Roadmaps: professions in data

3 main tracks in data work: DS/ML, analytics, data engineering. Each track has its own skill map, roadmap and possible career paths.

Analytics

The most common way into data

An analyst stands between a question like "why did revenue drop" and an answer someone can lean on when deciding. About half the work goes into getting the data and making sure it means what you think it means. The other half is explaining the result to the people who will decide. Closer to middle, experiments join in: A/B tests, unit economics and the conversation about what to do with the product next.

A typical week
  • Queries and extracts for the team’s questions
  • Recurring reporting and dashboards
  • Investigating anomalies: is the data broken or is this real
  • Designing and reading A/B tests
  • Research behind a specific decision: feature metrics, funnels, cohorts
Tools
Get it
SQLExcel / SheetsAmplitude / Mixpanel
Crunch it
Python / pandasGitΠŸΠ»Π°Ρ‚Ρ„ΠΎΡ€ΠΌΠ° A/B-тСстов
Show it
BI: Metabase, Superset, DataLens, Tableau
Where the jobs are

Banks and fintech, IT, e-commerce and marketplaces, retail and FMCG, telecom, consulting, advertising, medtech

Hiring is rarer in the public sector, utilities, energy, heavy industry, agriculture, hospitality, culture and non-profits. The data and the questions are there, but dedicated roles are few and the main tool is usually Excel.

Not your track if
  • You make regular slips in numbers and it does not bother you. Reports get used for decisions, so an error costs the company more than it looks.
  • You do not want to learn someone else’s business and only the methods interest you. Without the domain, a number can be read any way at all.
  • Explaining what you computed is hard work you would rather skip. The explaining takes about half of the working time.
  • You need a definite answer. A noticeable share of tests ends with "no visible difference", and that is a result too.

The path

Each stage is broken down by area of knowledge: SQL, Python, statistics, experiments, and so on. The same area comes back at the next stage at greater depth. Timings are a rough guide and depend on your background and how much you practise on real data.

The shared foundation

The first months look nearly the same on all three tracks. While you are working through this block you do not need to pick a specialisation: it will be useful on any of them.

Check: Take an open dataset on a topic you care about and find five facts in it in one evening. Each fact needs the query you computed it with. If that works out, the foundation is closed.
Where people get stuck

Learning SQL as syntax on trainer sites without touching real data. The syntax takes a week; the months go into understanding what specific tables mean.

Career paths

The first three grades look alike everywhere. After senior the ladder splits into an expert path and a management path, and what those two hold differs from track to track.

People arrive here from
  • BI-Π°Π½Π°Π»ΠΈΡ‚ΠΈΠΊΠ° β€” to add: statistics and hypothesis testing
  • ΠœΠ°Ρ€ΠΊΠ΅Ρ‚ΠΈΠ½Π³ ΠΈ web-Π°Π½Π°Π»ΠΈΡ‚ΠΈΠΊΠ° β€” to add: working-level SQL and Python
  • БистСмная ΠΈΠ»ΠΈ бизнСс-Π°Π½Π°Π»ΠΈΡ‚ΠΈΠΊΠ° β€” to add: SQL, Python, statistics, visualisation
  • UX-исслСдования β€” to add: the quantitative half: metrics and experiments
Junior0–2 years

Reporting upkeep and simple tasks: routine queries and dashboards built on definitions someone set before you.

Middle2–4 years

Owns a task end to end and sharpens the request themselves. Stakeholders stop spelling out what exactly to count.

You can step aside into
Product analyticsBI and reportingMarketing analytics
Senior4+ years

Owns the conclusions and how the whole team counts. Negotiation and reviewing other people’s numbers weigh as much here as Python.

You can step aside into
Product manager

After senior the ladder splits, and the choice is made once:

The expert path

No direct reports: methodology, metric definitions and reviewing numbers across the company. On the Russian market a separate grade often does not exist, and such a person is titled simply senior.

Lead analyst (IC)

Owns a whole domain, sets metric definitions and settles disputed numbers.

Staff analyst

Methodology and calculation standards across several teams. Found in companies with a large analytics function.

The management path

People, priorities and budget. You get to compute things yourself less and less, which is the main reason people step back off this path.

Team lead

A team of three to eight: priorities, reviews, hiring and onboarding.

Head of analytics

Analytics as a function: several teams, standards, budget, talking to leadership.

CDO

Data as a company asset. A rare grade: the role exists at large companies where data makes money directly.

Lead and head do not exist in every company. In a team of three analysts the ladder ends at senior, and the next step means moving to a larger company. Ask at the interview how many people are on the team and who runs it.

Roles inside the track

One track covers several job titles. They share the base but differ in subject and in what the person owns; in a smaller company they merge back into one job.

SQLExcel / SheetsPython / pandas

SQLPythonΠŸΠ»Π°Ρ‚Ρ„ΠΎΡ€ΠΌΠ° A/B-тСстов

SQLDAXMDX

ЯндСкс ΠœΠ΅Ρ‚Ρ€ΠΈΠΊΠ°, Google AnalyticsCRMΠ Π΅ΠΊΠ»Π°ΠΌΠ½Ρ‹Π΅ ΠΊΠ°Π±ΠΈΠ½Π΅Ρ‚Ρ‹

Google Analytics, ЯндСкс ΠœΠ΅Ρ‚Ρ€ΠΈΠΊΠ°Adobe AnalyticsΠœΠ΅Π½Π΅Π΄ΠΆΠ΅Ρ€Ρ‹ Ρ‚Π΅Π³ΠΎΠ²

Possible traps

A job title says less than the task list and the stack under it. Always read those two before you apply.

Data analyst, but the main tool is Excel
You will sometimes meet data analyst postings where the whole stack is Excel and exports from an accounting system. Such jobs exist, but they will not grow you: no SQL, no statistics, no marts, and a year later there is nothing to show for the experience.
A posting titled just "Analyst"
That word often stands for a systems or business analyst. It is a different profession: requirements, integrations and process descriptions, with no sampling, metrics or statistics. The stack gives it away: BPMN, UML, Swagger and Postman have nothing to do with data analytics.
A product analyst with no product
If the company has no digital product with a flow of users, there is no one to run an experiment on. There is plenty to count, but half the toolkit of the role does not apply, and at the next interview you will not be able to back up any A/B experience.
Where to start right now: take the first statistics lesson (mean and median), open a metric tree for an industry you know, and keep the glossary open for the words you meet along the way.
Breakdowns like this live in the channel

Data analytics in plain words: how to count metrics, how not to fool yourself in an A/B test, what gets asked in interviews and how this site gets built. The channel is in Russian.

Open the Telegram channel