Averages can damage your site

"Frostbite in one leg and the other's on fire - on average you're fine." Sometimes averages can mislead...

Here's a web example: an average visitor to an insurance website takes 3 days before she converts. The average visitor looks at 3 pages before converting.

Does this mean that an average visitor will look at the Monday page, then the Tuesday page, then convert on the Wednesday page?

No. It means that average is the wrong way to look at both of these figures.

Percentile is often a better tool when something has no upper limit. Like income or number of pages before converting or order value.

Economists talk about the 50th percentile of income which is the amount where 50% of the population earn this level or less. Average is a misleading measure of income - it gets distorted by a small number of very wealthy people. The average person earns less than the average wage. A lot less.

For web analytics the 80th or 90th percentiles are commonly used. On the insurance site mentioned above I found that 86% of visitors who ever converted did this on their first page which was (obviously) on their first day on-site. This was a specially-built Landing Page.

The average pages and the average time were distorted by the strange behaviour of some visitors who would prowl round the site over lots of days and occasionally convert.