The best first step in statistics is not a course, a book or a video. It is this: take a real dataset you care about and calculate its average and its spread. Then ask yourself whether the average told you anything useful on its own.
That exercise, which takes about twenty minutes, teaches the central lesson of the subject and gives everything you learn afterwards something to attach to.
Why starting with a real dataset works
Statistics taught in the abstract feels pointless because it is - the formulas are answers to questions you have not asked yet. Starting with your own data reverses the order. You have a question, you hit a wall, and the technique that gets you past the wall is immediately memorable.
Good starting datasets: your own sales or invoice history, your website analytics export, your team's ticket times, a personal spending export from your bank, or an open dataset from a national statistics office on a topic you care about.
The first week, in five sittings
Sitting 1: describe it. Count the rows, find the average, the median, the minimum and maximum of your main number. Notice whether mean and median differ, and think about why.
Sitting 2: draw it. Make a histogram. Look at the shape - is it symmetric, skewed, does it have two humps? The shape is often the finding.
Sitting 3: split it. Break the data into groups (by month, region, product, customer type) and compare. This is where interesting things start appearing.
Sitting 4: measure the wobble. Calculate the standard deviation, and work out the range your data usually falls in. Ask whether last month's figure is actually unusual.
Sitting 5: write a paragraph. State one thing the data shows, one thing you are unsure about, and one thing the data cannot tell you. That third sentence is the beginning of statistical maturity.
What to learn next, in order
- Sampling and bias - who is missing from your data
- Confidence intervals - how to express uncertainty
- Hypothesis testing - whether a difference is real
- Correlation and regression - relationships between variables
- Forecasting - projecting forwards honestly
There is a full 30-day self-study plan if you want the detailed version.
What not to do first
- Do not start with probability theory. It is the logical foundation but a demoralising entry point.
- Do not learn Python or R before you understand what you are computing.
- Do not buy a thick textbook you will read two chapters of.
- Do not try to memorise formulas. Understand what question each one answers; the formula is looked up.
The mindset to bring
Curiosity plus scepticism. Every time you produce a number, ask "what would make this wrong?" That habit is the actual subject. The techniques are just how you answer.
When to get taught properly
Self-study gets you a long way but cannot tell you when your reasoning is subtly wrong. When you want feedback and structure, the Statistics for Business course at London School of Business UK teaches from first principles with no assumed background. Enquire today.