The charming Brett Crosby from google analytics, with some fanfare, announced several new google analytics features at emetrics last week in Washington – which you will have heard about by now, being ability to report on internal site search, ajax and event tracking, outbound link tracking and new Urchin software fixed price, all of which will be launched in the next few weeks. But here I have outlined a guide to how to use and getting the most out of Site Search.

Being able to track site search is a great step forward. Internal search refers to the keywords that people use while exploring your site (not the keywords they use on the search engines such as Google). For many websites, in particular holidays, recruitment, publishing and large retailers, internal search can be the most important and used feature on the website and can account for 50% of all pageviews on the site. So this is pretty amazing stuff that a “free” analytics solution is letting us do this – I’m really excited.

How do you use Google Analytics internal search:
1. Switch on ”Site search” on your Google Analytics profile(s)
2. You can find Site search in the Content section of Google Analytics reporting interface
3. Site Search reports show you the keywords that people use and the pages from which people begin and end their searches.
4. You can also filter search on your site against site usage, conversion rates, and e-commerce activity.

MY GUIDE TO GETTING THE MOST OF SITE SEARCH 
1. What do people search for and do any keywords stand out?
2. What searches results in failed searches and what proportion are failed searches?
3. Let’s do some segmenting and why it is so fundamentally important

1. Here is how to get your long tail of internal search keywords:
First, create a report of all the internal search keywords and unique visits for a 2 month (dependent on site) period and upload to excel. Grab all the keywords and visits into smoothed line chart with data points (so that you can easily see the keywords that stand out). Then try making a list of the keywords that stand out. These are words that need to be looked at carefully as they will be benefit from being presented on the site in an easy to find way so that your visitors do not always need to search for them.

2. Which searches result in failed searches
A failed search is when a visitor doesn’t find what they are looking for. For example, keyword searches on products or services that you do not offer would be a failed search as would a time sensitive product or service that is not available within the results of the keyword search made. For example, visitors that click on the back button after making a search would be classified as failed/frustrated searchers. To reduce failed searches make sure the site reflects at the minimum easy to find information on the more significant failed search keywords (this is just a quick fix and not the solution if only information is presented but is a needed first step until a good solution is found).
Then look at the percentage of all visitors that have a failed search. In addition, you can create a visitor segment where the search results page is also the exit page and compare this against all failed searches to see how many “failed search” visitors, leave the site immediately.

3. Then we segment, to confirm our suspicions and insights
Assuming we can label visitors segments with specific keyword searches, we begin to drill-down further. If on a recruitment agency site, a noticable search is for “web analyst”, we can see which were the most popular pages they visited before searching. From this, we could learn that they visited the “marketing jobs” page and the “Web jobs” page and as a result of not being to find what they were looking for, searched for the term “web analyst” and subsequently left the site. Therefore, it would appear that both of these jobs page would benefit from having information about web analyst jobs on them (until the recruitment site started posting web analyst jobs that is).

We can segment against new versus returning visitors, time spent on site, by navigation path etc. For example, we can see how these visitors came to the site in the first place by looking at search engine keywords. If there is a noticable percentage of visitors who arrived at the site after having searched for “analyst job” or “web analyst job” on search engines, then it is clear that the hopes and desires of visitors coming in from the search engines is not being met by the site – as well as a PPC (pay per click) overspend on keywords that are resulting in a high number of failed searches and exits from the site. The key is to reflect on the site what visitors are looking for, in a holistic and thorough way.

Google Analytics “Site Search” will not be offering the same level of functionality as the big paid for boys in analytics solution. But amazing that, internal site search will now be offered because it is such an important part of so many websites and not being able to track it with GA was a waste. So excited, but will wait and see once it performs when actually does launch.

Emetrics has come to a close after a few hectic days in DC, interspersed with seeing the Dalai Lama in George Town, the solar powered homes exhibition on the mall, the White House, hours chatting in the omni shoreham lobby bar and swimming in the invigorating heated outdoor pool at sunrise.

