Chinalytics Longmarch Blog

When I first started to learn the ropes of web analytics, I turned to Avinash Kaushik’s blog (Occam’s Razor) and book (Web Analytic’s: One hour a day) for a great deal of insight and actionable advice. One thing that stuck with me early, was Avinash’s emphasis on Bounce Rate as “The sexiest metric ever“. With all the caveats of generalizing metrics across different websites, bounce rate analysis is still a great place to start, when you plan to optimize your website.

Many companies and analyst have followed Avinash’s lead and are now prioritizing the reporting of Bounce Rate metrics. Talking to many clients in China I noticed a common question on everyone’s lips, tough: “My bounce rate sucks. What can I do?“.

Over time I developed a standard approach to address this question. Take a look at my 5 step guide:

Step 1: Does your bounce rate really suck? (Benchmarking)

In order to understand if you need to take immediate action to improve your bounce rate (as opposed to focusing on other KPIs), it is critical to benchmark your site’s performance.

Since user behavior and web design varies greatly among cultures, it is critical to find relevant local benchmarks for your site, ideally in your industry. In the US, services like compete.com provide valueable data. In China we have to do without any reliable 3rd party benchmark (what a shame). Even Google Analytic’s Benchmark function is not relevant, since it compares sites by industy, but does not provide country specific numbers.

A rule of thumb based on my experience in China (and please leave your ideas in the comments segment):

1. For micro sites for branding campaigns with mainly banner traffic: 85% to 90%
2. For landing pages of search marketing campaigns 25% to 40%
3. For landing pages of targeted direct marketing campaigns (20% – 30%)

If your numbers are higher, your bounce rate really sucks and you do need to take immediate action.

There are 4 common drivers for bounce rate.

Lets take a look at each of them.

Step 2: Landing page segmentation

Bounce Rate is calculated by dividing the number of single page visits  to a page (bounces) by the number of overall entires (visits that started on this page) to that same page. Bounces can only occur on landing pages (the first page a visitor sees on a visit to your site). So when your overall site shows a high bounce rate, you should first look at which landing page contributes most to your overall site bounce rate.

The most effective way to do that, is to calcualate the weighted bounce rate of all your landing pages. Stephane Hamel wrote the defining post about the methodology in 2007 on his Immeria blog. In effect you calculate the impact the bounce rate of each landing page has on the overall site bounce rate, by weighing it according to each pages importance (measured by the number of page views).

Use this formula

Bounce Rate * (Page Views/Total Page Views).

to calculate the Weighted Bounce rate of each landing page.

Take Action: Focus further analysis and optimization efforts on the landing pages with the highest weighted bounce rate. Check if your problem landing page is implementing best practices, usability test it, make changes, then A/B test the new version vs. the old version.

Step 3: Traffic Source Segmentation

Another driver for a high bounce rate on your site is low traffic quality. If your advertising efforts drive visitors to your site that are not interested in what your site has to offer, the best landing page cannot convert them. So before to start getting all excited about remodeling the landing experience, take a look at the traffic sources for your problem landing page. Many web analytics tools (regrettably not Omniture) allow you to easily segment your bounce rate by traffic source and / or type of traffic.

When doing this segmentation, look out for high volume traffic sources that drive traffic with a very high bounce rate. Very high is relative and a good benchmark is usually the bounce rate of your direct and search traffic. Visitors from these sources are usually highly targeted. If their bounce rate is high, your landing page likely has a problem. If these traffic sources have a low bounce rate whereas others, especially banner ads, partnership links etc have a very high bounce rate, don’t change your site, change your (paid) traffic sources.

Step 4: Creative Segmentation

When seeing high bounce rates for banners or SEM campaigns, it makes sense to dig one level deeper. Often these campaigns run with multiple creative executions of the banner or multiple copy executions for the text ad. Sometimes that creates a situation where one banner’s creative or call to action or one text ad is not relevant to offer made in the landing page. That is turn leads to a high bounce rate.

To understand if that happened to your campaign, you first need to make sure that your banners and text ads are comprehensively tagged (Google Analytics: UTM _content; Omniture SAINT tags) to differentiate between different creative versions. In the next step, A/B test your various creative version in multiple spots, to measure which one leads to the higher bounce rate.

