These slides have minimal content. I've transcribed some of what I said during the presentation in the slides that follow.

See the link below for some Google Analytics dashboards and custom reports that I discussed at the end of the presentation.

http://www.matthewculnane.co.uk/posts/2013/3/13/improving-bounce-rate

Trying to make sense of data in Google Analytics is hard.

The very many blogs and websites that cover analytics will tell you that you can draw easily very clear conclusions. But often it can be nightmarishly difficult to conclusively prove things one way or another. There are *so* many variables.

What we can do is dig deep into the data, spot patterns, and extrapolate some basic principles.

Overall, we hope we are serving our audience well. That will bring us success.

We've all seen bounce rate, but what does it actually mean?

The key (often missed) is that a bounce only involves cases where a user visits a single page then immediately leaves.

The bounce rate is the number of bounced visits expressed as a percentage of all visits that begin with that particular page.

Just because a user leaves after looking at a particular page, it doesn't mean it is a bounce if they looked at other pages beforehand. That's an exit, not a bounce.

The exit rate is the number of times a page is exited expressed as a percentage of all visits containing that particular page.

5 separate visits, 3 pages. What are the bounce rates and exit rates?

Page: bounce rate | exit rate

A: 33% | 0%
B: 50% | 33%
C: 50% | 100%

People can bounce for good as well as bad reasons.

The bad: irrelevant content, dislike the design, poor usability, page slow to load.

The good: They may have found all the information they need, or performed the one task they arrived to do. For OpenLearn this might be ordering a free booklet off the back of a BBC programme.

It's difficult to meaningfully measure the two groups, but time spent on a page is a partial indicator.

When users have finished reading/watching/playing with the page, ensure there's a logical next step for them to continue their learning.

On occasion we have no clear instruction of what to do next; no manually determined related content. Users might think there isn't anything else on that subject, and bounce.

On the flipside, sometimes we have too many choices. Maybe it's a content page with 12 options for further study. This is too many and some users will instead choose to bounce.

We've all been let down before when following a link, clicking on a search result or clicking an ad. Sometimes the page we land on isn't the one we expected.

We need to investigate how people are finding our content and ensuring those links accurately reflect the page content.

This is a connections tool on OpenLearn—users can explore what connects different C20th composers.

It had a high bounce rate (>80%).

My hypothesis was that the links under the content to further study were a little passive. There wasn't any real CTA.

I've amended the page to give more direction.

At the time of this presentation, not enough time had passed to measure whether it was successful. I added an annotation in GA to come back to it in a month.

A collection of resources about childhood and youth studies to support the 2013 series of Child of Our TIme.

I was asked by my academic colleagues to add links to dozens of pages of content that users might also like.

This would have made for a very long and link-heavy page, so I grouped the content into 5 categories and added links to those 'landing' pages instead.

This ensured that users were more likely to click through to a specific area of interest than if they had to scroll though dozens of links.

Each landing page was given 'noindex, follow' for robots.

This was imported content from a prior version of the website. It was a timeline but displayed in a table.

It had very high unique pageviews but an equally high bounce rate.

I turned it into an ordered list (with list-style:none), used the HTML5 `time` tag, added anchor links at the top of the page for important parts of the timeline, and added links to related content within the body of the timeline.

Again, too soon to draw meaningful conclusions, but the bounce rate appears to be dropping as users find more content they're interested in.

The anchor links are showing up in Google SERP snippets, which is increasing the number of unique pageviews from organic search.

The number of ratings and the average refer to a 1-5 rating tool on our site.

We all know this, but useful to remind ourselves

What is the best, most useful thing you can point users to after they view your content?

Our academic colleagues want us to point our users towards lots of different bits of content.

We have data that shows this isn't always a good idea. Show them.

Pick one main thing, and a couple of other related bits of content.

Always guide the user who wants to learn more, but don't ram dozens of possibilities down their throat.

Use GA dashboards and standard reports to identify pages that could use improvement.

Use custom reporting to find out the exact source. Is there a problematic referrer? Is organic traffic to blame? Is the content unsuitable for a specific device (or type of device)? Is the page ranking well for less-relevant keywords?

Find this out. Don't look at the big numbers. Find the story.

We should sweat the individual pages.

The aggregate data will be reported and we will be judged on it.

By paying attention to our individual projects, we can improve the overall picture, and we all win. (Especially users.)

Look for the worst performing pages and do something about them. Seek and destroy problems.

Make sure there's an obvious next step.

When you see problems, hypothesise what might be causing them.

Test out solutions. Give enough time for meaningful data before deciding whether it worked.

If it did, try it on other pages.

If it didn't, reverse any changes and try a different hypothesis.