In this issue:
Segmentation is essential in conducting analysis of web analytics data. Leveraging segmentation allows companies to answer complex web traffic questions that stem from basic reporting. While virtually every web analytics solution in the market today offers some form or method of segmentation, they all differ in their approach. Thus it is important to understand the concept of segmentation, your needs in regard to using segments, and the ability of various solutions to meet those needs.
What Is Segmentation
The concept of dividing traffic into smaller subsets, called segments, and evaluating metrics for those segments is a core process required in the web analytics industry. Segmentation can be thought of as dividing up the pie. Some common predefined items in today's tools which can be segmented include:
-
Pages Viewed
-
Files Downloaded
-
Marketing Campaigns Seen
-
IP Geo-location Areas
-
Visit/Visitor/Page Referrer
-
Page vs. Organic Search Referred
-
New vs. Returning Visitor
-
Visitor Frequency
-
Buyers vs. Non-Buyers
-
New vs. Repeat Buyers
-
Purchase Frequency
-
Screen Resolution
-
Browser Type
Advancing beyond overall traffic numbers and high level metrics to understand web traffic requires delving into such segments. It's helpful to know a site's overall conversion rate, but how about the conversion rate for new vs. return visitors? Or even more granularly, how about the conversion rate for visitors who visit 1, 2, 3, 4, … times? How do session times differ between buyers vs. non-buyers? What's the conversion rate for visits which view your FAQ page compared to the visits which don't? What are the values of the success metrics for test cell A vs. test cell B in a controlled experiment? All of these questions are examples of using segmentation.
One of the most common examples of segmentation seen in today's web analytics tools is the marketing campaign. To demonstrate value to online marketers, tools have incorporated various methods for dividing up visitors based on what marketing campaign they entered the site through. In most tools you can then compare various metrics (such as shopping cart conversion, average visit duration, etc..) across the marketing campaign segments. A typical segmentation report could be generalized to look something like this:
.jpg)
Perhaps the most complex idea behind segmentation is that it can span and answer an unlimited number of web analytics questions. Very quickly you will find yourself and others asking increasingly complex questions.
Common Benefits
Identifying Underperforming Segments
By comparing different segments of visitors, underperforming segments can be identified. Segmentation is commonly used in this manner for identifying poorly performing online ad campaigns, referrers, user groups, or pages.
Exploring User Behavior
It is very common to define visit/visitor segments which categorize users based on their overall behavior. For example, on an e-commerce site "Engaged Visitor" segment might be defined as: "All visitors which browsed more than 15 product pages but did not place an order." A customer support site might define differing segments of visitors based upon what information is being sought in the visit (store location, product support, tech support, product returns, etc.). Grouping users in this manner allows discovery of the correlation between groups of users and overall success metrics.
Email Remarketing
Companies are often looking for more ways to reach out and grab additional business from website users. Email remarketing does just this. Given that users have in some way supplied an email address, it becomes possible to remarket to users based upon actions they did or didn't make on the site. For example, a company can setup a reoccurring export of all known visitor email addresses for visitors who view their most profitable products. Then they can load this export into their email solution and very quickly send offers (coupons) to these visitors relating to the items they viewed. The results can be staggering. Additionally, some web analytics tools allow for the complete automation of this process.
Conducting Controlled Experimentation
In order to evaluate most experiments, there must be a method for segmenting visitors based upon which test cell(s) they belong to. With this in place, success metrics can be evaluated for each test cell segment. This has the ability to get fairly complex when exploring multiple variables in a single test, but there are solutions which can aid in managing these segments.
Common Problems
Robust Data
A very basic requirement across solutions for conducting segmentation is that the data needed to define the segments must be contained within the data being collected. No matter how much money you spend on an analytics tool, it is not going to be able to automatically divide your visitors into marketing campaigns if there isn't any way within the data being collected to identify which campaign they should belong to! Thus it is crucial to devise methods for placing users into segments, whether it is by using query string parameters to be parsed within a web analytics tool, or by adding web-based logic to assign cookie values which can be used for segmenting, or by any number of other methods.
For example, to segment users based on which online campaign they came to the site through it is common to add a query parameter identifying the campaign to the link in the campaign ad. Rather than linking to http://www.stratigent.com we might instead link to http://www.stratigent.com?cid=SegNewsletter. This "?cid=SegNewsletter" is what is known as a query string. Solutions' data collection methodologies, whether its page tags, server logs, or network data collection, will record this data for processing. Most robust solutions will then allow information to be displayed on all sessions/visitors which came in with this query string.
Robust Tools
Another common frustration with segmentation stems from the differing levels at which web analytics tools will allow segmentation. If the tool uses an open database for storing web data, segmentation might be done by writing custom SQL queries. In some more structured report based tools, segmentation might require defining all aspects of the report in advance or it would require the reprocessing of data to create the desired segments. Truly flexible tools will allow exploration of the data by defining and evaluating segments on the fly.
A common distinction made in many solutions is that between a "persistent" segment versus a "dynamic" segment. A persistent segment refers to its ability to be stored in the tool indefinitely for continuing analysis and reporting. This type of segment is typically made available to view as part of any reports that the tool produces. On the other hand "dynamic" segments are normally only applicable to the one report for which it was setup. Many solutions are limited in the amount and complexity of persistent segments that they can handle. Persistent segments require significant processing overhead which can translate to slow processing times for software based solutions or in the case of ASP's lead to an increase in fees charged to compensate for the additional hardware needed.
Conclusion
Segmentation allows for the flexible extraction of crucial web data for analysis. This topic is of critical importance to the user of web analytics data as segmentation is foundational to almost all advanced web analytics techniques. Additionally, this topic is critical when considering a new web analytics vendor as solutions have various methods for supporting segmentation, which vary greatly and should be understood in detail before committing to a product or vendor.
.jpg)
Josh Manion
Chief Executive Officer
Stratigent, LLC
.jpg)
.jpg){kind=small}
WebSight Newsletter.jpg){kind=small}
Newsletter Archives