The reason most people who want to learn data science never finish is not that the subject is impossibly hard. It is that they have no map. They jump from a YouTube video on deep learning to a blog on SQL to a course on statistics, learning everything in the wrong order, understanding none of it deeply, and eventually deciding they are "not smart enough." They were smart enough. They just had no roadmap. Here is one.
Follow this order. Resist the temptation to skip ahead to the exciting parts, because every stage here is the foundation for the next one, and a foundation you skipped becomes the wall you hit later.
Stage 1: The foundations (do not skip these)
Statistics and probability
This is the actual heart of data science, and it is the stage most beginners try to skip because it feels less glamorous than building models. You do not need a mathematics degree, but you do need to genuinely understand averages, distributions, correlation, probability, and what it means for a result to be significant. Every model you will ever build rests on these ideas. Skip them and you will be someone who runs code without understanding what it says.
Python programming
Learn the focused core of Python — variables, data structures, loops, functions, and how to work with libraries. You are not aiming to become a software engineer. You are aiming to be fluent enough to manipulate data and build models without fighting the language.
Stage 2: Working with data
Data manipulation and cleaning
Here is a truth the courses undersell: real data science is mostly cleaning messy data, not building glamorous models. Data arrives incomplete, inconsistent, and full of errors. Learning to load, clean, reshape, and prepare data is the skill you will use every single day on the job. Master it early and take it seriously.
SQL for pulling data
Most real-world data lives in databases, and SQL is how you get it out. It is far less intimidating than it sounds and pays off immediately, because almost every data job assumes you can write a query. A week or two of focused practice covers what most roles need.
Data visualisation
Being able to turn numbers into a clear chart is not decoration; it is how you understand your own data and how you communicate findings to people who will never read your code. An analysis nobody can understand is an analysis nobody will act on.
Stage 3: Machine learning
Only now — with statistics, Python, data skills, and visualisation in place — does machine learning make sense instead of feeling like magic. Start with the core ideas rather than the fashionable ones.
- Understand the difference between supervised and unsupervised learning, and when each applies.
- Learn the workhorse algorithms first — regression, decision trees, and the like — before anything exotic.
- Learn how to evaluate a model honestly, so you know when it is genuinely good and when it only looks good.
- Understand overfitting deeply, because it is the single most common way real models fail.
Stage 4: Projects and specialisation
This is where learning turns into employability. Take everything from the earlier stages and build complete projects — from raw data all the way to a result you can explain. Build three to five of them, cover a few different types of problem, and document each one clearly.
Once you have built a handful, notice which parts you enjoyed most and let that guide a specialisation: perhaps analytics and business insight, perhaps machine learning engineering, perhaps the newer world of generative AI. Specialising makes you findable and memorable, whereas being vaguely good at everything makes you forgettable.
How long does the whole roadmap take?
For someone starting from zero and putting in steady, part-time effort, six to twelve months to reach a genuinely job-ready level is a realistic estimate. The variable that decides where you land in that range is not talent. It is consistency — the boring, unglamorous habit of showing up several times a week for months.
The rules that keep you on the road
- Learn in order. The stages build on each other for a reason, and skipping ahead is the fastest way to get lost.
- Build as you go. Do not wait until you have "finished learning" to start a project, because you never finish learning and projects teach what tutorials cannot.
- Go deep before you go broad. One concept understood fully beats ten skimmed.
- Track your progress somewhere visible, so that on the hard days you can see how far you have already come.
Data science is not reserved for geniuses. It is reserved for people who follow a sensible order and keep going. Now you have the order. The rest is showing up.
CPLC’s data science programme follows exactly this build-in-order approach, with mentors who keep you from skipping the foundations. Book a free demo to see the full roadmap in action.
