ClickTale

It is also important to understand how visitors use your chosen test page, so you can be more confident that you are testing changes to important elements.

We use ClickTale for in-page analytics before and during testing.  It tracks mouse movement and clicks, and reports this information for a large sample of visitors.  This gives very valuable information about the important elements of a page, and the unimportant.

You can see how far the average visitor scrolls down the page, which is very useful to know if there is important information below the fold.

The biggest benefit of ClickTale is simply its ability to add some meaning to the results that you see.  Pages that fail can be analysed and you can learn so much about why this happened, which is invaluable.

User Testing

ClickTale shows us where visitors are moving their mouse cursor and clicking, but it doesn’t tell us why they act that way.  We can guess, but it is much better to ask real visitors.

We simulate real visits by asking paid participants to complete tasks on the website typical of an average user.  They are also asked to talk us through their rationale as to why they are using the site in the way that they are and to highlight any thoughts about the site in general.

Whilst we understand that this is never going to be 100% true to life, the sheer volume of ideas generated from the feedback is perfect for our needs.  The feedback simply acts as a guide for things to test, starting with the elements that caused the most friction.

You can also use a survey tool such as KissInsights to generate a small amount of very valuable feedback from real visitors.

Constructing Tests

With all the analysis concluded, you should be in a position to choose a page and a test to run on the page.  It is always a good idea to create a hypothesis of your expected outcome for the test:

A more visible call to action will simplify the buying process and increase the likelihood of a visitor buying a product

You will be running variations of the page which aim to prove this hypothesis.  We ran a test similar to this, and used a range of sizes and colours in our variations to establish which size and colour led to the highest conversion rate (Large & Red incidentally).

Here are some key areas to focus on:

Friction  – is it easy to perform the actions necessary to complete the goal?

Anxiety  – are there trust elements (such as secure payment) that will remove fear of purchasing?

Focus      – is it clear what the goal is & how to achieve this (are your goals the same as your visitor)?

Incentive   – is there a reason why I should buy from this company?

Value      – what more can I expect?

Motive   – Why should I buy now?

Often you will need to have pages (or changes) designed and developed for the test, which requires the analyst to produce clear wireframes.  It is important to be very clear about your requirements as a misunderstanding will lead to a test page that does not prove your hypothesis.

Running Tests

There are a number of testing tools in the marketplace which allow you to monitor the difference in performance between different test pages.  It is important to choose a tool which allows you to track the important metrics for your objective.

Google Website Optimiser is great for simple tests involving a single objective such as conversion rate, whilst other tools like Omniture allow you to see revenue impacts and segment across different audiences.

It is good practice to clear your cookies and enter the test a few times to see if it works as expected.  You should then be prepared to stand back until a reasonable number of conversions have gone through the test.  The worst mistake is to make assumptions, or panic very early on based on only a few conversions.  Conversions will always fluctuate dramatically and only over time will you see the true picture of which page is performing best.

We often aim for 99% significance after at least 100 conversions per variation (so we can stop the test early), or 95% significance after 200 conversions per variation.  You should also wait a minimum of 2 weeks in case there are weekend effects on the website.

This is a very real, robust process regardless of how well your website already performs.  Amazon is the site it is today through this very process, which has seen it evolve slowly over years from a very basic ecommerce bookstore to a mega online retailer.  Google is always testing changes to all their products including the search page.  This alone should be enough to convince you it is worth investing in, but just in case, this document is an in-depth look at our process.

Before you start, you should answer these two questions:

• What goal do I want my visitor to complete?
• Which pages are involved in the process of a visitor completing this goal?

With the answers in mind, you should be able to name a few pages that are worth running tests on.  However, we don’t yet know what changes to test, or whether the test will be viable.

We use a range of tools to help us to choose the right pages and the right tests to run.

Google Analytics

This is great for seeing traffic to a page and goals completed.  You will first need to use the Content section and look at the Top Content.  This will show you the pages with highest overall traffic going through them.  The higher volume of traffic running through a page, the more impact any improvement will have on your objective.

