Data analyst
From statistical literacy through SQL, Python and BI tools to experiments and data storytelling: the path to a first analytics role.
Sign in to track how many of the 38 roadmap tasks you have done.
For students and juniors who want to draw conclusions from data, not just make tables, and who are aiming at a data or business analyst role. The roadmap goes from statistical literacy and interpretation through tools (spreadsheets, SQL, Python or R, BI) to experiments and presenting results. By the end you will be able to answer a business question from raw data to a clear recommendation, and defend every number in it.
Steps
Step 1: Statistical literacy
Mean, median, variance, conditional probability and the p-value are the vocabulary an analyst speaks. Without them you can produce a chart but not a conclusion, and conclusions are what analysts are paid for. You are done when, for every summary value, you can say what it hides, and from a description of the data you can tell which measure is misleading.
Topics on the portal
Resources
Step 2: Interpreting and visualising data
Percentages vs percentage points, truncated axes, correlation vs causation: most errors in decision making arise when reading numbers, not when calculating them. Visualisation is both an analyst's strongest tool and the easiest way to mislead. You are done when, for a given type of data and question, you choose a suitable chart, can defend it, and can spot manipulation in someone else's chart.
Topics on the portal
Resources
Step 3: Spreadsheets: Excel and Google Sheets
external resourcesThe spreadsheet is still the most widespread analytical tool in the world and, in most companies, the first place where data lands. Pivot tables, lookup functions and conditional aggregations solve a surprising number of tasks without code. You are done when you build a pivot table from raw data, join two tables with XLOOKUP or VLOOKUP, and can say when a spreadsheet is no longer enough.
Resources
Step 4: SQL
SQL is the language you use to ask a database questions, and the database is where company data lives. SELECT, JOIN, GROUP BY, subqueries and window functions cover most analytical work. You are done when, without help, you write a query that joins three tables, aggregates by month and calculates each category's share of the total.
Tasks
Resources
Step 5: Python with pandas or R
When data does not fit in a table or an analysis has to be repeated every week, scripting comes in. Python with pandas is the standard today, and R with the tidyverse is an equally good choice, especially in academic and statistical settings: pick one and go deep. You are done when you load a CSV, clean missing values, aggregate by group and plot the result, all in one script.
Tasks
Resources
Step 6: BI tools: Power BI, Tableau, Looker
external resourcesA dashboard is a way to hand an analysis to people who will not read your code. BI tools combine a data model, visualisations and sharing, and many analyst job ads ask for at least one of them. You are done when you build a dashboard with three linked visualisations on real data, with a filter and a clear answer to one business question.
Resources
Step 7: Experiments and A/B tests
When you want to know whether a new version of a page or campaign really works, you need a controlled experiment, not a "before and after" comparison. Sample size, the p-value, the confidence interval and typical pitfalls (multiple testing, stopping early) decide whether a result is real. You are done when you design an A/B test including the required sample size and interpret its result correctly.
Tasks
Resources
Step 8: Data storytelling and business context
An analysis that nobody understands changes no decision. A good presentation starts with a question somebody asked, continues with one main finding and ends with a recommendation, and it understands the metrics of the business it is made for. You are done when you can summarise the result of an analysis on one page for a manager who has five minutes, without losing honesty towards the data.
Tasks
Resources
Step 9: AI tools for analysts
Language models can now write a SQL query, explain someone else's code, suggest a chart or summarise data, but just as easily they invent a column that is not in the table. An analyst who understands the limits of models uses them like a fast junior, not as an authority. You are done when you use AI to speed up routine analysis and can independently verify every calculation it generates.
Topics on the portal
Resources