Posted on 27 November 2007 by Hurol No Comments

Web Data Segmentation – a pre-requisite for gaining actionable insights

We are observing increased sophistication in web analytics. This year, participants to our one day Web Analytics Workshops knew significantly more about web analytics. Many were conversant with at least one tool. We didn’t spend hours discussing the definitions of hits, visits and page views!

However, nearly all participants were very frustrated!

One of their frustrations stemmed from the struggle to gain actionable insights from the web analytics data. As we delved more into their web analytics practices, it emerged that what was missing from their analysis was data segmentation. They were trying to understand website users as a whole and finding it difficult to interpret. Even with the best interpretation skills, you will be stuck. Web analytics data has too much noise.

Take any reasonable size corporate website. Job seekers, media, students, friends and family all visit it alongside the customers, prospects and the business partners. When you mix all their activities together, even the best analyst will throw the towel.

Where to start

There are many opportunities to segment the web data. Any behavioural aspects of a site’s usage is a good place to start, such as where the visitor came from – referrers and geography, what content area they visit on the site, or what task they engaged in, to name a few. Drill down further into this data and combine it with other insights and soon you will acquire actionable and valuable business intelligence. Here are some examples to get you started.

Segment by method of arrival

In Bienalto’s web site analysis and evaluation projects, the first segment we always create is on how the visitors arrive on the website. We then study the behavioural differences should the user arrive directly by typing the website domain, referred from other websites or search engines. This study helps us to understand the website owner’s ability to bring users to the website as it:

• shows collectively the effectiveness of the marketing reach strategies
• illustrates the strength of the online brand
• shows us very quickly how good a job the business does with their search marketing.

For example, a large government department recently told us about major branding issues. They complained that many people still associated them with their old name, which was changed to a more positive, more customer engaging one many years ago. They reported a lack of brand awareness amongst their prospective customers. There were also rumours that their brand was associated with a negative image, that only desperate people resort to their services and many refer to it by its slang name.

Our initial segmentation proved a contrarian view. Upon analysis we found:

• There was significant brand awareness amongst the site users with 65 per cent entering the site directly.
• We estimated between 25% and 45% of the 6 million customers used the website regularly.
• There was a very high concentration of search keywords – 73 per cent of the searches were in top 100 and keywords contained the new brand elements with organisation’s proper name and / or product names dominating the searches.

Taking it one step further, segmentation also helps us to isolate unwanted / untargeted traffic from the rest. These are often the untargeted punters from a search engine. A careful look into the search engine referrals reveals to us in all our website health check projects considerable volume of untargeted visitors are brought by search engines.

For instance, on the website of a computer hardware company, we noticed that “telephone icon” appeared in the frequently used search terms. This company has nothing to do with telephone icons or telephones. They sell servers. A close examination revealed that a search engine liked the alt-tags used for the call me function on all of the product pages and indexed the site accordingly.

This shows that even organisations that are mature in search marketing accidentally attract many untargeted visitors to their websites. These people visit your website for completely wrong reasons. They come and go. They should not be there at all. These visitors should be isolated, the underlying cause rectified, and importantly removed from the analysis.

Segmentation by audience type

Many websites contain products, services, content that helps us segment the website usage by audience types.

For example, on a large global travel services company website, we segmented the usage by country of origin and observed significant marketing intelligence for this company. We found that the North American users, although they have dominated the website usage, had:

• less brand awareness about the company, its products and services
• relied on search engines heavily to find the website, significantly higher than the visitors from the UK, the continental Europe and Australia
• conversion rates in much lower proportions.
• lower participation rates in user generated content.

We added secondary and tertiary dimensions to the geographic segmentation to find other significant insights for marketing. The destinations of interest, for instance, differed significantly. Top three destinations for the North Americans were Mexico, Costa Rica, Brazil while the UK audience were interested in Rome, Thailand and Germany (the world cup effect!).

In another example, an insurance company client of ours offer special insurance to overseas students, various health cover options, dental and eye care services at their own centres. As each customer category wants something different from the website we drill down into these segments and gain further insights based on:

• frequency of visit
• tasks undertaken
• response to different promotions and search engine keywords.

As each of these segments represents a different commercial value and service priorities to our client it makes sense to segment website usage by these customer categories. If not, the results and intelligence will be skewed.

If your web site does not require registration or profile sign-in for access, you could infer segments based on the content viewed or tasks the visitor undertakes to accomplish their goal. For a simple example, if a visitor chose “view the latest XP updates” link, you may wish to identify them as XP users.

The above examples, I hope, illustrate clearly how data segmentation can assist in gaining valuable and actionable insights from web analytics.

What you need

To be able to segment the web data, you will need a flexible tool which allows creating segments:

• using various behavioural elements such as arrival method, pages viewed, geo-locations of the users, IP addresses, duration of visit, cookie values, etc
• at multiple level so that you can combine the behavioural elements to narrow a particular behaviour or incident
• retrospectively so that you can interrogate your data iteratively, on the fly based on clues, insights you observe without having to create tons of custom reports up front.

Consult your vendor for assistance on how to configure your tool to better capture segmentation information. There are tools out there that can provide what you need.

If you are serious about getting actionable significant insights from your web data, segment it. Otherwise Web Analytics will stay in the too hard basket. Contact us if you have any queries.