How to become an analyst: comparing the four main career paths

The four analyst tracks

"Analyst" is one of the most overloaded job titles in tech. A data analyst at Stripe writes SQL all day, owns a few dashboards, and runs ad-hoc cohort cuts for the growth team. A business analyst at a Fortune 500 insurer maps claims workflows in BPMN and writes requirements for a vendor RFP. A product analyst at DoorDash designs A/B tests, reads experiment readouts, and argues with PMs about whether the lift on day 7 is real. A systems analyst at a fintech translates "we want Apple Pay" into sequence diagrams, OpenAPI specs, and Jira tickets the engineering team can actually build from.

Same word. Four very different jobs. Before you spend six months learning the wrong stack, decide which track you actually want — because the day-to-day, the tools, and the salary band differ by 30-60% between them.

This guide compares the four most common analyst tracks side by side, lists the tools and salary ranges for each at US tech-company rates, and gives you a 90-day entry plan per track. The aim is not to convince you one is better; the aim is to make the trade-offs explicit so you choose with eyes open.

Side-by-side comparison

Here are the four tracks compared on the dimensions that matter when you're picking one to enter.

Salary bands above are US national medians for mid-level (2-4 YoE) roles at tech and tech-adjacent companies, sourced from levels.fyi and Glassdoor data through early 2026. New-grad roles run 25-35% lower; senior IC roles run 40-80% higher.

A few things to notice. The product analyst band is the highest because the role sits next to revenue decisions and equity often adds $30-80k/year on top of base at Airbnb, Uber, and Stripe. The business analyst band is the lowest in cash but BA roles are the easiest to enter without a CS degree and the most recession-resilient. Systems analyst sits in the middle on both salary and difficulty; it's the most underrated track if you came from QA.

Which track fits your background

Match your starting point to the track where your existing skills shortcut the learning curve.

Coming from a non-technical background (marketing, ops, sales)

Start with data analyst or business analyst. Both have shallow technical floors and reward people who already understand a domain. A marketer who can write a CASE WHEN query against the events table will out-deliver a CS grad who doesn't know what a conversion funnel is. Your domain knowledge is the moat; SQL is the bridge.

Coming from finance, banking, or consulting

Business analyst or product analyst. The structured-thinking habits from consulting (MECE, hypothesis trees, executive summaries) port directly. If you already have an MBA-adjacent toolkit and want to stay close to strategy, product analyst at a growth-stage company is the highest-leverage move — you'll be doing the same memo-writing but for retention curves instead of M&A decks.

Coming from software engineering or QA

Systems analyst or product analyst. You already read code, you understand APIs, and you've debugged production systems. The systems analyst track lets you keep that fluency while moving toward requirements and design. Product analyst is the alternative if you want to move further from code and closer to business decisions — but expect to spend the first six months learning experimentation stats properly.

Coming from a science PhD or research

Product analyst at tech, data analyst in traditional industries. PhD-level stats fluency is a genuine differentiator — half the PA candidates can't explain why a p-value isn't the probability the null is true. Lean into variance reduction, multiple-testing correction, and effect-size estimation in interviews.

Coming from BI or reporting roles

You're already a data analyst in everything but title. The next move is product analyst (add experimentation and unit economics) or senior data analyst (deepen Python, learn dbt). Don't restart from zero.

Entry plan for each track

Each plan assumes 10-15 hours/week of focused study and project work. Double the pace if you can study full-time.

Data analyst — 16 weeks

Weeks 1-6: SQL deep practice. Window functions, CTEs, gnarly joins, performance tuning basics. Drill 100+ problems on real datasets, not toy ones. Build muscle memory for the patterns that show up in interviews — top N per group, running totals, retention cohorts, funnel conversion.

Weeks 7-9: Product metrics and experimentation literacy. Learn DAU/MAU/WAU, retention curves, funnel math, what a controlled experiment actually proves. You don't need to design A/B tests yet — you need to read them without getting confused.

Weeks 10-12: Python and pandas. Just enough to clean a CSV, merge two tables, and produce a chart. Don't go down the ML rabbit hole; that's not the job.

Weeks 13-14: One BI tool deeply. Tableau or Looker. Build three real dashboards from public datasets and put them on a portfolio site.

Weeks 15-16: Mock interviews and applications. SQL screens dominate; practice them out loud, not just in a text editor.

Business analyst — 14 weeks

Weeks 1-3: Process modeling. BPMN 2.0 notation, swim lanes, event-driven process chains. Map three real processes you know (your old job's onboarding, an account-opening flow, an expense-report pipeline) end to end.

Weeks 4-6: Requirements engineering. User stories, acceptance criteria in Given/When/Then, the BABOK chapters on elicitation. Practice writing requirements for a feature you'd add to a public app — what would you write for adding 2FA to a banking app?

Weeks 7-9: Excel and basic SQL. Pivot tables, XLOOKUP, basic SQL SELECT/JOIN/GROUP BY. You won't write window functions, but you need to read the output of queries an engineer hands you.

Weeks 10-12: Stakeholder skills. Interview techniques, workshop facilitation, conflict resolution. These are the actual job; everything else is supporting cast.

Weeks 13-14: Portfolio and applications. A two-page case study showing one end-to-end requirements artifact — current-state process, gap analysis, future-state design, acceptance criteria — beats a CV bullet list.

Product analyst — 28 weeks

Weeks 1-10: Everything from the data analyst plan. Non-negotiable foundation.

Weeks 11-16: Experimentation theory and practice. CUPED, stratification, sequential testing, sample-size calculation, peeking corrections, holdout design. Read the Microsoft and Airbnb experimentation papers; they're public and dense and worth the hours.