This table shows that the Home Page and Category 1 Page have high volumes of traffic, so if they are on our path to conversion, then an improvement would have a large impact.  Also, the Home Page has a high exit rate compared with other pages, so it is slightly problematic (although this is often characteristic of a home page).

We have highlighted the Basket and Checkout as important even though they have few statistical problems, so they are on the test list.  Other potentially good test pages are those with high bounce or exit rates. If you choose a page with a high bounce rate, you should check the volume of traffic landing on the page is high enough for the stat to be significant – use the Top Landing Pages tab.

Next it is important to check how many goals are completed by visitors going through your chosen test page.  This will allow you to estimate how long your test will need to run for.  Ideally, you should be aiming for 200 goals completed per variation you create:

This table shows estimates of how many variations you should use, depending on the page you are testing and the length of the test you are running.  If the page has few goals completed, you will be forced into a longer test.

To work out goals per month, you need to use segmentation in Google Analytics.  Create a segment that matches exactly page=example_page.  Then go to the goals section to see how many goals were completed in a month.

If you literally have less than 100 goals per month on your highest converting page, then you can choose a different goal to optimise for. This is not ideal, but if you choose a goal high in the conversion path, you should improve conversions. For example, an ecommerce site may use the first checkout page as the goal, since more visitors will reach this point than the actual completed sale.  This is not perfect, but if you can drive more visitors to the first page of the checkout, then it is likely you will make more sales.

We’ve explored how you choose the right pages to test on in Part 1. In Part 2, we will explore how you understand your visitors so you can create the tests that are likely to produce great results.

Why pay for a website optimisation tool when you can use Google Website Optimizer for free?

However, a year has passed and I’ve gained experience with a range of tools, including Omniture Test & Target, which has proved its worth to me.

As a tool, Test & Target offers a range of options which are not available as standard in Google Website Optimizer.  The ability to choose a winning variation based on revenue per visitor is a crucial advantage of the tool.  More sales is not always better if you are driving lower average order values, so revenue per visitor is often a much better metric.

It is true that you can manipulate Google’s tool if you have the skill, but most people don’t.  Therein lies another advantage of Omniture – your payment gives you access to an account manager and technician who can help you setup the tests.  This has proven to be extremely useful as tests never seem to be as simple as I planned!

Segmenting traffic has always been an issue on Google Website Optimizer, but Test & Target allows you to setup segmentations and has a few important ones such as weekends/weekdays and browser type as standard.  For example, the results of a test in Firefox, compared with Internet Explorer can often be very different, which gives you a greater insight into the way your visitors behave.

A very cool feature is the option to show unique content to a visitor based on their previous actions.  I have yet to explore this in full, but it is a very powerful method.  If a visitor tells you they are interested in men’s clothes on a unisex site, then next time you can show them a site tailored to men.  This can dramatically improve the experience for the visitor, who is then much more likely to buy.

There are still some flaws with the interface that I would like to see improved.  Ironically, some of the more important features are a little hidden, begging the question “do they optimise their website?”.  There was a recent update which included a nice new dashboard, so perhaps the site is being reworked over the next year.

Two weeks ago, Google hosted their annual conference for Google Analytics Authorised Consultants (GAAC’s), to which Epiphany sent along 2 employees. This conference has been taking place for around 5 years now, and is typically a chance for leading authorities of Analytics to share ideas and practices. As well as dining all week at Google’s expense – attendees are
given exclusive previews of the latest developments to Google Analytics.

This year’s conference was packed full of influential speakers and thought leaders, who explored new tools for Website Optimisation, better ways to use Google Analytics and the future of search. Our two representatives returned with a better understanding of how Google intends to shape its future and also new, innovative ideas which will add value for our clients. This knowledge will ensure Epiphany stays ahead of the curve in 2010.

