Data analysis methods are easiest to choose between when you start from the question, not the method name. Is it "what is typical?", "how much does it vary?", "are these two things connected?", "could this be chance?" or "where is this heading?". The tables below group the common methods that way. The values come from one 30-day sample of orders from a made-up store, so you can see what each method actually returns; each name links to a page with the formula, a worked example and code for Excel, SQL and Python.
Descriptive and diagnostic: what happened, and what stands out
Descriptive methods summarise data you already have. Diagnostic methods look for where a result came from. They are the base of nearly everything else, and many business questions end here.
| Method | Answers | On the sample |
|---|---|---|
| Mean | What is the average day? | 32.73 |
| Median | What is the middle day, ignoring extremes? | 32 |
| Standard deviation | How far do days usually sit from the average? | 7 |
| Percentile | What value do most days stay below? | 38.1 |
| IQR outliers | How many unusual days are there? | 1 |
| Z-score | How unusual is one day (here, day 17)? | 4.04 |
The mean (32.73) sits a little above the median (32) because day 17, a one-day sale, pulls it up. When the two disagree, there is usually something skewed or extreme worth looking at before you quote either.
Relationships: are two things connected?
| Method | Answers | On the sample |
|---|---|---|
| Correlation | Do ad spend and orders move together? (−1 to 1) | 0.723 |
| Linear regression | How many more orders per ₹1,000 of spend? | 6.419 |
| R-squared | How much of the movement in orders does spend explain? | 0.523 |
Correlation and regression describe association, not cause. Both are also easily moved by a single extreme point, so run them with and without it.
Uncertainty and testing: could this be chance?
| Method | Answers | On the sample |
|---|---|---|
| t-test | Is the website–app difference bigger than chance? (p-value) | 0.0002 |
| Standard error | How precise is the average? | 1.28 |
| Confidence interval | What range is the average likely to sit in? (± half-width) | 2.61 |
A small p-value means the gap would be unusual if there were no real difference. It does not say the gap is large or important: always report the size of the difference next to it.
Trends over time: where is it heading?
| Method | Answers | On the sample |
|---|---|---|
| Moving average | What is the recent level with day-to-day noise smoothed out? (7-day, last day) | 36.57 |
| Linear trend forecast | If the straight-line trend holds, what is day 31? | 39.75 |
Predictive and prescriptive
Beyond the descriptive methods are predictive ones (forecasts and models that estimate what is likely next) and prescriptive ones (which choose between actions). They build on the same ground: a forecast is only as good as the history it was tested against, and a recommendation is only as good as the forecast beneath it. Before relying on one, ask how it did on data it had not seen.
How to choose
- Write the question as one sentence ending in a number or a yes/no.
- Pick the row of the tables above that matches it.
- Look at the data first (a line chart, a sorted list): a method cannot tell you that the data is wrong.
- Report the value with its size, its sample and what you checked it against.
Questions people ask
- What are the main types of data analysis?
- Descriptive (what happened), diagnostic (why), predictive (what is likely next) and prescriptive (what to do). Descriptive and diagnostic methods are the most common in everyday business work.
- What is the difference between a method and a technique?
- In practice the words are used interchangeably. A method is a defined calculation such as a mean or a t-test; a technique is usually a way of applying one or several of them, such as comparing periods or segmenting customers.
- Which statistical method should I use?
- Choose by question: averages for "what is typical", standard deviation for "how variable", correlation for "are these connected", a test for "could this be chance", and a trend for "where is it going".
Read more: All methods, with formulas and code, or try the live demo with no account.
Related guides
Methods used in this guide
All guides · Statistics methods · Business metrics · Free tools
Legal entity
Devansh Yadav
Moradabad, Uttar Pradesh
Governing law: Courts at Moradabad, Uttar Pradesh
Phone: +91 8433475698
Legal: legal@analyzeit.in