. Built, Shipped, Closed, Designed, Wrote, Coded, Trained, Sold, Negotiated, Reduced. Avoid verbs that suggest you helped someone else do the thing: Helped, Assisted, Supported, Participated, Worked with, Tasked with. Those words signal "was in the room" not "did the work."

First word matters more than any other word on the line. Make it a verb that names what you did, not what someone else needed you to do.

The anti-patterns: 5 phrases recruiters skim past

These openings make a recruiter's eye keep moving. Each has a reason it's weak, and each has a fix.

• "Responsible for"implies a duty assigned to you, not an outcome you produced. "Responsible for managing the support team" becomes "Built and led a 6-person support team that hit 96 percent CSAT across 2,500 monthly tickets."
• "Helped with" and "Assisted with"minimize your own role. If you genuinely contributed, "Contributed to" or "Co-led" carries more weight. If you owned it, name what you owned.
• "Duties included" is a 1990s job-description format. Replace with a concrete verb plus an outcome.
• "Experience in" is a skills- section construction misused as a bullet opener. Either show the experience as a real bullet or move it to the skills section.
• "Worked with"tells the reader you were on a team, not what you did on the team. "Worked with marketing to launch the campaign" becomes "Built the lead-scoring model marketing used to launch the September campaign, which sourced $1.2M in Q4 pipeline."

The action-verb death pool: 8 verbs to retire in 2026

These verbs used to read as strong. In 2026 they read as AI-generated, because language models have over-produced them since ChatGPT's launch. A 2025 Florida State University study confirmed the spike. A separate 14-million-abstract analysis (Kobak et al.) found delves appearing 25 times more often than the pre-2022 baseline, showcasing 9 times more, underscores 9 times more.

The retire-these list:

• Spearheaded: the single most-cited AI tell on a resume. Replace with Led, Built, Launched, Started, Owned.
• Orchestrated: co-occurs with Spearheaded in 100 percent of recruiter-flagged AI clusters. Replace with Coordinated, Ran, Rebuilt, Organized.
• Leveraged: as a verb, deprecated. Replace with Used, Applied, Built on, Drew on. (The noun is still fine.)
• Showcased: padding. Replace with Presented, Shipped, Demonstrated.
• Architected: only acceptable when literally describing software architecture. Otherwise replace with Designed or Built.
• Synergized: there is no replacement, because the word never meant anything. Delete.
• Drove: diluted. Still acceptable when paired with a number ("Drove $4.2M ARR"), but "Drove growth" or "Drove engagement" reads as filler.
• Delved: never use. The flagship AI vocabulary tell since 2024.

The verbs that still work in 2026 are the boring concrete ones, because they always meant something specific. Built, Shipped, Closed, Hired, Fired, Sold, Bought, Wrote, Coded, Trained, Negotiated, Reduced, Increased, Grew, Cut, Eliminated, Standardized, Automated, Migrated, Replaced, Consolidated.

The plausibility test (and why round numbers are now suspect)

Every quantified bullet has to pass three checks before it goes on a resume. We call this the plausibility test because the cost of failing isn't the screener (which won't catch you), it's the back-channel reference call. Top retained search firms now do off-list reference calls, and roughly 5 to 10 percent of offers are rescinded after the reference catches a resume claim that doesn't hold up.

1. The team-size test: would the team and role you held plausibly produce that number? "Increased revenue by 40 percent" from a junior analyst on a 50-person sales team will not survive the call. Either bound the claim ("Contributed analysis to a 40 percent revenue lift over Q3") or replace the number with a scope signal you genuinely owned.
2. The arithmetic test: does the percentage reconcile with company size, market context, and your timeline? "Grew revenue from $0 to $5M" on a 6-month internship is not arithmetically plausible. Use ranges and timeframes that match the real scope.
3. The back-channel test: could your old manager or peer corroborate this in 30 seconds if asked? If not, the bullet should either become a range, a non-numeric scale signal, or be cut.

Bullets when you can't use a number

Some work doesn't produce a measurable. Creative roles, classified work, government, anything where the numbers are proprietary or unreleased. The Muse's three frames cover almost every case.

• Range: "Supervised 7 to 12 research students each year, all of whom went on to graduate study in physics or astrophysics."
• Frequency: "Reviewed 40 to 50 articles per week; decided which advanced to editorial and which returned to authors with revision notes."
• Scale: "Chaired the promotional committee of 12, presenting to weekly senate meetings of 40 to 60 students."
• Before and after: "Standardized a previously ad-hoc onboarding process for new analysts, reducing time-to-first-independent-task from four months to six weeks."

Rough numbers are fine. Nobody verifies whether you saved 19 percent or 22 percent. The bullet only has to be approximately correct, and it has to survive the plausibility test above.

Bullets by role (10+ each)

Each line below either passes the XYZ formula or names a specific outcome with a credible scope. Pick the ones that match your work and adapt the numbers to yours.

