Can You Learn Data Science Without a Maths Background?

McTaba runs coding and data bootcamps in Kenya. This guide is written to be useful whether or not you ever join one.

Many people stop at this question. They remember a D in KCSE mathematics, a teacher who rushed through simultaneous equations, and a paper they never finished. They conclude that data science is for someone else.

The honest answer has two halves. You can start data science with very little maths. You cannot finish it with none. Both halves matter, and the useful part is the order in which the maths arrives.

This page sets out that order. It names what you need in the first month, what you need by month three, and what you can leave until the work demands it. It also says what our own programme asks of you at the door. For the wider career map, see the data science career map for Kenya.

One note on who is writing. Our founder holds a BSc in Mathematics and Computer Science, so none of this comes from someone who fears the subject. It comes from someone who has watched many beginners learn it late and do fine.

Think of maths in data science as four rooms you walk into one at a time. You never stand in all four at once.

When	The maths	What it looks like in real work
Day one	Arithmetic, percentages, ratios	Working out what share of this month's till takings came from airtime
Month one	Averages, medians, spread, simple charts	Seeing that one large transfer drags the average of a chama's contributions upwards
Month two to three	Probability, sampling, correlation, the idea of a fitted line	Asking if a drop in sales after a price change is real or just a slow week
Later	Linear algebra, calculus, optimisation	Understanding how a model improves its guess a little on every pass over the data

Look at the first two rows. A person who can read a Safaricom statement and tell whether a number is big or small already holds most of the maths they need to begin.

The third row is where beginners meet resistance. Probability feels strange because the answers are not yes or no. "Seven in ten" is the answer, and the work is deciding whether to trust it. That habit of mind matters more than any formula. Which statistics a data scientist uses goes through each idea with Kenyan data.

The last row sounds frightening, and it is the one most people overrate. In practice a library does the calculus for you. You need to know what the machinery is doing, not rebuild it with a pencil.

If you want to see how the early rooms feed the later ones, read the data science roadmap for beginners.

A lot. Much of a junior analyst's week uses no maths beyond month one. Here is what you can do with just a spreadsheet, SQL and a little Python.

• Clean data. Fix phone numbers typed three ways, strip "Ksh" from amounts, and make two date formats agree. This is most of the job, and it needs care, not maths.
• Count and group. Total sales by week, by branch, by product. SQL's GROUP BY does the arithmetic.
• Chart a pattern. A line of daily takings shows the Friday rush without any formula.
• Write the sentence. "Deposits at the branch fell in March because two agents closed." Clear writing is a data skill.

Try it this week. Request your own M-Pesa statement, load it into Google Sheets and total what you spent per week. That is a real analysis, and the only maths is addition.

Then do the same in Python. How to analyse an M-Pesa statement with Python shows each step. If your first love is spreadsheets, how to move from Excel to Python is the gentlest bridge.

Doing the work first has a second benefit. When a maths idea arrives later, you already have a pile of data where it answers a question you cared about. "Standard deviation" stops being a word in a textbook and becomes the number that tells you whether last Saturday's sales were normal.

A good rule: learn each piece of maths on the day a problem needs it. A beginner who studies probability for two months before touching data usually gives up. One who meets it while checking a real result usually stays.

Avoiding maths altogether leaves gaps that show up in interviews and in work.

You will trust a model that is wrong. A fraud model that is right 99 times in 100 sounds excellent, until you remember that nearly every transaction is honest. A model that always says "not fraud" scores the same. Spotting that takes basic probability.

You will misread a result. A sales rise of 4 per cent after a promotion might be noise. Without the idea of spread, you cannot tell.

You will hit a ceiling in machine learning. You can run a model from a library with little maths. You cannot tune it, explain it to a sceptical credit manager or notice when it has drifted without some feeling for what it does.

So treat maths as a debt you pay in small instalments as you go. The size of each instalment is small. Skipping all of them is what hurts. If you are hunting for an honest sense of how much, read whether you need maths to code for the programming side, then come back here for the data side.

Our Data Science to AI Engineering programme sets aside a phase of its own for this. It is phase 3 of 9, "Math, Statistics and Time Series", and it runs in weeks 7 to 8. Its topics, in our words: linear algebra and calculus through code, gradient descent from scratch, Bayes for fraud alerts, A/B testing and time features.

Two things stand out. First, the maths comes after weeks of Python, SQL and pandas, so you already have data to apply it to. Second, it is taught through code. You run gradient descent from scratch and watch a number shrink, rather than copy a proof from a board.

The entry requirement is stated plainly: No coding, data or maths background needed. We mean it. You need a laptop, internet and the willingness to practise. A laptop and internet. Labs run on Google Colab and Kaggle.

It is fair to be wary of any course that says "no maths". The honest version is "no maths first". The programme runs 5 months full time or 8 months part time, at 288 contact hours, with 24 weekly projects, so the maths arrives in small pieces with something to build each week.

The next cohort starts on 11 January 2027, and the programme costs KES 140,000. Read the full phase list on the Data Science to AI Engineering programme page, and decide for yourself. We do not claim graduate results yet, because the first cohort has not started.

If you are not ready for a cohort, here is a plan with no cost beyond time. Imagine you give it four hours a week.

1. Weeks 1 to 2. Work with percentages and ratios on your own spending. Build a pivot table. Nothing else.
2. Weeks 3 to 6. Learn the mean, the median and the spread. Learn why the median beats the mean on a list with one huge amount.
3. Weeks 7 to 10. Learn simple probability with coin and dice examples, then with your own data: how often does a customer return?
4. Weeks 11 to 14. Meet the fitted line. Fit one on shop sales against temperature or rainfall using a library.
5. After that. Only go deeper when a problem forces you to.

Pair the plan with code from the Python a data science beginner needs. When the maths feels too thin, find a friend who is good at it and swap: you explain the data, they explain the formula.

Keep one rule: every maths idea must touch a dataset within a week of learning it. Ideas that never meet data fade.

If a weak maths grade is the only thing holding you back, it is not enough reason to stop. Start with the first two rows of the table, with your own statement, this week.

If you want a structured route with deadlines and review, look at the Data Science to AI Engineering programme. If you would rather first decide between data and code, read whether to learn data science or software engineering. No school can promise a job at the end, and we make no such promise.