That asides, what has been going on? Or as my American counterparts would say, what are some of the key takeaways in terms of consumer understanding and behaviour. I’ll do another post about Google Analytics and Microsoft’s Gatineau this weekend.

Jim Novo of Drilling Down fame, spoke about speaking the “exec level” language that CEO/CFOs understand. If we think about our sales pipeline, it is the predictive/future likelihood to happen that execs are interested in when it comes to understanding our online data, sales and consumer behaviour so you can focus your efforts, marketing spend and optimisation efforts where they will have the most impact. Which are your dreck customers, your former best customer, new customers and best customers – map them out on a two dimensional value map with an XY slope.

Use recency, frequency and latency (you can even begin looking at these with Google Analytics) to understand your best and worst customers and grow your best customers. And importantly, build your predictive customer performance pipeline with your CEO/CFO so that they understand it, help you build it – which helps significantly with buy-in. Buy-in let’s face it, can be the biggest obstacle to taking action in any company.

Joseph Carrabis, the web analytics association new anthropologist and neuro-behaviourist on the scene, spoke about really taking advantage of our hard wiring to make our audience do and think what we want them to think or do. As human beings we all apply our own stereotypical and prejudiced frame of reference to everything into which we come into contact. For example in the context of images on a webpage, which image and at what position and angle will trigger what emotions or thoughts at a subconscious level. If an image is positioned at an angle, it implies motion. A photo of a couple, an elderly man, a teenager and early thirties woman will also, all provoke different inferences from one’s audience. To illustrate this, Carrabis engaged 50 of us in a persona exercise where we had to sit down after he namecalled six photos to tell him which one we thought was the Economics professor in Beijing. Interestingly, most of us thought the middle-aged conservative looking white man, was that character – and we were right. The key thing being the inferences that we draw.

In terms of multi-variate testing, the weather channel, used a variety of different images, a couple, then a man and also a woman to see which image was working more successfully in terms of optimising the page for it’s audience and hence having the highest conversion rate. Interestingly, the web page version withe the image of a woman on her own had a much higher conversion rate than other versions tested. This can be linked back to Carrabis point about the power of assocations, inferences and our pysche hard wiring on what we think about images, positioning and sound on a webpage.

Neil Mason, a fellow Londoner, talked about segmenting one consumer segments into tribes (richly developed personas in other words), using datamining to provide statistically robust anomalies, patterns, associations that stand out from a business commercial perspective and use these to identify key drivers for purchase and identify the most valuable consumer segment. For example, with a case study on the Royal Mail, segments included price finders (10%), cottage industrialists (2%) and regular posters (1%) – which were the most valuable segment. They also indentified that visitors who “saved a quote” on their first visit were signifantly more likely to become and continue to buy from the website and be the website’ most commercially valuable segment (worth most money). They used these consumer segments to drive email marketing segmentation and discovered that emails sent 4 to 5 days after their last visit were most likely to convert. Less than 4 days was too soon (the visitor was still thinking about it) and more than 5 days and the conversion rate began to drop. It’s all about the timing – oh – it’s recency again.

 ”Where are most valuable customers, how should I optimise my site?”

Thanks for an interesting emetrics everyone and I look forward to meeting those I met again soon

I am a huge fan of management and marketing theory (not for the sake of it of course), but applying and finding ways to make the continuous job of improving overall marketing performance (web analytics of course) a little easier. I developed my Activity Based Scorecard (ABS) after working with the balanced scorecard - traditional management theory. In my “web analytics scorecard”, see image below, define KPIs, I identify relationships, I bench mark these relationships (with trends) against themselves: and ask the question how effective is my website and marketing performance? This is my correlation between effectiveness of my website under review and the web analyst’s approach to continuous improvement, of said company performance.

