Front page / By role / Data and analysis / Data Analyst

Recommended by role

Data Analyst

Not "write my SQL" but "know my definitions and show your work": skills for the intake, the metric, the query, the check, the readout and the memo, built from how analysts describe their week with agents.

More specific roles under it: Analytics engineer, BI developer, Product analyst, Marketing analyst

Install what you need, one listing at a time: every part of the work below links to the tools for it, and each listing has its own Add button. Installing a whole kit at once is not available yet.

Role

Product manager Technical product manager Product owner Product operations Program manager Project manager Scrum master or agile coach Software engineer Frontend engineer Backend engineer Full stack engineer Mobile engineer DevOps or site reliability engineer Platform engineer QA or test engineer Security engineer Machine learning engineer Data engineer Engineering manager Solutions architect Game developer Embedded or hardware engineer Data analyst Business analyst Data scientist Analytics engineer BI developer Product analyst Marketing analyst Product designer UX researcher UI or visual designer Graphic designer Content designer or UX writer Brand designer Marketing manager Growth marketer Product marketing manager Content marketer SEO specialist Performance marketer Social media manager Community manager Copywriter Account executive Sales development representative Sales engineer Account manager Customer success manager Customer support agent Partnerships and business development Operations manager Business operations or strategy Revenue operations Supply chain and logistics Procurement Executive assistant Financial analyst Accountant Finance manager or controller Investment analyst Risk analyst Recruiter HR business partner People operations Learning and development Lawyer or in-house counsel Paralegal Compliance officer Privacy officer Founder Executive or general manager CTO or VP of engineering Chief of staff Management consultant Writer or journalist Editor Technical writer Video producer or creator Academic researcher Teacher or educator Student Clinician Healthcare administrator

Industry

All industries SaaS and B2B software Consumer apps E-commerce and retail Fintech and payments Healthcare Marketplaces and platforms Games and live ops

Workflows for this work

The job in 5 areas. Each part is a workflow: open it for its steps and the tools that do each one. Built from occupational data and the field's own body of knowledge; sources at the end.

Data analyst209 skill packs · 65 MCP servers · 6 pluginsEach part is a workflow. The number beside it is how many tools help with it

1Ask the right question

Turn a vague request into a clear question with agreed definitions.

2Get and clean the data

Find the right data, pull it and make sure it can be trusted.

3Analyze the data

Explore the data, run the right tests and check the work before it goes out.

4Share what you found

Put results in front of people as dashboards, charts and short write ups.

5Help people act

Turn findings into decisions and leave the work reusable and safe.

Sources for this tree, 8

StandardThe AQuA Book: guidance on producing quality analysis for governmentUK Government Analysis Function (Government Operational Research Service, Government Actuary's Department, Office for National Statistics)
Body of knowledgeIBM SPSS Modeler CRISP-DM Guide (version 18.3): the six CRISP-DM phases and their tasksIBM
VendorAbout MetricFlow (define metrics once in a central place)dbt Labs
Occupational dataData analyst: role description and skillsGovernment Digital Service (UK), Government Digital and Data Profession Capability Framework
Occupational dataPerformance analyst: role description and skillsGovernment Digital Service (UK), Government Digital and Data Profession Capability Framework
Body of knowledgeGoogle Data Analytics Certificate: courses Ask, Prepare, Process, Analyze, ShareGoogle (Grow with Google)
Occupational dataData Scientists (15-2051.00), summary report: tasks and detailed work activitiesO*NET OnLine, U.S. Department of Labor
Occupational dataBusiness Intelligence Analysts (15-2051.01), summary report: tasks and detailed work activitiesO*NET OnLine, U.S. Department of Labor

O*NET occupations consulted: 15-2051.01, 15-2051.00. Read on 2026-10-04.

Stage by stage walkthrough, 10 stages, with what practitioners keep for themselves

Intake: from a vague ask to a briefshared with BA, PM

What you do

Turn the request into a decision question
Pin the metric, grain, filters and time window
Name the definition source before any SQL

DeliverableOne-paragraph analysis brief the stakeholder signs off

Stays with youThe ambiguity itself. "Active customer revenue last month" hides three undefined terms; resolving them is the job.