Our star pick of the new tools available is ‘AttentionWizard.com’ which Epiphany is currently testing. ‘Attention Wizard’, developed by Tim Ash [http://twitter.com/Tim_ash], brings eye-tracking technology to the masses at an affordable price. Using analysis of data from real eye-tracking tests, this is a tool which simulates the results of running a real eye-tracking test and will offer great value for Website Optimisation.

We cannot discuss these advances until they are officially released, but we are rapidly disseminating the information internally; and when the features and products are officially released, our clients will immediately see the benefits.

So here is the WORKING example of how to implement it.

Firstly, its key to note you cannot A/B split test across multiple domains, you have to use a multivariate test.

In this example we are testing the homepage (www.mysite.com) and we have 2 conversion pages; one on secure.mysite.com and another on secure.example.com. (If you only have one conversion page you can obviously only show the code relating to your situation).

Let’s start with the homepage:

At the very top of the page code u need to add the control script, which will look something like with:

<script type="text/javascript">// <![CDATA[
 _udn = ".mysite.com"; _uhash = "0"; _ulink = "1"; function utmx_section(){}function utmx(){} (function(){var k='1234567890',d=document,l=d.location,c=d.cookie;function f(n){ if(c){var i=c.indexOf(n+'=');if(i>-1){var j=c.indexOf(';',i);return c.substring(i+n.
length+1,j<0?c.length:j)}}}var x=f('__utmx'),xx=f('__utmxx'),h=l.hash;
d.write('<sc'+'ript src="'+ 'http'+(l.protocol=='https:'?'s://ssl':'://www')+'.google-analytics.com' +'/siteopt.js?v=1&utmxkey='+k+'&utmx='+(x?x:'')+'&utmxx='+(xx?xx:'')+'&utmxtime=' +new Date().valueOf()+(h?'&utmxhash='+escape(h.substr(1)):'')+ '" type="text/javascript" charset="utf-8"></sc'+'ript>')})();
// ]]></script>

To ensure our cookie can be passed and tracked across our multiple domains / sub-domains we need to add some additions to the control script (these are shown in bold above).

Immediately after the control script you need to add:

<!-- utmx section name="Test URL" -->
<script type="text/javascript">// <![CDATA[
 var b = utmx('variation_content', 'Test URL'); function filter(v) { var u = v[0].contents; if (b && u.substr(0,7) == 'http://' && b.substr(0, 7) != 'http://') { u = u.substr(7); } return u; } utmx('url', 'Test URL', 0, filter);
// ]]></script>

This will be used to load in our test page.

At the bottom of the homepage we then include the tracking script, but again we need to make some amendments to allow the cookie to be tracked.

<script type="text/javascript">// <![CDATA[
if(typeof(_gat)!='object')document.write('<sc'+'ript src="http'+ (document.location.protocol=='https:'?'s://ssl':'://www')+ '.google-analytics.com/ga.js"></sc'+'ript>')
// ]]></script>
<script type="text/javascript">// <![CDATA[
 try { var pageTracker=_gat._getTracker("UA-1234567-1"); pageTracker._setDomainName(".mysite.com"); pageTracker._setAllowHash(false); pageTracker._setAllowLinker(true); pageTracker._trackPageview("/1234567890/test"); }catch(err){}
// ]]></script>

On our test page (page B), we need to add the same above tracking code (not control or page section script, just the above piece of code). This records which page (A or B) has been seen by a user and can therefore record a conversion against it.

Because we are using multiple domains we MUST add the _link function to our code. This is placed on all links between www.mysite.com and the domain with the conversion page (in this example secure.example.com) – you don’t need this code for sub-domains.

On www.mysite.com any links to secure.example.com would look similar to:

<a onclick="pageTracker._link(this.href);" href="https://secure.example.com/purchase.php">Click Here</a>

Without this conversions on secure.example.com will not record.
We also need to ensure that these cookies are picked up on secure.example.com. So you must ensure your Google Analytics Tracking Script (GATC) is present on secure.example.com. Again the additions are in bold, one thing to note is the changing of the parameters passed to setDomainName this needs to be mirrored in your own implementation.