              (Click image for higher resolution) 

My ABS scorecard, large image above, measures the relationship(s) between:
1. Usability, multi-variate A/B testing, market research.
2. Management information and web analyst reports.
3. Web data and statistics.
4. Actionable insights and decisions.

These can all be classified as flows of information and should have separate agenda and metrics. Each component should be weighted according to its importance and the overall metric for assessment. In other words providing weighted scores for different components to provide a more accurate picture of their value to the overall picture, from the company perspective which would be aligned to their commercial objectives.

At the outset one should understand the objective, content, and overall aim of the website, in other words if everything went kaput, as Avinash often tells us, what is the most important thing for your business. My Activity Based Scorecard, analysis and decisions then flow from this. My analytics scorecard is a very similar assessment to the cash flow forecast of a company (finance people will love this) it takes basic inflows of information (whether product or service based) and links these inflows to outflows. A relationship Forms. Also, a cross relationship forms between the numbers of statistics and the reports generated and not to forget the decisions that flow from these relationships.

Separate metrics or cross relationships develop and can be assessed alongside the overall assessment. Usability can in some cases prevail over say the numbers leading straight through to decision flows which influences both the statistics and the reports generated. Through timely assessment of my ABS report trends (and metrics) can be established through each component and the overall assessment.

Actions can have a corrective or test feedback straight through to the inputs or changes of the core components. But, for me the most important factor is that the activities of the analyst have direct influence to the drivers of the website. This ultimately impacts the assessment.

Strategic objectives can be best described as the common ground between what we focus on, what we do best and ultimately what our passion is. We are however constrained by our financial drivers. Our marketing plan directly links with our strategic objectives. So when senior management ask: how is our website performing? - we have a context and a weighted activity based scorecard within which we can measure our true performance and ongoing actions and decisions.

The reality is that for any of this to work at all requires alot of hard work, perseverance and at the outset defining the key business objective(s) for the company. There are some tools on the market to help develop one’s scorecard, for example with Balanced Scorecard Designer you can create a set of KPIS and group, categorise and weight them.

Do let me know if you have any questions or if you agree or completely disagree. Thanks so much for reading.

Consumer Generated Media (“CGM”) is the term that encompasses all social media content on the Internet authored by consumers. This content ranges from blogs, to social networks, consumer review sites, message boards, and videos.

Social networking and connecting with customers is all the buzz, for example yesterday Forrester Research did a webcast on “Know your Customers’ Social Technographics and Craft the Right Social Marketing Strategy” with Charlene Li from Forrester. She shared her insights on understanding ones target audience attitudes and behaviors towards social technologies in order to craft the right social marketing strategy. These are great calls for marketers to learn more about getting their arms around social media, listening to the voice of the customer and engagement with consumers in social media.

Jeremiah, a fellow web analytics association social media committee member, is a social media stategist and gives us outlines of how to approach positioning one’s company in the wider social media ecosystem. From a web analytics perspective, how does one even begin to gauge the influence of these conversations on one’s brand?

Social media technographics report by Forrester research. 

Some other stats, from Pew Internet and Jupiter research:

One blog is created every minute
27% (32M) read blogs
22% (27M) post reviews/ comments
44% (53m) are content creators(running own blogs/sites, posting messages).

There are more than 1.5 billion comments per day, the collective voice of the consumer to influence brands and buying strategy has never been stronger and will continue to be strong.

There have recently appeared in the market, applications, such as Visible Technologies TruCast that enable companies to monitor social media conversations, gain valuable insights, and even engage with consumers in order to better allow companies to manage their brands online on social media sites. For companies, these online conversations represent a new opportunity and challenge for brand monitoring, reputation management, word-of-mouth marketing, and consumer engagement.

This is pretty powerful stuff, the ability to segment one’s potential customers by feeling and tone and message from the enormous pool of social media sites.

I wonder how scalable this tool or any tool is, because eventually with the increase of the blogosphere appearing to be exponential how much data will their databases be able to handle?
But assuming all social media data on the Internet, posts and comments are collected in a database with multi-tiered querying – there would be some pretty powerful information.

