Website: how long do users stay?

How long will users stay on a web page before leaving? It is a recurring question, but the answer has always been the same: not very long!

The average page visit lasts just under one minute.

When users rush through web pages, they only have time to read about a quarter of the text on the pages they actually visit (let alone all the ones they do not look at). As you will appreciate, unless your writing is extraordinarily clear and targeted, very little of what you say on your website will reach your customers.

However, while users are constantly in a rush on the web, the time they spend on individual page visits varies considerably: sometimes people bounce immediately, other times they linger far longer than a mere minute. Given this, calculating an average is not the best way to analyse user behaviour. Users are human beings — their behaviour is highly variable and cannot be accurately represented by a single figure.

Leaving web pages: the Weibull hazard function

Research by Chao Liu and his colleagues at Microsoft Research now provides a mathematical understanding of user page-exit behaviour. The scientists collected data from a "popular web browser plug-in", analysing visit durations across 205,873 different web pages for which they had captured more than 10,000 visits.

Result: the time users spend on a web page follows a Weibull distribution.

What is a Weibull distribution?

Weibull is a reliability engineering concept used to analyse the failure time of components. The hazard function of the model indicates the probability of a component failing at time t, given that it has been functioning correctly up to that point.

So, after replacing a spare part in a piece of equipment, Weibull analysis predicts when you will need to replace it again. It also allows you to carry out risk analysis beyond the mean time to failure. And if you have a large amount of equipment, you can use an overall analysis to manage your spare parts inventory, for example.

Of course, when analysing web visits, we simply replace "component failure" with "the user leaves the page". In their research paper, Liu and his colleagues provide an intensive statistical analysis demonstrating that the Weibull model closely matches the empirically observed behaviour of users.

According to earlier research, there are 2 different types of Weibull distribution:

• Positive ageing: the longer a component has been in service, the more likely it is to fail. In other words, the hazard function increases for larger values. This makes intuitive sense, because the longer something is used, the more it wears out. So something that has been in service for a long time is approaching its breaking point.
• Negative ageing: the longer a component has been in service, the less likely it is to fail. Here, the hazard function decreases for larger values. This makes sense in the fact that individual components vary in quality: poor-quality components generally fail early, so anything that has been in service for a long time is likely to be particularly robust and will generally continue to operate for some time yet.

Negative ageing: leave quickly or stay a long time

Researchers found that 99% of web pages display a negative ageing effect. In human-computer interaction (HCI) research, it is extremely rare to obtain such a robust conclusion, and Liu and his colleagues deserve credit for discovering a major new insight.

Why negative ageing? Because web pages vary enormously in quality. Users know this, and spend the majority of their initial time filtering pages to weed out potential "rubbish". It is rare for people to linger on web pages, but when users decide a page is valuable, they may stay on it for considerably longer.

The following graph shows the hazard function — that is, the probability of leaving — for the median Weibull parameters fitted across the full scientific dataset:

It is clear from this graph that the first 10 seconds of a page visit are critical in users' decision to stay or leave. The probability of leaving is very high during these first few seconds because users are extremely sceptical, having already encountered countless poorly designed web pages.

If the web page survives this extremely harsh 10-second first impression, users will look around a little. However, they are still likely to leave during the following 20 seconds. The curve becomes relatively stable after the 30-second mark. Of course, visitors continue to leave every second thereafter, but at a much slower rate than during those first 30 seconds.

Consequently, if you manage to convince users to stay on your page for more than half a minute, there is a good chance they will stay for much longer — often 2 minutes or more, which is an eternity on the web.

In summary, two scenarios emerge:

• Poor pages, which are closed within seconds;
• Good pages, which may be visited for several minutes.

Note: "good" or "poor" is a subjective judgement that each user makes within the first few seconds of arriving. The design implications are clear: to earn several minutes of a user's attention, you must clearly communicate your value proposition within the first 10 seconds.