Homepages - good signposting is critical

Regardless of content, features and design, visitor behaviour on university unit homepages is pretty much the same. They're looking for a link to click and they do it very quickly.

I've been using a click analysis tool for about 3 years now. Partly to reassure website owners that cutting the content and focusing on the navigation is the right way to go, and partly to investigate the effectiveness of link labelling and locations.

Recently I looked back at click analysis data from 11 homepages, going back about 3 years. Some were in the new corporate style, others were non-standard designs.

A selection of homepage click analysis screenshots

Most were link heavy, focusing on signposting content to the visitor, although 3 were text heavy with up to 400 words. Two included a gallery of images which change every few seconds.

Regardless of the features and the layout, the story in the stats on the time taken to click is always the same. Website visitors just want to get what they want and get on their way.

Many usability and content management experts say the same:

• Steve Krug talks about satisficing behaviour - users choosing the first link they see that seems to have a reasonable chance of giving them what they want.
• Jared Spool talks about trigger words - the importance of learning what they are and getting them in the right locations.
• Jakob Nielsen has reported on numerous studies that highlight just how little website visitors read.
• Gerry McGovern talks about the importance of getting link and heading text right - succinct and meaningful and aligned to what the user wants to do.

Click analysis data patterns

In most cases, over 50% of visitors had clicked in under 10 seconds.

In all, over two thirds had left within 20 seconds.

After 30 seconds one homepage had 22% still to click, but almost all others had under 15%.

The vast majority of visitors have left within 30 seconds

Pages with large amounts of text tended to have the most visitors still on the page after 10 seconds ( up to 57%) but by the time 20 or 25 seconds had passed the rate of clicking had caught up with the others. On these pages, the data showed that most of the copy was not being read by the majority of visitors.

The most effective pages for getting visitors on their way quickly were those employing lots of task and audience-focused keywords and links, with minimal descriptive text. In these cases, over 90% had clicked within 20 seconds.

Statistics for homepages with an automatically rotating gallery of images were not noticibly different to those without.

Caveats

• The data analysed here has been collected over a range of time periods and had significantly different sample sizes - from 250 visits to 10 000.
• Where possible, the visits were filtered to include external or internal only audiences as appropriate, but in some cases the data is mixed. This is important because many key external audiences tend to visit the University's websites a small number of times, while internal audiences may well be frequent visitors.
• The data was collected using the Crazy Egg click analysis service which frustratingly changed the way they present time to click data about a year ago. This means I now only see the top 14 time intervals with all other clicks lumped into a category labelled 'other'. So sometimes there are gaps in the data. However, were the missing data present, it would only further accentuate the trends I've commented on.
• I've adopted the practice of cleansing the data, advocated by Jakob Nielsen, where I removed all excessively long periods of time to click. This is because the patterns would be skewed by visitors who haven't given the page their full attention - maybe it's their homepage on opening the browser, maybe they opened it in a background window or tab, or maybe they just wandered off to make a cup of coffee at that point. I looked at the number of words on the page, calculated how long it would take to read every one at 200 words per minute and then doubled this timeframe. Anything outside of this range I discarded.

How little do users read? - article by Jakob Nielsen