Influence engagement metrics and advanced analytics:
Identifies the most influential consumers for a particular topic or issue
Determines the sub-topics of conversations
Interactive dashboard allows clients to determine specific sites and authors wielding the most influence in conversations.

What are they talking about (sentiment) scores:
For example, their intelligent sentiment technology evaluates the positive and negative sentiment and tone of conversations. Users establish sentiment criteria by scoring a sample of data, and TruCast automatically scores the rest. I’d like to put this to the test.

There are others such as Pythia which give trended social media data for free, so even for SMEs there are tools which can help.

I personally think the idea of engagement metrics within the context of the broader social media ecosystem and putting it to use to be able to positively impact on managing one’s company’s brand, social media reputation management, is something that we will all be doing in the not too distant future.

Any thoughts or questions or disagreements, please let the web analytics princess know.

Say for example, we would like to compare the performance of three different marketing campaigns against one another, by which I mean organic, ppc, online banner advertising, we would want to analyse the conversion rate of the campaigns over time.

First of all, we would take the average conversion rate on a daily basis, by average I mean the central tendency for our data to centre around a particular value, be that mode, median or mode. And then compare these average conversion rates over time for each individual campaign against one another.

However, when we are looking at how the data is dispersed, we may see some points that either stick out too high or too low. So how can we tell whether these are significant data points. If we look at all of the raw data that makes up the averages, it may be that there is one number that is so off the charts that it is completely skewing the whole average – these are called outliers.

It is important to get rid of outliers, otherwise you would have no idea whether or not one campaign is performing better than others.

I will post more about this later in terms of my key take aways overall. But for now, thanks so much to Rene and Aurelie for their interesting event Web Analytics day and their hospitality. Here are some take aways from the event.

Ian Thomas from Microsoft showed us some slides from Gatineau which will be launching in a couple of weeks. Click here if you would like to sign for Microsoft’s beta invite request.

The key USP of gatineau is the ability to segment your visitors by age and sex demographic from Live passport holders who visit your site (there are 30 million live passport holders so for your website this is likely to be a statistically significant segment). This demographic data is apparently completely anonymous. Plus some nice zooming in. As to whether this become a significant player in the market, like Google analytics, only time will tell. I do have some nice photos of the screen shots but will have to put those up later.

Actual web analytics client casestudies – I always think that actual results from actual clients are the meat on the bone, for example Eric Petersen’s presentation, Web Analytics is hard.

1. Top online retailer improved site design as a result of web analytics and multi-variate testing.
Analyst finds that visitors using site search are better customers.
Analyst recommend improving visibility of internal site search.
Controlled testing of different visibility and position of internal search within site
One successful experiment returns a 2% increase in visitors using site search.
Result: a six figure lift in weekly revenue, equal to a nearly 1% increaase in total company revenue over a year.

2. A well known brand analysed website to isolate a specific group of users on the web site.
Analyst shows that “product compare” functionality is under utilised.
Visitor segmentation is used to isolate visitors using “product compare” functionality.
Analyst discovers this segment has a 33% higher average order value.
Changes made bsaed on this analysis increase traffic to “compare product” functionality by 11%, increase purchases by 16% and reduce exits from these pages by 56%.
Result: an annual increase in revenue of $2.2 million

3. A Fortune 500 travel company used web analytics too and made millions by improving their messaging.
Analyst speculates that prospects were not seeing “best price guarantee” information.
Concerns were confounded by multiple messages and placements throughout site.
Multi-variate testing was used to test different messages and treatments.
The winning format was shown to convert at a rate of 0.6 points better than the control.
Result: an estimated life in online bookings of $30 million annually.