Weeks 17-22: Unit economics. LTV, CAC, payback period, contribution margin, cohort retention curves. Build a working unit-economics model in SQL for a hypothetical SaaS — get comfortable defending the assumptions.

Weeks 23-28: Product judgment and mock loops. Read Stratechery and Lenny's Newsletter to build vocabulary; then run mock loops across SQL, experiment design, product case, and metrics deep-dive. Each round is its own skill.

Systems analyst — 24 weeks

Weeks 1-6: Database and SQL. Same as the DA plan but skewed toward schema design — entity-relationship modeling, normalization, indexing intuition.

Weeks 7-12: UML and API design. Sequence diagrams, class diagrams, activity diagrams. REST principles, OpenAPI 3.x, GraphQL basics. Read three real public API specs (Stripe, GitHub, Twilio) and reverse-engineer what trade-offs the designers made.

Weeks 13-18: Distributed systems literacy. Microservices, message queues (Kafka/RabbitMQ), caching layers, the CAP theorem at a working level. You don't need to build them; you need to spec features that span them without breaking consistency assumptions.

Weeks 19-22: Process and tooling. Jira, Confluence, ADRs (architecture decision records), Postman for API testing, Mermaid for diagrams in Markdown.

Weeks 23-24: Portfolio. Pick one feature in a public product and write a full functional spec for it — user stories, sequence diagrams, API contract, edge cases, error handling. This is the artifact a hiring manager wants to see.

If you're drilling interview questions for any of these tracks day by day, NAILDD has SQL, A/B, and systems-analyst problem sets organized by track and difficulty.

Common pitfalls

The biggest mistake first-time analysts make is picking the track by salary alone. Product analyst pays best on paper, but if you don't enjoy arguing with PMs about metric definitions or spending three weeks debating one experiment design, you'll burn out by month nine. Pick the track whose median day-to-day you can sustain for two years, not the one with the prettiest comp band.

A second trap is chasing every shiny tool. Candidates burn weeks on Spark, Airflow, dbt, Snowflake, and Databricks before passing a junior SQL screen. Interviewers see through tutorial-deep familiarity in five minutes. Pick the one tool the JD lists, learn it deeply, move on.

The third pitfall is building toy projects no one would ship. A Titanic dataset notebook signals you watched a Kaggle tutorial. A dashboard that monitors your local coffee-shop's foot-traffic data (real or scraped) signals you can scope a real analytics problem. The bar is "would a hiring manager believe a PM actually asked for this?" — if no, replace the project.

Fourth, underestimating soft skills, especially for BA and SA tracks. Both roles are 70% communication. If you can't run a 45-minute stakeholder interview, summarize the findings into a one-pager, and present it without filler, you'll lose offers to candidates with weaker technical skills but stronger presence. Practice this on video and watch yourself back; it's painful and it works.

Fifth, applying before you can pass a take-home. Every rejection burns referrals, future-cycle eligibility, and morale. Wait until you can finish a 4-hour SQL take-home in 2 hours with documented assumptions. Applying early costs more than waiting two more weeks.

Related reading

• Complete guide to becoming a data analyst
• How to become a data analyst from scratch
• How to transition from developer to data analyst
• SQL window functions interview questions
• A/B testing peeking mistake

FAQ

Which analyst role pays the most in the US?

Product analyst at a high-growth tech company has the highest ceiling — total comp at Meta, Stripe, or Airbnb can reach $250-350k for senior ICs once equity is counted. Senior data analysts at the same companies sit around $180-250k. Systems analysts top out lower in pure salary but the role is more recession-proof. Business analysts have the lowest cash ceiling but the highest job security in regulated industries like insurance and healthcare.

Which is easiest to enter without a technical degree?

Business analyst, hands down. The technical floor is Excel plus basic SQL, and many BA roles weight communication and domain knowledge over coding ability. Data analyst is the second-easiest if you're willing to do four to six months of SQL drilling. Product analyst and systems analyst both require deeper technical chops and are harder to crack as a first analyst job — most people enter PA after one to two years in a DA seat.

Can I switch tracks later?

Yes, and cross-track moves are common. The typical paths are DA → PA (learn experimentation and product judgment), BA → SA (deepen technical skills around APIs and data modeling), and SA → PA or engineering manager (use systems fluency to move closer to product or people leadership). The hardest jump is BA → PA because it requires picking up both heavy SQL and statistics; budget 12-18 months for that one. Switching is easier inside the same company than across companies because internal teams already know your work.

Do I need a master's degree?

For DA and BA, no — a portfolio plus a passable behavioral interview beats a degree. For PA at top-tier tech companies, a quantitative master's or PhD helps signal stats fluency but is not strictly required if your interview performance carries it. For SA, a CS or engineering degree is the most common path but bootcamp-plus-portfolio candidates do break in. No analyst track requires a specific certification, despite what bootcamps tell you.

Will AI replace these jobs?

The boring half of every analyst job — first-draft SQL, dashboard summaries, requirements from a meeting transcript — is already partly automated. The work that survives is scoping the right question, judging which model is misleading you, and arguing the result with stakeholders. Analysts who use Copilot, Claude, and Cursor as accelerators will outproduce those who don't.

How long until I'm earning the mid-level salary?

Plan on 18-30 months from your first analyst offer to the mid-level band in the table above. The first year is learning the data model and earning trust; year two is when scope expands. Faster promotions happen at growth-stage companies where headcount is doubling. Switching companies after 18-24 months typically beats waiting for the internal promo by 15-25% in total comp.