SEO experts have debated the relevance of Google Analytics bounce rate to search engine ranking for years.
Yet, several misconceptions still surround the topic. The result is that many website owners optimise for lower bounce rates to improve their ranking, while lower bounce rates don’t simply translate into higher rankings.
A well-documented conversation between Rand Fishkin, Founder of Moz, and Andrey Lipattsev, Google’s search quality senior strategist, illustrates this point perfectly. It turns out that when Rand ran tested intentionally, increasing the bounce rate of various pages over a few days, the results were inconclusive. In fact, in an almost perfectly even split, half of the rankings on the search engine results page (SERP) did change while half of the rankings did not.
Furthermore, Matt Cutts, Google’s former head of Webspam, has denied Google’s use of bounce rates (and other Google Analytics metrics) in ranking algorithms.
The answer seems clear enough – bounce rates don’t affect SERP. Except, if that were truly the case, how do we explain Backlinko’s conclusion that low bounce rates are associated with higher rankings? Answering that question requires understanding three issues:
- First, what does bounce rate measure?
- Second, why doesn’t Google use bounce rate as a metric?
- Third, how does bounce rate relate to other SEO factors?
Finally, what does bounce rate measure?
Bounce rate is the percentage of single-engagement visits to your site.
That means what Google Analytics is actually tracking is the number of visitors who come to your page and leave without viewing any other page on your website or engaging with your page in any meaningful way (more on that in a previous chapter).
Marketers interpret this measurement to determine whether the webpage provided what the user was looking for.
Bounce rate is not a measurement of how long a user spends on your page. Much of the confusion arises from this distinction. You can have a great, engaging page AND a high bounce rate because bounce rate does not measure time spent on site.
This has two implications:
First, the bounce rate is not necessarily bad.
While bounce rate might be explained by ineffective content and/or low accessibility, it can also result from a mismatch between keywords and content or even the page’s purpose.
For example, a high bounce rate on a landing page or product launch page is basically inevitable (especially with the trend in one-page websites). However, you might also desire a high bounce rate for informational pages where users can find what they’re looking for and then move on. A good example of this is Wikipedia.
Therefore, optimizing for bounce rate does not necessarily mean you are improving the quality of your website or helping your website become more useful for your visitors.
Second, overemphasis on bounce rate could decrease the usability of your site.
Imagine turning every page you have into two pages and linking them together to decrease your site’s bounce rate forcibly. From the analytics end of things, you have improved a KPI. From the user experience end of things, you’ve turned a simple, accurate site into a mess. Your website’s roadmap and user funnel should be purposeful in design rather than bounce rate focused.
Why doesn’t Google use bounce rate as a metric?
Based on bounce rate measures, it is easy to see why Google doesn’t use the bounce rate you see on Google Analytics as a ranking factor on SERP. We can break it down into four main reasons.
1. Bounce rate is not a reliable measurement of quality:
As discussed before, bounce rate cannot accurately determine user engagement because it does not factor in time spent on-page. At the same time, bounce rate varies widely across industries. For example, according to Clicktale, blogs tend to have 70-90% bounce rates, content sites about 40-60%, and service sites only 10-30%.
It does not make sense for Google to punish certain websites based on their Google Analytics’ bounce rate if the page’s purpose is not to lead the user through the rest of the website.
One comment left by DatingSolutions on a creative IM post is indicative of this problem.
Their high bounce rate is explained by the display of many of their images in Google search results. When people search for their images and click on them, the image is displayed and counts as a hit for the website and a bounce when people close the image after viewing it.
Deciphering between the two becomes impossible if the same keywords are procced for the website and the image.
2. Google lacks the motivation to access Analytics data:
While we see Google Analytics data as integral to testing or analysis, Google isn’t likely to see things differently.
People who believe Google has reason to stalk Google Analytics data cite one of two reasons, each of which can be deconstructed (though do note, a certain degree of conjecture has been employed on both sides since Google’s actual algorithm is proprietary).
Argument #1: Google gives preferential treatment to Google product users.
The thesis is simple. Google should (in theory) want to reward users supporting its business, whether it be through Google AdSense, AdWords, or Analytics.
The problem with this theory is that it is complete conjecture.
More importantly, Google could easily run into a legal issue if it were to implement this system. Preferentially treating users of its software and incentivizing even more use of its search functions creates a positive feedback loop that may bring up monopoly questions.
Google would be prudent to avoid vague waters.
Argument #2: Google Analytics reveals valuable information to Google.
If Google Analytics gave Google extra information about your website, concerns about bounce rate factoring into SERP make sense. However, Google is already harvesting all this information on its own. It did start, after all, as a WebCrawler. As stated plainly, Google already has access to everything it needs – who do you think provides all the data in Google Analytics anyways?
More importantly, though albeit indirectly, Matt Cutts has stated Webspam’s avoidance of Google Analytics data. To quote: “You can use Google Analytics, you cannot use Google Analytics, it won’t affect your ranking within Google search results.”