Sometimes in web analytics, we get far too concerned with which tool should we use, which theory or thought leader are we following. Ultimately, we are in the business of making more money for our clients or our own website, be that in increased sales or increase lead generation – which is why I gain so much from hearing exactly how and why people are getting the results they are for their websites.

Thanks for reading and please do comment if you have thoughts for web analytics princess

This is a different type of post from normal because there will be lots of graphics instead of words. I am also ill at the moment which is really annoying. The question being, which online marketing banner is more effective, from a web analyst’s perspective of course, and which other bits of data or charts would you have liked to have had?

Bubble charts key (I like bubble charts because you can compare 3 datasets at the same time):

Size of bubble determines number of visits in (000s) driven to website.
Bounce rate is the number of visitors who leave site after less than 10 seconds and only visit 1 page in the site.
Conversion rate is total number of conversions (eg items sold) divided by total unique visits to the site.
Message effectiveness chart where each different coloured bubble represents a completely distinct marketing message and design on logo.
CTR (click through rate) is the total number of clicks on the banners divided by total number of impressions on media/publications/other websites where banners were placed.
Cost represents the total ad spend in (000s) with the various publications.

As usual, I very much welcome your feedback and thoughts and please do let me know which online banner you think is more effective and why and which other charts or data would you would have used?

Last weekend I went to the inaugural Podcamp UK and co-presented a session with Lucie Follett on the monetisation of podcasting and podcasting measurement using engagement metrics, in the auspicious surrounding of Birmingham’s NTI (new technology institute). It was fast paced and innovative. You may be able to spot us in one of the photos? There were a whole bunch of people there including all the top UK podcasters, Twitter guys (I twitter, do you?), bloggers, journalists and new media folk in general.

In the social media ecosystem, in which I would include podcasting, there is so much potential for businesses to use podcasting to generate brand awareness and interest in their product or service from a niche audience. At the same time, there is an increased awareness of the potential monetisation of podcasting, if it is done effectively. I am still a big believer in “Content is King” - ie create podcasts that genuinely interest and compel your target audience. And have seen examples where “view movie” (ie watch podcast) with the right kind of engaging content has resulted in a tripling of lead generation on a particular car company’s website, such as brochure requests. So podcasting can and does work for business when done in the right way - you need a good story, and definitely not my boss told me to do a podcast!

However, how do you begin to measure a podcast’s effectiveness?

Due to the nature of downloadable media, there are a number of difficulties when it comes to getting accurate metrics from podcasting and issues to consider which impede the efficient implementation of big marketing or advertising campaigns across multiple website.

-How many podcast downloads are there – if the podcast is embedded in the website, is it still considered to e a podcast?
-How many viewers actually watch or listen to the podcast once it is downloaded?
-What degree of the podcast is listened to, for example if you have ad in the podcast towards the end, how many people actually listen to it?
-What true influence or buzz is actually generated by the podcast, because link content (popularity) does not equate to influence.

The key things is to look at podcasting in the same way one does other social media.

Engagement metrics are key. Things to consider include (and please feel free to add any more via comments):
1. Visitor reviews of your podcast (for example on itunes).
2. Visitor comments – where a podcast:comment ratio is the most helpful one as it strips it down to pure engagement on a podcast, by podcast basis.
3. Social capital/Visitor influence – if an established reviewer ie top podcaster or specialist within the industry writes a review/comment etc, this will have a lot more influence than if Sam from Dunkirk did (sorry Sam).
4. Ranking on established podcasting platforms, such as podcasting news top 25 or podcast alley top 10.
5. Wisdom from the rest of the web, such as the reaction on the blogosphere, twitter-sphere, facebook-sphere, general search engine results etc.

The monetisation of podcasting, is not about corporates trying to strangle the life out of a vibrant, independent podcasting community – which will definitely continue to thrive, but a marketing journey where businesses who understand social media will use it to their advantage. Businesses that podcast will be able to measure those tangible or intangible (hence engagement metrics) benefits to their business, and where eventually marketers and advertisers will be able to efficiently implement advertising across multiple podcasts, similar principle (but very different at the same time) to the way google adwords has their content network advertising – where you can run campaigns on a keyword/sector basis, having illustrated the value of advertising on podcasts or websites running podcasts.