Software engineer

• Cut p99 latency on the checkout API from 2.4s to 380ms by replacing a synchronous fanout with a batched query and Redis read-through cache.
• Migrated the frontend build from Webpack to Vite, dropping local dev server startup from 45 seconds to 3 across a team of 18 engineers.
• Wrote and shipped 60-plus Storybook components for the design system, cutting average UI build time on new features by 35 percent.
• Refactored a 12,000-line legacy module into 6 typed services; deploy errors fell from 8 per week to fewer than 1.
• Built the GraphQL federation layer that consolidated 4 microservices behind one schema; freed 2 engineers from inter-team coordination meetings.
• Took the on-call alert volume from 70 pages a month to 18 by adding SLO-driven alert routing and killing 14 flaky alerts.
• Designed and shipped a feature-flag service used by 9 product teams; replaced 3 ad-hoc rollback systems.

Product manager

• Shipped checkout v2 to 14M users in Q3 2025; cart abandonment dropped from 71 to 58 percent over the 90-day measurement window.
• Owned the pricing model rewrite for the SMB tier, lifting trial-to-paid conversion from 4.1 to 6.8 percent.
• Ran the discovery for the API marketplace launch (38 customer interviews, 4 prototype rounds); 87 percent enterprise adoption inside 90 days of GA.
• Co-led a cross-team migration off the legacy admin app; cut weekly support escalations by 42 percent.
• Wrote the PRD and shipped the v1 onboarding redesign; new-user activation climbed from 45 to 72 percent in 30 days.
• Built and ran the quarterly OKR review process across 4 squads; cycle time on a typical roadmap commitment dropped from 11 weeks to 7.

Marketing manager / SEO / content

• Grew organic traffic 114 percent to 500K monthly visits in 12 months by restructuring site architecture around 9 topical clusters.
• Built and ran the editorial calendar for a 4-writer content team; published 92 posts over 11 months, lifting non-brand search clicks 3.2x.
• Designed and ran a 14-cell A/B testing program across email and landing pages; lifted click-through rate on the main nurture from 6.2 to 9.1 percent.
• Reallocated a $1.2M paid media budget across 3 channels in 8 weeks; cost per acquisition fell from $58 to $42 while volume held.
• Recovered 78 percent of organic traffic lost in the March 2026 core update inside 60 days by implementing an EEAT-driven content overhaul.
• Built backlinks to 8 priority pages through 23 published guest posts; average target-keyword position moved from 12 to 4.

Data analyst / data scientist

• Cleaned and standardized a 4.7M-row customer table in pandas, resolving 38,000 duplicate records and 12 inconsistent date formats; downstream model accuracy moved from 76 to 84 percent.
• Designed and analyzed 15 A/B tests for the product team, with sample-size calculations and post-hoc statistical significance checks; 9 of 15 shipped to GA.
• Built a forecast model in Prophet for SMB demand; cut MAPE from 18 to 9 percent over 8 months.
• Cut weekly executive-reporting prep from 12 hours to 90 minutes by building 7 dbt models that replaced manual spreadsheet pulls.
• Authored a pricing-optimization analysis adopted by sales leadership; the recommended changes generated $1.4M in net-new ARR over the following year.
• Analyzed 3.2M payment transactions for fraud patterns; surfaced 14 high-confidence rules that the rules engine then automated.

Sales (AE / SDR / CS)

• Closed $4.2M in new ARR across 48 deals in 2025 at 135 percent of annual quota; average deal size $87K, average cycle 35 days.
• Booked 45 qualified opportunities per quarter as an SDR, averaging 140 percent of monthly meeting quota across 6 consecutive months.
• Grew average deal size from $16K to $22K over 18 months by writing and running a structured expansion-discovery playbook.
• Held an 89 percent gross renewal rate across 75 SMB customers; identified $380K in upsell opportunities through structured business reviews.
• Cut net churn from 12 percent to 7.9 percent annually by building a Gainsight health-score that flagged at-risk accounts 90 days before renewal.
• President's Club 2024 and 2025.

Designer (UX / product)

• Ran 24 user interviews and 4 rounds of usability tests on the onboarding flow; activation climbed from 45 to 72 percent over the 30-day window.
• Redesigned the provider search experience; successful appointment bookings rose 35 percent over the next quarter.
• Shipped the v3 component library adopted across 6 product teams; UI build time for new features fell 28 percent on average.
• Lifted the System Usability Scale score on the dashboard from 62 to 81 across 3 iteration rounds and 120 testing sessions.
• Owned the accessibility audit for the marketing site; closed 47 WCAG 2.2 AA issues across 9 templates ahead of a contractual review.

Finance / FP&A / accounting

• Built monthly financial reports covering $45M in revenue across 3 business units; variance analysis surfaced $1.2M in addressable cost reduction.
• Owned the rolling 12-month forecast for a $200M operating budget; hit 92 percent accuracy against actuals for 4 consecutive quarters.
• Reduced the monthly close timeline from 12 days to 7 by restructuring the inter-company reconciliation workflow in NetSuite.
• Negotiated multi-year contracts with 4 key suppliers; price-lock provisions saved $2M over the contract period.
• Cut forecast variance from 10 percent to 4 percent across 4 quarters by introducing a driver-based model in Adaptive Planning.

