Data & Analytics
SQL, pandas, statistics and visualisation — plus the part that actually gets people hired: turning an analysis into a decision somebody acts on.
Finish this and you can take a messy dataset, establish what is true in it, quantify how sure you are, and present it to someone who will not read your code.
-
SQL foundations
BeginnerThe single highest-leverage skill in this field. Joins, aggregation and window functions, in that order.
- SQLBolt Twenty short exercises in the browser, no setup. The fastest way to go from nothing to writing joins.
- Mode — SQL Tutorial Written by analysts, against realistic data. The intermediate and advanced sections are where window functions land.
- PostgreSQL Exercises Progressively harder problems with solutions. Do these until joins stop requiring thought.
- PostgreSQL — Window Functions Running totals, rankings and period-over-period comparisons. This is the topic that separates junior from mid-level in SQL interviews.
Build this
An analysis answered entirely in SQL
Load a public dataset into Postgres and answer five business questions without leaving SQL — no pandas, no spreadsheet.
The constraint teaches you how much can be done at the database layer, which is the difference between a query that takes a second and a notebook that takes ten minutes.
Done when: Five documented queries with their results and a note on what each one tells you.
Any open dataset with a real schema -
Python for data
Beginnerpandas for reshaping, NumPy underneath it, and notebooks kept tidy enough that someone else can rerun them.
- pandas — Getting started The 10-minute tour, then the comparison-with-SQL page — it maps what you already know onto the new syntax.
- Python Data Science Handbook Free in full, notebooks included. Chapters 2 to 4 are the working core of the job.
- Tidy Data Twenty pages that will reorganise how you think about every table you touch. Read it early.
Build this
Clean a genuinely messy dataset
Find real data — government portals are ideal — with inconsistent dates, duplicate entities and missing values. Clean it, and log every decision you made and why.
That log is the deliverable. Anyone can call dropna(); explaining why dropping those rows did not bias the result is the actual skill.
Done when: A reproducible cleaning notebook and a written decision log.
Open Government Data, India -
The statistics you need
IntermediateDistributions, uncertainty, and the tests you will actually run — with enough understanding to know when the answer is 'this sample cannot tell you that'.
- Seeing Theory Visual, interactive probability. The best first exposure to distributions and the central limit theorem there is.
- StatQuest — Statistics Fundamentals Clear to the point of being disarming. Watch the p-value and confidence interval episodes twice.
- Think Stats Statistics taught through code rather than proofs. If maths notation loses you, start here instead of a textbook.
- A/B testing tools and pitfalls "How Not to Run an A/B Test" is required reading before you ever call a test early.
-
Visualisation
IntermediateChoosing the right chart, then removing everything from it that is not the answer.
- From Data to Viz A decision tree from your data's shape to the right chart, with the common mistakes attached to each. Bookmark it.
- seaborn — Tutorial Statistical plots without fighting matplotlib. Learn the figure-level functions and most charts become one line.
- Matplotlib — Tutorials You need this for the final ten per cent of control — annotations, axes and export for print.
- Storytelling with Data — blog Before-and-after makeovers of real charts. The single fastest way to develop taste in this area.
Build this
A chart makeover, with the reasoning
Find three genuinely bad charts in published reports or news articles. Rebuild each one, and write a short paragraph on what the original obscured and what your version reveals.
Done when: Three before/after pairs with the reasoning written out.
-
Warehouses and pipelines
AdvancedWhere analytics data lives once more than one person depends on it — modelling, transformation and tests that run on a schedule.
- dbt — Documentation Transformation as version-controlled, tested SQL. This is the standard practice in analytics teams now, and it is genuinely learnable in a weekend.
- Dimensional modeling techniques Facts and dimensions, from the people who defined them. Older than most of the tooling and still the vocabulary everyone uses.
- DuckDB documentation An analytical database that runs in-process with no server. Ideal for learning warehouse patterns on a laptop.
Build this
A small warehouse with tests that fail loudly
Take a raw dataset, model it into fact and dimension tables, and add data tests — uniqueness, referential integrity, accepted ranges. Then deliberately corrupt the source and confirm the tests catch it.
Done when: A modelled warehouse, a passing test suite, and evidence that it catches bad data.
-
Communicating the analysis
IntermediateThe last mile, and the one that decides whether your work changes anything.
- Google — Technical Writing Courses Short, practical, and it fixes the two habits that make analyses unreadable: passive voice and burying the conclusion.
- Writing that replaces a slide deck Written for public services, applicable everywhere: front-load the answer, cut the qualifiers, write for the reader in a hurry.
Build this
One page that changes a decision
Take your best analysis and compress it to a single page for someone who will not read the notebook: the recommendation first, the evidence second, the caveats honestly stated third.
Then give it to someone outside the field and ask them what you are recommending. If they cannot say, the page is not finished.
Done when: A one-page recommendation that a non-technical reader can restate correctly.
Not sure this is your path?
A path is only worth the hours if it is the right one. The counselling session tests the direction you arrived with rather than simply agreeing with it.
Find your directionClose, but not quite yours?
These nine are fixed on purpose — complete, sequenced, and ready today. If your direction sits between two of them, we will build the path around you instead.
Ask for a custom path