Therefore, what you see on Google Analytics is likely of no interest to Google search algorithms.
3. Google Analytics can be easily manipulated:
The number of posts published on how to detect and filter out bot behaviour in Google Analytics should be a powerful indicator on its own of Google Analytics’ inherently unreliable nature. If Google Analytics isn’t powerful enough to automatically filter out manipulative behaviour, there is no reason for Google to employ this “flawed” data in its ranking system.
4. Many websites do NOT use Google Analytics:
Because 94.6% of marketing websites use Google Analytics, it isn’t easy to conceptualize any website not using Google Analytics. But the fact is, by W3Techs’ estimates, only about 54.3% of all websites use Google Analytics.
Some sites don’t use any analytics tools; of the sites that use analytics tools, not all use Google Analytics. Sites also use Clicky, New Relic, Amplitude, Heap, Yandex Metrica, and WordPress Jetpack. Therefore, Google Analytics’ bounce rate cannot be used as a ranking factor.
However, we’ve been discussing bounce rate in terms of the percentage we see on Google Analytics up to this point. It should be intuitive to anyone that Google does not need to rely on Google Analytics to source data about the traffic and content of websites. Instead, Google employs its own tools to determine your actual website bounce rate: the pogo-sticking algorithm.
The pogo-sticking algorithm is Google’s tool for tracking user pogo-sticking. When a user clicks a link, realizes it is not what they’re looking for and returns to the search results page to click on another link, they are pogo-sticking – bouncing around from page to results to page to results and so on.
Hold up, pogo-sticking sounds ridiculously similar to bounce rate. Indeed, their similarity is often a point of confusion. However, whereas a high bounce rate means:
- The webpage is attracting the right audience and satisfying their needs on the landing page; or
- The webpage is attracting the wrong audience and causing them to leave unsatisfied
Pogo-sticking is only caused by the latter. This means whereas high bounce rates can sometimes be good, pogo-sticking is always considered bad.
According to Google’s “search pogo-sticking benchmarks” patent, Google tracks the number of times search results are selected before a user settles on a particular result as well as how many other results are clicked after a particular result.
If users click on many other links after visiting your webpage AND you have a high bounce rate, Google’s algorithm concludes you are causing users to pogo-stick. Your ranking is negatively affected as you leave your users unsatisfied due to ranking for irrelevant keyboards, using confusing webpage design, or providing inadequate content.
Google does not quantify the amount of pogo-sticking your webpage causes through bounce rate, but rather through the number of long clicks and short clicks your website generates.
A long click occurs when a user clicks on a result and stays on that page for a long time. They don’t return to the results page. On the other hand, a short click occurs when a user clicks on a link and quickly returns to the results page. Long clicks are the best sign of user happiness and what Google optimizes for.
For SEO, the implications are clear; ranking requires generating long clicks and keeping your long to short click ratio high rather than dwelling on bounce rate.
However, knowing why Google doesn’t use Google Analytics’ bounce rate brings us to a rather surprising conclusion: bounce rate still matters.
How does bounce rate relate to other SEO factors?
This question is the one most that SEO articles won’t talk about, but it’s probably the most important one.
At the end of the day, what’s important isn’t whether Google is tracking bounce rate through Google Analytics and changing your webpage ranking based on it. After all, direct implications are easy to track. Instead, what matters is how bounce rate indirectly affects the SEO factors Google does care about.
To start, let’s return to the pogo-sticking discussion. By this point, it should be clear pogo-sticking is something we cannot directly track through analytics tools. Google’s black-box algorithm isn’t exactly easy to analyze. However, we can work around this knowledge hole by combining two factors we do know: bounce rate and dwell time.
While you cannot analyze search intent with these two variables, you can generate a pretty good estimate of how much of your organic traffic classifies as long clicks.
, if your webpage has a high bounce rate, but users on your page also have a high dwell time, your webpage likely has a good long click percentage even though your bounce rate would typically be considered damaging. Thus, by taking dwell time into account, you can avoid unnecessarily optimizing for bounce rate.
Second, a high bounce rate is often a symptom of weakness in other SEO factors. Here are just a few SEO problems you might want to check for when your bounce rate is abnormally high:
- Slow loading speed
- Low-quality webpage design
- A mismatch between content and keywords
- Poor mobile optimization
These problems are treatable and often addressed when SEO experts “fix” bounce rates.
The problem occurs when the focus is on bounce rate (the symptom) rather than the underlying issue. Therefore, it makes sense why “fixing” your webpage bounce rate often helps your SERP ranking but sometimes affects nothing.
Conclusion: So what?
The key takeaway is this: while bounce rate doesn’t directly affect your page ranking, its bounce rate is still something you should understand and be able to improve upon.
High bounce rates (when calculated correctly) are often symptoms of deeper problems like user experience issues or poor targeting. So these are the things you should worry about. However, if you work on deeper problems like usability and customer targeting, SEO problems also improve.