Page views and bounce rates

Page views and bounce rate describe different aspects of website use. Page views count content loads; bounce rate estimates sessions that do not meet a platform’s engagement rule. Neither metric establishes whether a visitor achieved the intended outcome, so both must be interpreted alongside the site’s purpose, conversion events and measurement configuration.

When to use it

• Answer the key performance question: “How effective is our internet strategy?”
• Assess traffic and engagement within the marketing-and-sales perspective.
• Diagnose differences by page, audience, channel, device and campaign.
• Plan definitions, event collection, reporting frequency and data-quality checks.
• Evaluate trends against the visitor outcome the page is designed to support.

Origins

Page-view measurement grew from early web-server logs and later client-side analytics. “Bounce” was introduced as analytics tools began grouping individual requests into visits and distinguishing single-page sessions from continued navigation. There is no one universal bounce-rate definition: products and versions have used different session and engagement rules. Modern Google Analytics defines bounce rate as the inverse of engagement rate, while older Universal Analytics material used a single-page-session definition.

What it is

Perspective: Marketing and sales perspective.

Key performance question: How effective is our internet strategy?

Page views are the total recorded views of pages or screens. Average page views per session divides those views by sessions. A higher value may indicate useful exploration, but it may also signal confusing navigation or content split unnecessarily across pages. A lower value can reflect poor targeting, unmet expectations or a page that answers the question efficiently.

For publishers, page views may inform advertising inventory. For other organisations, they help evaluate how content, navigation or campaigns change observed behaviour. Repeated views, automated traffic, consent choices and tagging errors can all affect the count.

Bounce rate depends on the analytics system. In the older interpretation used by much of the inherited benchmark material below, a bounce was a visit that ended without another tracked page interaction. A visitor might leave through an external link, close the window, enter another address, use the Back control or allow the session to time out. Under current Google Analytics, a bounce is a session that does not qualify as engaged; qualifying engagement can be based on duration, a key event or multiple page or screen views.

A high or low rate is not intrinsically good. A single-page answer, phone call, download or completed off-site action may create value without another page view. Conversely, a low rate may result from duplicate tags or irrelevant events rather than genuine engagement.

How to use it

Measurement

Write a measurement specification before interpreting either metric. Define page-view events, sessions, engaged sessions, internal traffic, bots, cross-domain journeys and the actions that represent user success. Record platform changes so that discontinuities are not mistaken for behavioural shifts.

Data collection method

Use consent-aware analytics tracking, server logs or both. Validate tags and events in a controlled test, monitor missing and duplicate data, and avoid collecting personal information that is unnecessary for the decision.

Formula

The formula in the figure reflects the historical single-page-session convention. If the analytics product uses an engagement-based convention, calculate and label the metric according to that product’s documented definition rather than mixing the two.

Frequency

Collection is continuous, but reporting should match decision cadence and traffic volume. Use enough observations to distinguish durable movement from normal variation, and annotate campaigns, site releases and tracking changes.

Source of the data

Sources include web-analytics platforms and server logs, supplemented by event systems, experimentation tools, search data and customer-outcome records. Reconcile the sources when material decisions depend on them.

Cost/effort in collecting the data

Basic collection may be free, including through Google Analytics, but trustworthy measurement is not costless. Implementation, consent management, tag governance, data validation, event design and analyst time all require effort.

Target setting/benchmarks

The following figures are retained as historical examples, not current targets. An archived US retail report from Core Metrics included page views per session:

Source: www.coremetrics.com

Avinash Kaushik’s often-repeated observation was that a bounce rate below 20% was difficult to achieve, above 35% deserved attention and 50% or more was worrying. Other historical ranges were:

• Retail sites with targeted traffic: 20–40% bounce
• Simple landing pages with one call to action: 70–90% bounce
• Portals such as Yahoo! and MSN: 10–30% bounce
• Self-service and FAQ sites: 10–30% bounce
• Content websites with high search visibility: 40–60% bounce
• Lead-generation sites: 30–50% bounce

These values are not portable across analytics definitions, eras, consent regimes, devices or page purposes. Build a baseline within one stable configuration, segment it by intent and compare the metric with completed outcomes.

Example

Yahoo! historically tested homepage changes by randomly assigning one or two hundred thousand visitors to an experimental experience and several million to a control. With high traffic, behavioural differences could appear within minutes, and the company reportedly ran about 20 experiments concurrently. The useful principle is controlled comparison: define the outcome and guardrails, randomise exposure, pre-specify the analysis and evaluate uncertainty. A change in page views or bounce rate should not be treated as a profit improvement unless downstream evidence supports it.

Top practical tip

Measure the action the visitor came to complete, then use page views and bounce rate to explain that outcome. Relevant content, clear navigation, visible search, restrained pop-ups and fast loading can help, but validate each change with behaviour rather than assuming more page views are better.

Top pitfall

Never compare bounce rates without confirming the definition and page objective. An 80% rate may be acceptable for a page that answers a question or generates a phone call, while a seemingly excellent result may be caused by duplicate tracking. Treat legacy benchmarks and platform migrations with particular caution.

Further reading

Historical Core Metrics reference: www.coremetrics.com

Historical article, Bounce Rate Demystified: http://blog.kissmetrics.com/bounce-rate/?wide=1

Archived Google Analytics reference: www.google.com/support/analytics/bin/answer.py?hl=en&answer=60127