Analytics
Metrics, dimensions and reporting
Web analytics turns behaviour into evidence. Distinguish users (unique people or devices), sessions (visits, which end after inactivity) and pageviews (each page load; one session can contain many). Bounce rate is the share of sessions that ended on the entry page without interaction - it is not simply "left the site", and on a single-page site a high bounce rate may be normal. Conversion rate = conversions ÷ sessions × 100. Traffic source reporting splits channels into organic search, paid search, direct, referral, social and email; "direct/none" usually means untagged or unknown, not necessarily typed-in.
Dimensions describe (channel, country, device); metrics measure (sessions, revenue). KPIs are the small subset tied to objectives. Attribution assigns credit - last-click gives all credit to the final touch and undervalues upper-funnel work, which is why first-click and linear models exist. Set up goals or events and UTM tagging before launch, or data is lost. Filter internal traffic, watch for sampling and consent-driven gaps, and judge A/B tests on statistical significance and adequate sample size rather than early movement.
Traps
- Defining bounce rate as "time on site" or assuming high bounce is always bad.
- Comparing sessions with users as if they were the same unit.
- Trusting last-click attribution as a complete picture of channel value.
- Forgetting that untagged campaigns appear as direct traffic.
- Calling an A/B test after two days of data.
Rule of thumb
Before quoting any number, ask three questions: what is the unit (user, session, pageview), what is the time window, and what would this metric look like if tracking were broken? That last check catches most reporting errors.