<script type="text/javascript">// <![CDATA[
if(typeof(_gat)!='object')document.write('<sc'+'ript src="http'+ (document.location.protocol=='https:'?'s://ssl':'://www')+ '.google-analytics.com/ga.js"></sc'+'ript>')
// ]]></script>
<script type="text/javascript">// <![CDATA[
 try { var pageTracker=_gat._getTracker("UA-1234567-1"); pageTracker._setDomainName(".example.com"); pageTracker._setAllowHash(false); pageTracker._setAllowLinker(true); pageTracker._trackPageview(); }catch(err){}
// ]]></script>

Moving on to the conversion pages:

Depending on how many you have and which site they reside on not all the following will apply to you.

The conversion page on secure.example.com contains the following code:

<script type="text/javascript">// <![CDATA[
if(typeof(_gat)!='object')document.write('<sc'+'ript src="http'+ (document.location.protocol=='https:'?'s://ssl':'://www')+ '.google-analytics.com/ga.js"></sc'+'ript>')
// ]]></script>
<script type="text/javascript">// <![CDATA[
 try { var pageTracker=_gat._getTracker("UA-1234567-1"); pageTracker._setDomainName(".example.com"); pageTracker._setAllowHash(false); pageTracker._setAllowLinker(true); pageTracker._trackPageview("/1234567890/goal"); }catch(err){}
// ]]></script>

The additions are in bold and the setDomainName call is again domain specific.

The conversion page on secure.mysite.com contains the following code:

<script type="text/javascript">// <![CDATA[
if(typeof(_gat)!='object')document.write('<sc'+'ript src="http'+ (document.location.protocol=='https:'?'s://ssl':'://www')+ '.google-analytics.com/ga.js"></sc'+'ript>')
// ]]></script>
<script type="text/javascript">// <![CDATA[
 try { var pageTracker=_gat._getTracker("UA-1234567-1"); pageTracker._setDomainName(".mysite.com"); pageTracker._setAllowHash(false); pageTracker._setAllowLinker(true); pageTracker._trackPageview("/1234567890/goal"); }catch(err){}
// ]]>
</script>

The difference in the 2 codes being the setDomainName call.

Now it’s time to do the online / offline validation, once thats done you need to create your variation.

Click to add a new one, enter a name (Test URL) for example. In the text box, enter the URL for page B (your test page) – single line, no whitespace. Save and preview.

You’re ready to launch

Should anyone have any questions or need some help with their set-ups I’d be more than happy to help – you can contact me via the comments.

It is no longer enough to get hundreds or thousands of visitors a month to your website it’s the action your visitors take when they get to your website that’s important.

As an account manager working for a search marketing agency, my recent discovery of the new Website Optimiser tool has provided me with a great way of adding value to the service we offer our clients. Any business that relies on their website for sales and lead generation will find this tool remarkable and I am sure it will become an essential part of the marketing function. It is another one of Google’s innovative tools designed to help any business large or small and it’s absolutely free.

Google Website Optimiser has allowed me to take the SEO and PPC service we offer our clients to the next level. We already track conversion rates to measure the volume of traffic we drive to a site and how much of it converts into a lead or sale. Website Optimiser now means we can drive the conversion rate even more by constantly testing different pages to maximise on the number of conversions made.

I am not technical at all, but using web optimiser is really easy and can have a massive impact on your conversions. Quite simply you select which page you would like to test and then create the variations. You can decide on how often each variation is served so in case you are worried one of the changes will reduce your conversions you can opt for a 70%-30% split for instance, which can put your mind at ease. By continually monitoring and changing the pages you will eventually find the perfect page where you will hit your maximum conversion rate where everything on the page is motivating your visitors to take the action you want them to take.

It doesn’t get any better than this – what will Google think of next?