The 40 Action Verbs ATS Scores Highest (With Examples)

The Myth About Action Verbs and ATS

A common claim online is that ATS systems reward specific action verbs and downgrade others. This is mostly false. The major ATS systems — Workday, Greenhouse, Lever — score on keyword match, skills coverage, and years of experience. They don't maintain an action-verb preference list.

So why does verb choice matter? Because after the ATS passes your resume through to a human, the recruiter spends 10 to 20 seconds on it. The verbs at the start of your bullet points are the first thing their eye lands on. Weak verbs ("worked on," "assisted with," "helped with") make strong experience read as passive. Specific verbs ("rebuilt," "shipped," "cut") signal ownership and outcome.

The Verb Test

A good action verb answers one question: what did you actually do? Before using any verb, test it by asking whether a specific person could have done the opposite. If yes, the verb has meaning. If not, it's filler.

"Led" passes the test — someone else could have followed, delegated, or stayed uninvolved. "Responsible for" fails — it describes your position, not your action.

Engineering: 10 Verbs That Signal Ownership

• Built — Built real-time pricing service in Go handling 12k RPS
• Shipped — Shipped v2 of checkout flow across web and iOS in 11 weeks
• Rebuilt — Rebuilt legacy payment module, cut incident rate 80%
• Migrated — Migrated 2.3TB from MySQL to Postgres with zero downtime
• Designed — Designed event-sourcing architecture adopted by 4 adjacent teams
• Refactored — Refactored auth service, reduced p95 latency from 420ms to 90ms
• Optimized — Optimized Postgres queries on order table, cut DB cost 38%
• Automated — Automated CI/CD for 17 microservices using GitHub Actions
• Scaled — Scaled notification service from 50k to 4M daily sends
• Debugged — Debugged memory leak in production gateway, recovered 42% capacity

Product and Program: 10 Verbs That Signal Direction

• Led — Led onboarding redesign across 3 engineering squads
• Launched — Launched self-serve tier in 5 markets, 22% of new signups within 90 days
• Prioritized — Prioritized 180-ticket backlog into quarterly roadmap with 4 cross-functional leads
• Defined — Defined activation metric now used by product, marketing, and finance
• Drove — Drove activation from 24% to 41% through onboarding experiments
• Shipped — Shipped billing v3 with usage-based pricing and grace-period handling
• Aligned — Aligned pricing, legal, and engineering on enterprise tier in 6 weeks
• Scoped — Scoped and wrote PRDs for 14 shipped features over two years
• Reduced — Reduced time-to-first-value for new users from 11 days to 4 days
• Negotiated — Negotiated vendor contract change that cut cost 28% with zero feature regression

Data and Analytics: 10 Verbs That Signal Rigor

• Modeled — Modeled customer LTV by cohort for pricing committee decision
• Analyzed — Analyzed 11M orders to identify fraud pattern missed by rule-based system
• Forecast — Forecast revenue by product line with 4% MAPE over 6 quarters
• Instrumented — Instrumented product event schema adopted by 7 downstream teams
• Experimented — Experimented on 14 onboarding variants, 3 shipped to 100%
• Built — Built dbt model for order lifecycle consumed by 40+ Looker dashboards
• Surfaced — Surfaced pricing anomaly costing $1.2M/year in manual refunds
• Quantified — Quantified impact of paywall redesign: +18% conversion, -3% refund rate
• Automated — Automated weekly finance close from 9 hours to 40 minutes
• Investigated — Investigated 30% step-drop in funnel, traced to mobile asset regression

Operations, Marketing, and Customer-Facing: 10 Cross-Role Verbs

• Onboarded — Onboarded 240 new enterprise accounts over 18 months
• Retained — Retained 94% of mid-market renewals in first full year
• Closed — Closed 38 enterprise deals at $1.4M total ARR
• Negotiated — Negotiated 11-contract renewal cycle; raised ACV 22% on average
• Recovered — Recovered $420k from at-risk accounts through structured QBR program
• Streamlined — Streamlined vendor onboarding from 14 days to 4 days
• Resolved — Resolved 1,800+ support tickets with 96% CSAT across 2 years
• Wrote — Wrote help center documentation for 80+ features used by 120k MAU
• Trained — Trained 14 new CSMs on onboarding playbook over 3 cohorts
• Reported — Reported weekly NRR/CSAT dashboard to leadership for 2 years

Verbs to Avoid — and Why

Some verbs feel substantive but carry very little information. They flood resumes, dilute stronger bullets, and create the impression of a generic profile. Replace them wherever possible.

• "Worked on" — describes presence, not action. Replace with Built, Shipped, or Contributed to.
• "Assisted with" — signals peripheral involvement. If the work was peripheral, it likely doesn't belong on the resume at all.
• "Responsible for" — describes the job description, not what you did. Replace with the specific action you took.
• "Helped" — vague. Replace with Supported, Collaborated, or the actual action (Wrote, Tested, Debugged).
• "Involved in" — the weakest verb on most resumes. Remove the bullet or rewrite it around a concrete action.

The Bullet-Point Formula That Works

Every bullet should pass three tests: it starts with a specific verb, it names the system or problem, and it ends with a measurable outcome or scale marker. If you can't fill all three, the bullet isn't ready yet.

Template: [Action verb] + [system / problem] + [measurable outcome or scale].

Example: "Rebuilt auth service — cut p95 latency from 420ms to 90ms across 180k daily active users."

Rewriting a Weak Bullet Into a Strong One

Seeing the transformation end-to-end is more useful than any verb list in isolation. Here are three real-world examples of rewrites, pulled from common resume patterns.

Example 1 (engineering):

Weak: "Worked on improving the checkout experience and fixing various bugs on the payments team."

Strong: "Rebuilt checkout state machine in TypeScript; cut dropped-cart rate from 11% to 4% and eliminated 3 recurring payment-timeout bugs."

Same underlying work. The strong version names the specific change, quantifies the outcome, and uses a verb that signals ownership. The weak version reads like someone who was on the team while others did the work.

Example 2 (product):

Weak: "Helped drive activation by running experiments and working with design."

Strong: "Ran 14 onboarding experiments with design and engineering; shipped 3 winners that moved day-7 activation from 24% to 41% across 180k monthly signups."

Example 3 (data):

Weak: "Assisted the team with analysis and reporting for the finance department."

Strong: "Built dbt models feeding weekly finance close dashboards; cut close cycle from 9 hours of manual work to 40 minutes of automated runs."

How Many Bullets Per Role, and How Long?

More bullets is not better. A senior engineer with eight years of experience who lists eleven bullets under their most recent role dilutes the strong bullets with weak ones. A recruiter reading the first three will decide whether to keep going; if the first three don't land, bullets four through eleven don't save you.

Reasonable guidance for 2026 resumes:

• Most recent role: 4 to 6 bullets. This is where the bulk of your evidence lives.
• Role 2 (one back): 3 to 5 bullets.
• Role 3 and beyond: 2 to 3 bullets each. Older roles get compressed unless they contain something unusually relevant.
• Bullet length: One line, maybe two. Three-line bullets with semicolons and qualifiers read as noise.

The total resume length follows from this: one page for most people with fewer than 10 years of experience, two pages for people with more or for academic and engineering leadership roles where publications and open-source contributions need their own sections.

Test Your Rewritten Bullets

Verbs and outcomes handle the human-review half. The ATS half still requires keyword match. After rewriting bullets with stronger verbs, paste the updated resume and the target JD into ATSGuard. It surfaces any keywords still missing from the new phrasing so you can fix both halves — ATS pass and human pass — in one sitting.