Action: Run a creative A/B test before launching a campaign to ensure maximum performance.

Step 5: Loading time (Geo Segmentation)

Another very important factor for bounce rate performance is the loading time of your landing page. Especially rich landing experiences (often Flash based) require the download of large amount of data before they are ready for consumption. The longer visitors have to wait before the experience begins, the more likely they are to bounce. So far so easy.

The key challenge for web analysts is that loading time data is not available in any  web analytics tool. In order to get reliable data, you need to buy the services of companies like Gomez, who specialize in web performance measurement (see last weeks Web Analytics Wednesday). This data is especially important in China, where loading times can vary widely across provinces and cities due to a unique network layout (see ChinaNetCloud’s presentation on SlideShare).

A good indicator for loading time challenges is a large variation of bounce rates across provinces in China. In order to get Google Analytics to show you the bounce rate by province in China, go to the map overlay report and click on China. This will go directly to the “by city” breakdown. Then go to the URL bar of your browser and replace the term “city” with the term “region” (** here magic happens **).

Action: If you see a large variation (especially between northern and southern provinces) you have a good indicator that your need to improve your hosting infrastructure to address your bounce rate problems.

These are my five steps. What are yours? Did I miss anything important? Let me know in the comments.

Related articles by Zemanta

• The Context of “Intent” (fastforwardblog.com)
• Webtrends 9: Google Analytics Eat Your Heart Out (readwriteweb.com)

One of the most elelgant features in Google Analytics is Advanced Segments. It allows analysts to quickly and easily segment visitors based on their behavior on your site and answer questions like:

• What share  of overal conversions comes  from visitors that were searching for brand search term vs. non branded terms?
• Which pages do visitors read that do not convert but don’t bounce? (Maybe you missed a goal, or the opportunity to add a call to action to critcal page)
• What is the conversion rate for visitors from groups of untagged referrers (e.g. social networking sites, blogs) vs. tagged traffic sources and search
• What is the best indicator for the difference between “super engaged visits” (>10 PV/Visits) and “normal engaged visits” (>1 to 9 PV/Visits). Is it the traffic source? Is it the landing page? Is it the conversion for a specific goal?

Just create 1 or 2 segments and compare the bahavior difference in almost any report in GA, instantly. Anyone who every tried to answer similar questions in Omniture or WebTrends will feel nothing but gratitude to Google to make segmentation that easy. You can find much more detail in Avinash’s great Google Analytics Releases Advanced Segmentation: Now Be A Ninja! post.

What Avinash didn’t tell us in his post is how to supercharge our advanced segment using user defined variables. I learned about this power of this combination, when trying to segment out the behavior of registered visitors. GA has no build in function that can identify registered visitors, but Google Analytics Help had the solution. User Defined Variabeles!

Adding  a small piece of JavaScript to your login script

<script type=”text/javascript”>pageTracker._setVar(‘registered_user’);</script>

tags this visitor as a member of the “registe

red user” segment by setting a variable in the GA cookie that is handed over with each tracking image request. That sounds complicated but in the end it just means that GA now allows you to create a custom segment based on this User Defined Variable and voala, you can segm

ent each report for registered visitors. Sweet!

But why stop there? Advanced Segments allow an unlimited number of variations. For example

• Compare “Converting Registered Visitors” (who convert to a Goal) vs. “Non Converting Registered Visitors”, to improve conversion
• “Returning Registered Visitors” (more than 1 visit in the time period) to “Single Visit Registered Visitors”, to improve loyalty.
• etc..

To take matters further, Convurgency provided a list of ideas for User Defined Variables in their Google Analytics – User Defined visitor tracking post in Juy 2007. They include ideas like:

1. Visitor Type Segmentation (Business Users, Technical Users, etc) based on form inputs
2. Simple A/B testing by setting a user defined variable for each landing page
3. Referrer Segmentation

This was before the new GA code and before advanced segments became available. Respect! Now implementing all these ideas became even easier.

What are you waiting for? Go segment!

Any ideas for cool segments? Please leave them in the comments.

Related articles by Zemanta

• What does the Google Analytics API mean for you? (cmswatch.com)