Thanks so much for reading and do let me know if you have any thoughts or ideas, or if you completely disagree.
 

This is a continuation of my series of google analytics tips that I began a few weeks ago with naming your webpages. Now, I am looking at tracking visitors across different domains or subdomains for the same website. You will need to add some extra variables to your javascript google analytics tracking tag. Otherwise, google analytics won’t realise that they are all part of the same site and your visits will be inflated. Every time a visitor goes from one subdomain or to another domain, they will be treated as new visitor (see below for “what are cookies”).

For example, earlier this year when I began working with a global recruitment agency (no names, I’m British) where they had set up google analytics internally, I discovered that visits weren’t 120,000 a month as they thought – but instead closer to 50,000 visits a month. The reason being that there were 16 domains/subdomain so every time a visitor went from one country site to another (with different domains) google analytics was treating them as new visitors. So quite a big problem that needed to first resolved technically – in order to get the correct numbers – and then with tact (obviously). You also need to make sure you add these extra variables to every webpage/tag where visitors can navigate from one domain or subdomain to another.

For example, if you have a product website, a shopping cart on a subdomain and a company website,
1. www.product. com: product website
2. store.corporate. com: shopping cart
3. www.corporate com: company website

You will need to add three variables to your standard GA tracking tag so that when one visitor for example navigates from corporate.com to product.com to store.corporate.com they will be treated as one visitor and not three which is what would happen otherwise.

The three variables you have to add:
First, utmLinker. This variable should be set to “1″ which means on. The reason is that utmLinker is the actual mechanism that transfers the cookies from one domain to the other. Without utmLinker you’d have two sets of cookies with different data – google analytics would not identify the same visitor crossing domains without it. With utmLinker, two different sets of cookies have the same data so that google analytics knows that this is the same visitor.

Second, uhash. This variable is named _uhash and should be set to “off”. This variable adds a hashed (or encoded) version of the domain name to the tracking cookies. This value is used during processing by the GA system.

Third, udn. This variable is named udn and should be set to “none” for the product website and “company name” for the store.company.com and Company.com

This is what you would add to the google analytics tracking tag for each site: 

On product.com
 _udn=”none”;
 _uhash=”off”;
_ulink=1;

On Store.company. com:
 _udn=”Company. com”
 _uhash=”off”
 _ulink=1

On Company.com:
_udn=”Company. com”
_uhash=”off”
_ulink=1

N.B What is a cookie:
Cookies are used by Web servers to differentiate users and to maintain data related to the user during navigation, across multiple visits to a website and as a way of remembering what visitors do on a website for example buying, removing items online or logging in to a site. For example, they were originally developed as a virtual shopping cart/basket into which visitor could “place” items to purchase, and as they navigate add or remove items from their shopping basket at any time. 

To Conclude:
This may seem somewhat dry a subject, but as I highlighted in my introduction – it is vital to get it right. Anyway, we are the new breed of marketers aren’t afraid of a bit of a javascript and statistics:)

If you have any thoughts or questions, or happen to know of any other ways of doing things then do please let me know.

It appears that an usurper to my title of web analytics princess has appeared, in the shape of my daughter Leah. She has been making friends in high places in the web analytics world. And in fact, Leah has recently been caught on camara whilst researching/contemplating web analytics conundrums in Avinash’s book Web analytics an hour a day and is now on his blog. 

To quote Avinash “Beautiful English countryside. Check. Wonderful cottage in the background. Check. Pretty little Leah. Check. Obligatory Web Analytics book. Check!” 

She is my inspiration and ultimate motivation for everything that I do – so please forgive a little bias.

Please feel free to ask her any questions, she is particularly keen on colouring in standard deviations.