“The system can write perfect SQL and still miss the business meaning entirely.”Jose Parreño Garcia, senior data science lead ↗

Skills that do it

Metric definitions and the semantic layershared with PM, BA

What you do

Draft metric definitions people can agree on, with lineage
Reconcile two numbers that disagree
Keep the semantic layer the single source for BI and agents

DeliverableMetrics dictionary, semantic model (dbt or YAML), reconciliation notes

Stays with youAgreeing the definition with the business. Anything outside the semantic layer is unsupported by design.

“Skills protect against plausible but incorrect output.”dbt Labs (vendor) ↗

Industry notes for SaaS and B2B software, E-commerce and retail, Fintech and payments, Marketplaces and platforms, Healthcare, Games and live ops. Pick an industry above to read them.

Read industry notes

SaaS and B2B softwareAccount grain, not user grain: activation, trial-to-paid, feature adoption, net dollar retention as expansion plus reactivation minus contraction minus churn, and renewal-period retention.

E-commerce and retailGMV before returns and fees, AOV, conversion, and revenue per visitor as the reconciling metric; cohort repeat purchase and AOV by first-order month.

Fintech and paymentsApproval rate, fraud loss rate, chargeback rate and false-positive rate need the exact denominators regulators and bank partners use.

Marketplaces and platformsNet revenue is GMV times take rate; GMV is active buyers times transactions per buyer times average transaction value; track liquidity and search-to-fill.

HealthcareReadmissions, length of stay, throughput and denial rates, computed on de-identified extracts with small-cell suppression.

Games and live opsD1, D7, D30 retention, ARPDAU by genre, and LTV to CPI; benchmark against public genre medians before calling a number good.

Skills that do it

Acquire, clean and check the data

What you do

Pull from warehouse, API, CSV or Excel
Run the quality audit: nulls, duplicates, ranges, row counts
Log every cleaning step and every column touched

DeliverableClean table with a cleaning log and a row-count and column diff

Stays with youWatching for silent drops. Agents have quietly lost columns and fetched a third of the rows because of pagination; the audit catches what the chat did not mention.

“Claude Code isn't going to replace data engineers (yet).”Robin Moffatt, after a dbt run silently dropped columns ↗

Industry notes for Healthcare, Fintech and payments. Pick an industry above to read them.

Read industry notes

HealthcareWork on de-identified or limited datasets under a data use agreement; log lineage from source to report.

Fintech and paymentsLeast-privilege access and PAN redaction before any extract leaves the warehouse.

Skills that do it

SQL writing and reviewshared with PM

What you do

Draft the query from the brief, with the SQL shown
Review joins for fan-out, grain, nulls and date boundaries
Explain the query in English and reconcile row counts

DeliverableReviewed query saved as a reusable view, with an English explanation

Stays with youThe review. Text-to-SQL fails silently and confidently; most wrong queries are schema-level mistakes.

“I know the relevant libraries well enough to read and verify the code, which is an important distinction from blindly using a black box AI.”Data analyst at a geology firm, Hacker News ↗

“The real danger is that text-to-SQL fails silently and confidently.”Omni (vendor) ↗

Industry notes for Fintech and payments, Healthcare. Pick an industry above to read them.

Read industry notes

Fintech and paymentsAdd a PII and card-number check to every review; a query that returns raw PANs fails review whatever its logic.

HealthcareEvery query carries lineage and a suppression rule for small cells before it reaches a report.

Skills that do it

Exploratory analysis and root causeshared with PM

What you do

Run the EDA and the quarter-over-quarter comparison
Decompose a miss into drivers with a waterfall
Split claims the data supports from hypotheses

DeliverableFindings doc with charts and an explicit supported-versus-hypothesis split

Stays with youSign-off on every claim. Agents have confidently attributed a gap to sales effectiveness with nothing in the data to support it.

