What is Web Analytics?

Web analytics is the measurement, collection, and analysis of website usage data: traffic sources, sessions and users, pages and screens viewed, paths taken, time on page, conversions, and exits, gathered automatically from every visitor rather than from a recruited sample. It is quantitative and observational: no manipulation, no questions asked, just the behavioural residue of the whole audience, which makes it uniquely complete on what people did and uniquely silent on what they meant, wanted, or felt while doing it. Its instruments are the tags, scripts, and server logs that capture events, and the platforms that turn them into reports, and its output is the dashboard most product and marketing decisions begin from.

What It Measures Well, and What Has Changed

The classic metrics (page views, sessions, bounce rate, conversion rate, funnel drop-off, traffic source) remain the vocabulary, but the measurement ground has shifted. Cookie restrictions, consent requirements, ad blockers, and browser privacy changes mean that a meaningful share of visitors are now unmeasured or measured incompletely, so analytics numbers describe the consenting, trackable population rather than everyone, a coverage problem hiding inside data that looks like a census. Modern platforms have moved from page-centric to event-based models (any interaction can be an event, not just a page load), which blurs the line between web analytics and product analytics, and machine-modelled estimates now fill the gaps consent leaves. The practical consequence: treat analytics figures as very large but imperfect samples, and check whether the missing visitors differ from the measured ones.

The What/Why Gap

Analytics will tell you that 60% of visitors leave the pricing page within ten seconds. It cannot tell you whether the price was too high, the page confusing, the audience wrong, or the visitors merely comparing tabs. That is the structural limit of behavioural traces: rich on frequency and sequence, empty on intent and experience. Every serious analytics finding is therefore a research question in disguise, answered by watching people use the page, asking them what they were trying to do, or running a short intercept study with the visitors themselves. The reverse pairing is equally valuable: a usability problem found with five participants gets its prevalence from the funnel data. Analytics says how many; research says why; the loop between them is where mixed-method product work lives.

Using It Well

1. Instrument the decisions, not the site.
Define the handful of events and conversions that map to real goals before tagging everything; an analytics implementation designed around questions produces answers, one designed around coverage produces reports.

2. Version and document the tracking.
Changed tag definitions create false trends; the analytics changelog is a research instrument's revision history.

3. Segment before concluding.
Aggregates hide the story: new versus returning, device, source, and market routinely reverse an overall pattern.

4. Respect the causal line.
Analytics correlations ("users who visit the docs convert more") are associations until an experiment says otherwise.

5. Close the why-gap deliberately.
When a metric moves, run the small qualitative study that explains it: a prototype task with recorded reactions, or open-ended video answers from the affected audience (a fast Ballpark study is the natural companion to a surprising dashboard), so that decisions rest on mechanism rather than on a guess about a number.

The Takeaway

Web analytics is the behavioural record of the whole trackable audience: complete on what happened, blind on why, and increasingly a large sample rather than a census as privacy reshapes measurement. Instrument around decisions, version the tracking, segment before concluding, keep correlation in its place, and pair every important movement with the research that explains it. The dashboard poses the questions; it rarely answers them.