Bounce rate is one of the most useful website traffic metrics. It can be defined as the percent of users that arrive to your website and leave immediately. The typical scenario is as follows: a user is searching for some specific information, then he clicks on a link pointing to your website, and he instantly discovers that your website has nothing to do with what he was looking for.

According to this definition, having a 0% bounce rate would be great. That would mean that every one of your visitors found what he was really looking for, and then he stayed in the site quite a while.

Nevertheless, real life situations are even more complex than they seem. It all started with a customer who was trying to increase the traffic and number of sales through his ecommerce website. When asking him about how well was his site currently performing, he answered that he had a bounce rate of less than 5%.

At first, I thought he was talking about the bounce rate of a specific landing page. Achieving low bounce rates for specific pages is possible (and it's specially desirable for pay per click campaigns). Anything lower than 20% or 15% would be a good job, so a bounce rate of less than 5% seemed like a great job of single page optimization.

But then I found that such bounce rate of less than 5% was not a single page measure: my customer was talking about the bounce rate of the whole website. In my experience, having such a low bounce rate for a whole ecommerce website is a utopia. So hearing such news really raised red flags.

The context of the website had to be taken into account in order to produce reliable traffic reports. The key was to analyze the bounce rate for each one of the traffic sources. Then the answer to this mystery seemed obvious: the website was getting most of its traffic from users that directly typed the website address on their browser address bar. In other words: the website had very little referring links, and was nearly unknown to search engines.

After some search engine submission and search engine optimization the overall bounce rate of the website moved to more normal amounts. Before the optimization, only users who really knew the product were coming to the website. But now the product was being known by lots of new users.

Measuring the results in absolute numbers, the data demonstrated that more people were staying longer on the site, and that sales number increased. Nevertheless, tendencies are more important than absolute data. At first, one could be scared by the high spike in the overall bounce rate metric. But in fact, the bounce rate for direct traffic was kept in a very low percent, which demonstrates that things didn't got worse after performing the changes. And the growth of staying visitors was higher than the growth of the whole bounce rate itself. So the conclusion is that the optimization process did work.

I always advise to take into accounts these factors when using bounce rate metrics:

• Make sure what your “bounce rate” metric is actually measuring. Some bounce rates refer to “just 1 page visitors”, while my preferred bounce rate is the number of users who “instantly leave” your website, because it gives useful information for nearly any kind of website.

• Track your bounce rate per traffic source. Distinguish the bounce rates of search engine traffic, referral links and direct traffic.

• A bounce rate lower than 25% is a good job, while a bounce rate higher than 40% needs optimization. Any bounce rate over 60% would be a real problem. And, besides it may look like a dream come true, an overall bounce rate of less than 5% is a bad sign that requieres further analysis.

The idea is that you should not be obsessed about a single metric. Real world situations are even more complex than they seem. Analyze your data in the context of each specific website to find out true useful results.