“Every number, every claim, every conclusion still needs a human sign-off.”Data analyst, after matching three days of work in an hour ↗

Industry notes for E-commerce and retail, SaaS and B2B software. Pick an industry above to read them.

Read industry notes

E-commerce and retailDecompose revenue into traffic, conversion, order value and returns; separate discount-driven volume from margin.

SaaS and B2B softwareBridge net revenue retention into expansion, contraction and churn before blaming acquisition.

Skills that do it

Experiments and statisticsshared with PM

What you do

Size the test from baseline, effect and traffic
Check sample ratio mismatch and segment effects
Compute effects with intervals; use CUPED or sequential tests where they apply
Write the go, no-go or partial decision

DeliverableExperiment readout with primary and guardrail metrics and a decision

Stays with youChoosing the primary metric and the exposure unit, and refusing to peek. The skills run the arithmetic and the checks.

Industry notes for E-commerce and retail, Consumer apps, Games and live ops. Pick an industry above to read them.

Read industry notes

E-commerce and retailPrimary metric is revenue per visitor with conversion and order value as the decomposition; guardrail on margin and returns.

Consumer appsHigh experiment velocity means multiple-comparison discipline; keep the exposure unit consistent across onboarding, push and paywall tests.

Games and live opsRead live-ops offers on ARPDAU delta and retention delta, then D30 impact, not on first-day revenue.

Skills that do it

Cohorts, funnels and forecastsshared with PM

What you do

Build cohort and retention views
Map and measure the funnel
Forecast with a stated method and a stated error

DeliverableCohort tables, funnel report, forecast with intervals

Stays with youPicking the cohort key and the funnel definition; both encode a business decision.

Industry notes for E-commerce and retail, Marketplaces and platforms, SaaS and B2B software, Games and live ops. Pick an industry above to read them.

Read industry notes

E-commerce and retailCohort by first-order month on repeat purchase rate and AOV, weekly for consumables and subscriptions; overlay the promo calendar.

Marketplaces and platformsCohort supply and demand separately and review both in the weekly marketplace review.

SaaS and B2B softwareCohort by signup month and plan tier at account grain; build the NRR bridge from the same table.

Games and live opsFit retention curves per acquisition cohort; LTV is the area under the curve times ARPDAU.

Skills that do it

Dashboards and recurring reportsshared with BA

What you do

Write the dashboard spec: metric, grain, filters, owner, cadence
Build it in the BI tool or as an HTML page
Produce the weekly report pack

DeliverableDashboard spec, the dashboard, the weekly pack

Stays with youChart choice and layout judgment; publication figures still need hand tweaks.

Industry notes for E-commerce and retail, Games and live ops. Pick an industry above to read them.

Read industry notes

E-commerce and retailDaily tracking through holiday periods with a promo-calendar overlay; margin next to GMV, always.

Games and live opsLive-ops dashboards read event participation, ARPDAU and retention deltas per event.

Skills that do it

Narrative and stakeholder handoffshared with PM, BA

What you do

Write the insight memo: findings, confidence, caveats, next steps
Produce the three-bullet summary for leadership
Translate method into business language

DeliverableInsight memo, executive summary, the answer to the follow-up question

Stays with youThe recommendation. Agents write a competent professional's memo; what to do about it is yours.

“Claude, on the other hand, sounds like a competent professional.”Jess Ramos ↗

Skills that do it

Documentation and QA

What you do

Document the analysis, its assumptions and its data
Run the QA checklist and a peer review
Keep the catalog entry and the dbt docs current

DeliverableAnalysis doc, assumptions log, catalog entry, reviewed notebook

Stays with youDeciding what is reusable. Reproducible scripts beat chat state; sessions do not remember.

“With learning to use these tools comes learning WHAT to hand over to them and what to keep tight reins on yourself.”Analytics engineer ↗

Skills that do it

Packages for this role

Ranked like the ladder: by the GitHub stars each source repository gained this week. Install any of them into any supported client.

Recommendations are curated in catalog/professions.json (updated 2026-09-28) and reviewed like code. Quotes link to the practitioner who said them, and vendors are marked as such. Nothing here is ranked by payment.