Customer support

• Held 96 percent CSAT across 2,500 monthly interactions on Intercom, email, and Zoom; 4-hour median first response time.
• Resolved 50-plus tickets per day with 98 percent CSAT over 6 consecutive quarters.
• Cut average handle time from 8 minutes to 5.5 minutes by writing 14 new macros and retiring 8 outdated ones.
• Raised team NPS from 42 to 61 over 6 months by introducing structured follow-up sequences after escalations.
• Wrote 32 help-center articles cited in 23 percent of inbound tickets the following quarter; deflected an estimated 480 tickets per month.

How many bullets per role (and how long)

Two related questions with consistent answers across recruiter surveys.

• Bullet count per role: 3 to 5 for entry-level, 5 to 7 for mid-level, 6 to 8 for your most recent senior role. Older roles shrink to 2 to 3 bullets each. Anything over 8 starts to read as padding.
• Bullet length: 15 to 30 words per bullet, one or two lines. Two lines is fine if it stays one coherent thought. Three lines and beyond get skipped during the initial skim.
• Total resume bullets: 18 to 28 for a one-pager, 30 to 45 for a two-page. More than 50 is almost certainly the wrong density.

When you didn't own the metric (the "we vs I" allocation)

On most teams, the big numbers are team numbers. Pretending you single-handedly drove a 40 percent revenue lift on an 8-person sales team is a back-channel call away from being caught. Three honest patterns for IC and junior bullets.

• Contributed to(when you were one of many): "Contributed to the 20 percent sales lift in EMEA by building the data pipeline that powered the weekly territory model."
• Owned X within Y(when you owned a sub-step of a team result): "Built the customer health-score query inside the CS team's retention dashboard; used weekly by 4 CS leads."
• 1 of N on Z(when scope is the honest unit): "1 of 4 engineers shipping the v3 platform rebuild, which released to 200K users with 99.95 percent rollout uptake."

Roughly 30 to 40 percent of bullets on most CVs should have collaborative language. The rest can be solo verbs. A page of nothing but "I built, I owned, I shipped" reads as exaggerated unless you're a solo founder.

The parody bullet: why this exact sentence fails

Pull any AI-generated draft and you will see this archetype somewhere on the page:

Spearheaded a cross-functional team to deliver results that drove measurable impact.

Every component fails a separate test:

• Spearheaded: AI vocabulary cluster (see the death pool above).
• A cross-functional team: no team size, no named functions, generic.
• To deliver results: "results" is a filler noun. What results?
• That drove: a second AI verb stacked on the first.
• Measurable impact: empty without the measurement. The whole point of the bullet is the measurement.

The fix, any of the following work:

• Led a 6-person eng-PM-design pod to ship checkout v2; cart abandonment dropped from 71 to 58 percent in 90 days.
• Coordinated a 4-person team across engineering, PM, and legal to launch the GDPR-compliant cookie banner in 6 EU markets in 8 weeks.
• Co-led the dashboard rebuild with PM plus 2 engineers; cut weekly leadership report-prep from 6 hours to 45 minutes.

Related reading

For the verb-by-verb rule on what now reads as AI on a resume, see our AI detection guide. For getting bullets through the resume scanner before the human reads them, the ATS resume guide is the parser-by-parser detail. The keywords-by-industry post has the role-specific phrasing your bullets should echo.

Key takeaways

• The XYZ formula (Laszlo Bock, 2014): accomplished [X] as measured by [Y] by doing [Z]. Use it on every bullet. STAR is for the interview.
• First word matters most. Lead with a concrete ownership verb (Built, Shipped, Closed, Reduced). Retire the AI cluster (Spearheaded, Orchestrated, Leveraged, Showcased).
• Every metric must pass the plausibility test: team size, arithmetic, and back-channel verifiable. Round numbers (50 percent, 30 percent) now read as fabricated.
• If you can't use a number, use a range, a frequency, a scale signal, or a before-after state. Rough numbers are fine, fake precise numbers are not.
• 3 to 5 bullets entry-level, 5 to 7 mid-level, 6 to 8 most recent senior. Older roles shrink to 2 to 3. Each bullet 15 to 30 words, one to two lines.
• When you didn't own the metric, use "Contributed to," "Owned X within Y," or "1 of N on Z." Honesty about scope beats exaggeration on every reference check.
• Avoid the parody bullet: "Spearheaded a cross-functional team to deliver results that drove measurable impact." Every word fails.

If you want the formulas applied to your CV in one pass, the free CV score flags weak bullets in 90 seconds, and Glow Up rewrites them using the same XYZ logic above, with the AI cluster retired and the plausibility test built in.