How to write a data analyst resume
Contents:
Why your resume gets 30 seconds, not 30 minutes
A recruiter at a mid-size tech company opens 80 to 150 resumes for a single data analyst opening. They are not reading — they are scanning for disqualifiers. The eye tracks down the left margin, snags on titles and company names, then jumps to the first bullet of the most recent role. If that bullet says "performed data analyst duties" you are out of the funnel.
The bar is not high, but it is specific. Hiring managers at Stripe, Airbnb, DoorDash, and the rest of the well-paying tier look for three signals in the first five seconds: a relevant title, a tool stack that overlaps with the JD, and at least one number in the first bullet. Miss any and the resume goes to "maybe" — which means "no", because the "yes" pile already has fifteen candidates.
The rest of this guide is about engineering those three signals on purpose.
Load-bearing rule: every bullet in your most recent role must contain a metric, a stack, or a stakeholder. Ideally all three. Bullets that contain none of those are decoration, and decoration gets cut.
The structure that works
The boring single-column resume outperforms the Canva multi-column template in almost every ATS we have seen parsed. Pick one of the standard templates — Google Docs default, the LinkedIn export, or a plain LaTeX class — and stop optimizing the visual layout. The optimization happens in the words.
Header
Name, the title you want (not the title you have, if they differ), city plus remote-readiness, and three contact channels. Keep it to four lines maximum.
Jane Doe
Senior Data Analyst / Product Analyst
San Francisco, CA — open to remote (US)
email | linkedin.com/in/jane | github.com/janeIf you are targeting roles outside your current country, write "open to relocation, sponsorship not required" if it applies. Recruiters filter on that string.
Summary (optional, 2 to 3 lines)
A summary block earns its space only if it sharpens your positioning. Vague summaries that could fit any analyst are worse than no summary at all.
Senior product analyst with 4 years in B2C marketplaces. Owned the experimentation
platform at a 12M MAU app — shipped 60+ A/B tests, reduced false-positive rate from
14% to 3% via CUPED, mentored 3 junior analysts.Experience
This is where 80% of the screen-pass decision happens. Format every role identically:
Company Name City / Remote
Job Title MM/YYYY — MM/YYYY
— One-line company context (only if niche; skip for FAANG/known brands)
— Bullet 1: highest-impact result with a metric and a dollar figure if possible
— Bullet 2: a technical achievement (the hard thing you built)
— Bullet 3: a cross-functional achievement (the thing you shipped with PM/Eng)
— Bullet 4: optional, a scope marker (team size, data volume, geography)
— Stack: Postgres, Snowflake, dbt, Looker, Python (pandas, statsmodels)Four bullets per role is the sweet spot. Three is fine for older roles. Six or more and your most important bullet gets diluted.
Education, certifications, projects
Education is one line for senior candidates and two lines for juniors. Certifications go below education and only if they are recognized — Google Data Analytics Certificate, AWS Certified Data Analytics, Databricks Certified Data Engineer Associate. Random Udemy completions do not belong here.
Projects matter most for candidates with under two years of experience. Two or three solid pet projects with a deployed demo or a public GitHub repo beat a dozen course completions. List the stack, the data, the problem, and the result — same shape as a job bullet.
STAR bullets that survive a skim
STAR — Situation, Task, Action, Result — is a framework for spoken interview answers. For resume bullets, compress it to a single sentence with a Result-first shape:
Pattern: [Result with metric] by [Action with stack] for [Stakeholder or business context]
Compare the same bullet in three voices:
| Voice | Bullet |
|---|---|
| Vague | Ran A/B tests on checkout |
| Action-first | Designed and analyzed 12 A/B tests on the checkout funnel using Python and the company's experimentation platform |
| Result-first | Lifted checkout conversion from 3.5% to 4.2% (+20% relative, +$3.1M annualized) by designing 12 A/B tests on the cart-to-confirmation funnel with the growth PM |
The result-first bullet front-loads the number that the recruiter is hunting for. The vague bullet gets you cut. The action-first bullet is better than vague but still demands the reader do work to find the impact — and they will not.
Resume bullets are not the place to be modest. If your team shipped the win, claim your slice with first-person verbs.
Required hard skills by level
The expectations have shifted in the last two years. AI assistants have collapsed the bar for syntax recall but raised the bar for judgment, scoping, and validation. Hiring managers no longer test whether you can write a window function from scratch — they test whether you know when one is the wrong tool.
| Level | SQL | Python / stats | Experimentation | BI / pipelines |
|---|---|---|---|---|
| Junior (0-2y) | JOIN, GROUP BY, basic CASE WHEN, simple subqueries |
pandas basics, mean/median/variance, a t.test you can defend |
Read an experiment readout, explain p-value in plain English | Build a dashboard in Looker / Metabase / Tableau |
| Mid (2-5y) | Window functions, CTEs, EXPLAIN ANALYZE, query optimization |
scipy / statsmodels, regression, bootstrap, basic causal | Design a test, pick the right metric, debug an SRM | Own a dbt project, schedule with Airflow or Dagster |
| Senior (5y+) | Everything mid + cost-aware queries on Snowflake/BigQuery, modeling | Productionized notebooks, CUPED, variance reduction, light ML | Run the experimentation review, coach PMs on power | Own a data platform piece end-to-end, mentor 2-3 analysts |
The biggest mistake at the senior level is listing the same skills as the mid bullet, just bolder. Senior signal is about ownership, ambiguity, and stakeholders, not about knowing one more SQL function.
ATS keywords that actually get parsed
Most ATS in 2026 do basic keyword matching with some stemming — Workday, Greenhouse, Lever, Ashby. They do not do semantic search well, so the literal string matters. If the job description says "experimentation platform", do not write "A/B testing framework" instead. Write both.
