3Course 3Business analyticsIntermediate
Statistics for Business Decisions
Enough statistics to tell a real change from noise, size a sample, and read an A/B test result without being fooled.
- Chapters
- 10
- Time
- 8 hours
- Format
- Self-paced
- Tools
- Spreadsheets and Python notebooks provided; no maths beyond school algebra
What you will be able to do
- Describe a metric with the right summary: mean, median, percentiles, and spread.
- Explain variance and why last week's dip is probably nothing.
- Set up an A/B test: hypothesis, metric, sample size, duration, and stopping rule.
- Read a p-value and a confidence interval correctly, and know what they do not tell you.
- Recognise confounding, survivorship, and selection bias in your own data.
Chapters
10 chapters. Chapter 2 is free below.
- 1
Describing data
Centre, spread, shape. Percentiles as the everyday tool.
- 2
Variation is normalFree to read
Why every metric wobbles, and how to tell wobble from change.
- 3
Sampling
What a sample can and cannot tell you about the whole.
- 4
Confidence intervals
A range you can defend instead of a single number you cannot.
- 5
Hypothesis tests
The logic of the test, p-values, and the mistakes everyone makes with them.
- 6
Designing an A/B test
Metric, minimum detectable effect, sample size, duration, and when to stop.
- 7
Reading an A/B result
Significance, practical size, and the peeking problem.
- 8
Correlation and causation
Confounders, reverse causation, and the questions to ask before acting.
- 9
Biases in business data
Survivorship, selection, and Simpson's paradox with examples from Indian consumer businesses.
- 10
Capstone: a pricing experiment
Design, analyse, and write up a test on checkout conversion.
Free chapter 2 of 10
Variation is normal
Every metric you track moves every day, and most of that movement means nothing. Conversion was 3.1% on Monday and 2.8% on Tuesday. Someone asks what went wrong. Usually the honest answer is 'nothing; that is what conversion does', but very few people can say that with evidence.
The evidence is the metric's own history. Take the last ninety days of daily conversion and look at how far it usually moves from one day to the next. If a 0.3 point swing happens a few times a month, Tuesday is unremarkable. If it has never happened before, Tuesday is worth a look. This is the whole idea behind control charts, and it needs nothing more than a spreadsheet.
What makes this hard is not the arithmetic. It is that dashboards show today's number in large type with a red or green arrow, and the arrow does not know about variation. One of the most useful things an analyst can do is add the band of normal variation to the chart, so the arrow has context.
This chapter builds that band step by step, first by eye, then with standard deviation, then with percentiles, which turn out to be more robust when the data has spikes. By the end you will have a rule for your own metrics: how big a move has to be before it deserves a meeting.
Who it is for
Analysts and managers who present numbers that drive decisions and want to know when a difference is real.
Before you start
Data Analysis Foundations or equivalent comfort with spreadsheets. No prior statistics.
Taught by
Paul Montero
Founder and instructor
Questions about this course
How long do I have access?
Is the dataset included?
Can I get a refund?
incl. GST, 7-day refund