How to read common web analytics metrics
Analytics reports are full of numbers that sound obvious and are not. This page defines the common metrics, gives the formula behind each one, and works through a single example month so you can check your own reports.
Counting things: users, sessions, pageviews
- Pageview: one load of one page. Reloading counts again.
- Session: a visit, meaning a group of interactions by one browser that ends after a period of inactivity or, in some tools, at midnight. One person can produce several sessions in a day.
- User: an estimate of distinct people, usually based on a browser cookie or identifier. The same person on a phone and a laptop counts as two users, and clearing cookies makes a returning visitor look new. Treat user counts as approximate.
The example month
All the calculations below use the same made-up figures for a small shop website:
| Figure | Value |
|---|---|
| Sessions | 1,200 |
| Pageviews | 3,000 |
| Sessions that viewed only one page and did nothing else | 540 |
| Orders (conversions) | 36 |
| Revenue | $1,440 |
Ratios and their formulas
3,000 ÷ 1,200 = 2.5 pages per session.
540 ÷ 1,200 × 100 = 45%. Some analytics tools define bounce differently, for example as a session that was not engaged by time, clicks or a second page. Always read your tool’s definition before comparing numbers between tools.
36 ÷ 1,200 × 100 = 3.0%. If your tool divides by users instead of sessions, the result will be higher, because there are fewer users than sessions.
revenue per session = revenue ÷ sessions
$1,440 ÷ 36 = $40.00 per order. $1,440 ÷ 1,200 = $1.20 per session.
Points versus percent
Suppose next month the conversion rate rises from 3.0% to 3.6%. That is an increase of 0.6 percentage points. As a relative change it is 0.6 ÷ 3.0 × 100 = 20%. Both statements are true; saying “conversion went up 20%” without saying which you mean confuses people.
Click-through rate
For search results, email links or ads, click-through rate is clicks divided by impressions. Example: 84 clicks from 2,800 impressions gives 84 ÷ 2,800 × 100 = 3.0%.
Reading the numbers sensibly
- Small samples swing. On a day with 40 sessions, one extra order moves conversion rate by 2.5 points. Judge rates over enough traffic.
- Averages hide spread. A few very long sessions can pull average session time up while most visits stay short.
- Ad blockers and consent choices mean analytics never sees every visit. Server logs and order records are useful cross-checks.
- Compare like with like: the same weekday, season and campaign period.
Questions
Is a high bounce rate bad?
Not always. A visitor who reads a single article fully and leaves satisfied counts as a bounce under the classic definition. It matters more on pages meant to lead somewhere, like a product listing.
Why do two tools show different session counts?
They use different session timeouts, bot filters, consent handling and tracking methods. Pick one source for trends and stick to it.