How bad is a bounce on your page? What does that high bounce rate mean?
Interpreting bounce rate is not as simple as some guides put it.
- there are different kinds of bounces: bad, very bad, and not bad
- bounce rate is not an indicator of content engagement
How Google Analytics measures bounce rate
By default in Google Analytics all one page visits are called bounces, whether the user read the page or not.
So, the way Google Analytics measures it, a bounce could mean anything between these two extremes
- The visitor saw the page and left immediately – (described by Google as a “puke”)
- A one page visit where the visitor read the page and went away happy
Unfortunately, by default, Google Analytics doesn’t distinguish (1) from (2). From a site management point of view, the difference can be key:
- For a page whose purpose is to deliver information – product support, government, not-for-profit, etc – a bounce of type (2) can represent a perfectly satisfactory visit for everyone involved.
- Even on an ecommerce site, where the main aim of a page is to get the user to make that next click, there is a world of difference between the two types of bounce:
- a type (1) bounce is probably someone in the wrong place, if you are paying for this traffic, you probably shouldn’t be.
- a type (2) bounce is a user who has engaged with the content, but hasn’t made the next step – there may be a dozen ways to improve this situation.
The issue is the way GA measures bounce rate. In his “standard metrics revisited” discussion of bounce rate Avinash Kaushik talks about two ways to measure bounce rate:
- The percentage of website visitors who see just one page on your site.
- The percentage of website visitors who stay on the site for a small amount of time (usually five seconds or less).
Kaushik says he prefers the second definition (as do I!). However, Google Analytics uses the first – it simply counts every one page visit as a bounce
Worse, when a visitor bounces, both time-on-page and session-time are recorded as zero (because of the way GA measures session time and time on page).
The upshot is that, for a high bounce rate page (or traffic source)
- you are left to wonder what that bounce rate means – in terms of how many are type (1) bounces and how many are type (2)
- your time-on-page and session-time stats are likely to be a long way from the truth
The solution: measure actual time on page / in-focus, and scrolling using event tracking
When you measure actual time on page, and scrolling using event tracking, you can see how many of your ‘bounce’ users are spending enough time to read the page, and scrolling to the bottom of the page.
If you choose to, you can make these events ‘interactive’ so that a bounces are not counted where a visitor scrolls or spends time on a page (effectively implementing Kaushik’s second definition)
How to measure time on page
I use a modified version to LittleData’s time on page tracking script here. I’ll cover my modifications in a future post
How to measure scrolling
See my post on measuring scrolling for a script and implementation details
Example
A client has a site that publishes information as a public service – on a not for profit bases. This is the story of two real pages on their site. Both pages are getting thousands of pageviews a month – this is not weird effect of a tiny sample
Numbers from GA out of the box
Bounce rate | Avg time on page | |
Page A | 83% | 03:45 |
Page B | 27% | 03:12 |
We see that page A has a worrying bounce rate, while both pages have a similar average time on page, with page B being about 14% shorter.
Numbers from event tracking
% scroll to bottom | % of visitors staying on page for 1 minute or more | median time on page | |
Page A | 81% | 39% | 00:39 |
Page B | 36% | 19% | 00:27 |
As far as content engagement goes the numbers are almost the reverse of what you might infer from bounce rate!
81% of page A visitors scroll all the way to the bottom, while only 36% of page B visitors reach the bottom of the page.
Once the bounce visitors are included in the time on page metrics we also see a much more stark difference between page A and B – with visitors engaging far longer with page A.
When we started measuring time on page and scrolling, we found this in a lot pages – particularly information pages. The key point is that a high bounce rate does not always mean an unsuccessful page.
Bounce rate, as measured by Google Analytics out of the box, is a very poor proxy for content engagement. If you want to measure content engagement, use event tracking to measure page interactions and time in focus.