Sanity check: open the job description, paste it into a word-frequency tool, and confirm that the top 15 nouns and verbs each appear at least once in your resume body. If they do not, you are losing the keyword match before a human reads anything.
The terms that consistently parse well across ATS for analyst roles:
- Tools:
SQL,Python,pandas,NumPy,SciPy,statsmodels,dbt,Airflow,Snowflake,BigQuery,Looker,Tableau,Metabase,Mode,Hex,Git,R - Methods:
A/B testing,experimentation,causal inference,cohort analysis,funnel analysis,regression,time series,forecasting,CUPED - Domains:
product analytics,marketing analytics,growth,retention,LTV,CAC,churn,activation,North Star metric - Stakeholder verbs:
partnered with,owned,shipped,mentored,presented to,defined
Do not keyword-stuff. If the only mention of Spark is a single line in your skills section and you cannot explain repartition in a phone screen, take it off.
Common pitfalls
The most common failure mode is bullets without numbers. "Improved processes" tells the recruiter nothing. "Reduced p50 dashboard load time from 14s to 1.8s by rewriting the underlying query and adding a materialized view" tells them three things — you can profile, you can rewrite, and you understand percentiles. Scan every bullet and ask "what number would have made this concrete?" If you do not know, get it from your BI tool before you submit. Made-up numbers are worse than no numbers — senior hiring managers smell them in the phone screen.
A close second is the technology grab-bag. A skills section listing twenty tools tells the reader you have done a tutorial in each and committed to none. Hiring managers prefer six tools you can defend over twenty you cannot. Delete any tool you would be uncomfortable being grilled on for ten minutes in a technical screen.
The third pitfall is the one-resume-fits-all submission. You do not need a new resume per role, but you do need a new top — title line, summary, and the order of bullets in your most recent job. If the JD emphasizes experimentation, lead with an experimentation result. If it emphasizes modeling, lead with dbt or modeling. This takes ten minutes per application and roughly doubles the screen-pass rate in cohorts we have tracked.
The fourth is misjudged length. Juniors pad to two pages with coursework. Seniors compress eight years into one page out of misplaced modesty. The right shape is one page for under five years and two pages above that. A two-page resume with one strong page and one weak page reads worse than a tight one-pager.
The fifth is the headshot trap. In the US, UK, Canada, Australia, and most of the EU, photos on resumes invite bias concerns and look unprofessional. Drop the photo unless you are applying in a region where photos remain the local convention.
Before-and-after rewrites
The fastest way to internalize the pattern is to rewrite three of your own bullets right now. Here are the templates:
Bullet A — reporting work
| Version | Bullet |
|---|---|
| Before | Built dashboards for management |
| After | Automated 15 weekly executive reports via Airflow + dbt, saving the team 20 hours/month and delivering data 2 business days earlier |
Bullet B — SQL work
| Version | Bullet |
|---|---|
| Before | Worked with SQL |
| After | Authored 200+ production SQL queries on Postgres and ClickHouse over tables of 10B+ rows, with cost-aware patterns audited via EXPLAIN ANALYZE |
Bullet C — experimentation work
| Version | Bullet |
|---|---|
| Before | Helped with A/B testing |
| After | Owned the analysis layer of 40+ A/B tests across onboarding and pricing, cut false-positive rate from 12% to 3% via CUPED, and trained 4 PMs on test design |
The pattern is consistent: a verb you own, a number that quantifies the work, and either a dollar figure, a percentage, or a stakeholder count to anchor scale.
Cover letters and the screen call
A cover letter is rarely required in 2026, but the three-paragraph version is worth keeping in a snippet manager. Paragraph one: one sentence on why this company. Paragraph two: one concrete past result that maps to the JD. Paragraph three: soft close with availability. Six sentences total.
Your resume also scripts the first ten minutes of your screen call. "Walk me through your resume" means reading your bullets back, expanded into STAR stories. If a bullet is too vague to expand into a 90-second story, it is too vague to be on the page.
If you want to drill the SQL and experimentation questions that come up after the screen, NAILDD launches with 500+ analyst interview problems across these patterns.
Related reading
- How to become a data analyst from scratch
- How to become a senior data analyst
- How to land a FAANG data analyst role
- SQL on data analyst interview
- Resume without analytics experience
FAQ
One page or two pages?
Under five years of experience, aim for one tight page — the recruiter scan is calibrated for that length and a second page almost always means the first page got padded. Five to ten years, two pages is fine but the second page should be load-bearing (projects, publications, open-source), not more bullets of the same shape. The rule of thumb: if you cannot point to why page two exists in one sentence, it should not exist.
Are pet projects required?
For candidates with under one year of professional analyst experience, yes — two or three substantial projects with deployed demos or public repos substitute for the work history you do not have yet. The best pet projects look like real analyst work: a question, a data source, a SQL or pandas analysis, a visualization, and a short writeup. For mid and senior candidates they are optional, though a current technical blog or popular GitHub repo can still earn its place.
Should I include my GPA or graduation year?
If you graduated in the last three years and your GPA is at or above 3.5 on a 4.0 scale, include it; otherwise drop it. Graduation year is recommended within five years of graduation, after which the year of your most recent degree becomes a proxy for age. Dropping the date of your bachelor's degree is one of the highest-leverage age-bias edits you can make.
How do I show impact when confidential?
Use relative numbers and ranges instead of absolute figures. "Reduced report generation time by 80%" is just as strong as a raw second count and reveals nothing proprietary. For revenue impact, "low seven figures annualized" or "single-digit percentage of segment revenue" gives the reader scale without exposing the dollar amount.
How often should I update my resume?
Quarterly, even when you are not job-hunting. Bullets written within a week of shipping are far more specific than bullets reconstructed from memory six months later — that specificity separates a strong resume from a generic one.