A 90-Day Roadmap to Learn Data Analysis

You do not need a degree to become genuinely useful with data. This 90-day plan takes you from spreadsheets to SQL and Python, ending with a real portfolio project.

Data analysis has a reputation for being hard to enter, guarded by statistics and computer science. In practice, the work that most teams actually need is far more approachable: cleaning messy information, asking sharp questions of it, and explaining what you found to people who do not want to read a spreadsheet. A focused learner can become genuinely useful in about ninety days by following a sequence instead of collecting random tutorials.

What data analysis really is

Strip away the jargon and the job is a loop. You get a question, you find or assemble the data that could answer it, you clean that data until you trust it, you explore it for patterns, and then you explain the result clearly enough that somebody can make a decision. Tools change constantly, but that loop has not changed in decades. Beginners struggle not because the tools are hard, but because they learn the tools without ever running the loop end to end.

The core tool stack

• Spreadsheets. Still the universal language of business data, and the fastest way to sort, filter, and pivot a few thousand rows before reaching for anything heavier.
• SQL. The language for asking questions of databases where the real data lives, and the single skill most often listed in analyst job ads.
• Python (with pandas). Where you go when the data is too big or the cleaning too repetitive to do by hand, and where you start automating your own work.
• Visualisation and storytelling. Charts and short explanations that make a finding obvious, because an insight nobody understands changes nothing.

Days 1 to 30: get comfortable with data

Spend the first month inside a spreadsheet and resist the urge to jump ahead. Learn how tables are supposed to be structured so each row is one record and each column one variable. Practise sorting, filtering, and building pivot tables, then learn the handful of formulas that do most of the heavy lifting: lookups, conditional logic, and basic aggregation. Most importantly, find a real dataset you care about — your own spending, a local sports league, public health figures — and answer three small questions about it. The discomfort of messy real data is the actual curriculum.

Days 31 to 60: SQL and your first scripts

In the second month, move to SQL. Install a database you can break without consequence, load a public dataset into it, and drill the basics until they are automatic: filtering with WHERE, grouping with GROUP BY, joining tables, and summarising with aggregate functions. Write the same question in SQL and in a spreadsheet and notice which feels clearer. Around week seven, begin Python gently. Learn variables, lists, and simple loops, then let pandas do the boring work of loading files, filtering rows, and grouping. Do not try to learn all of Python — learn just enough to reshape a table and you will already outpace most self-taught beginners. Expect this month to feel slowest. You are building vocabulary, and vocabulary always feels clumsy before it feels obvious. Keep a small file of queries and snippets you actually reuse, because that personal library will save you more time than any course certificate.

Days 61 to 90: build one real project

The final month is where you turn knowledge into evidence. Choose one question that genuinely interests you, gather a dataset large enough to be awkward, and take it through the entire loop: clean it, document every decision, explore it, and produce a short write-up with two or three charts and a clear recommendation. Publish it somewhere public, imperfect and all. One finished project that shows your reasoning beats twenty certificates, because it proves you can work without a tutorial holding your hand. Along the way, get comfortable explaining your choices out loud, since the ability to defend an analysis matters as much as producing it.

The bottom line

Ninety days will not make you a senior analyst, but it will make you someone who can take a messy question, find honest answers in data, and communicate them without overselling. Pick a boring schedule you can keep, run the full loop at least once from beginning to end, and let each project make the next one easier. Fluency here comes from repetition on real questions, not from finishing another course.