Web Analytics Tutorial

Lesson 9 – Incorporating Business Goals

IN THIS LESSON

Goals and Desirable Outcomes

E-commerce

Effective Search Terms

Further Study

Events and Traffic

Cluster Analysis

Further Study

Events and Traffic

So far we have discussed using Goals to track specific desired outcomes of a visit. Goals can also be useful for tracking quantities that may not be an outcome, but rather a side-effect of a visit. For example, if your site requires users to register to use a particular service, you can make the “registration succeeded” page a goal and use that to count the number of visits by new registrants. Using the Monthly Metrics report you can then see how many visitors you had each month and how many of those were new registrations.

Sometimes you will see a spike in traffic on your reports and wonder where it came from. If you notice a the spike in your daily traffic report, you might be able to correlate it with an advertising or other marketing campaign. As we covered in Lesson 2 - Where Are My Visitors Coming From?, you can then use the referrer report to find out where this spike originated and possible find the source. It could be a press release or a news story, in which case, by the time you notice it, it may no longer be visible.

If you know that a story is coming out or you are planning a marketing campaign of some sort, you can have any URLs that these contain connect to a particular page on your site. If you define a different arrival page for each campaign or press release, then you can make all of these Goals in your configuration. Using the Goals column, you can easily tell how many of your visitors came from one of your promotions. And the Goal report will tell you the relative popularity of each promotion.

Cluster Analysis

Target groups can sometimes be hard to identify. If you have designed your site for specific tasks, then usually these are a good choice. If your site seems to be all interconnected or you have several distinct goals for all visitors, you may want to consider using a cluster analysis technique to decide how to divide your visitors into meaningful groups. Cluster analysis is an advanced statistics tool for aggregating similar objects into distinct taxonomies. You may have already determined what the goals of your site are, but want to know what the customers’ goals are and whether they are distinct from the design. Using cluster analysis you can analyze your visitors based on patterns or values and perhaps discover distinct groupings for marketing or further analysis.

MORE ON

Sourcing Traffic Spikes

To illustrate the application of cluster analysis, consider a typical e-commerce site using cluster analysis to group search engines that are sending qualified visitors. In Lesson 3 - Search Engines we discussed qualifying search engines based on how many visitors they send who reach a defined Goal or based on the number of Average Steps visitors from each engine took. In this example, we want to look at both. Average Steps per visit indicates the value of the visitor for advertising – the more pages each visitor sees, the more effective the advertising. The Goal is defined as a desirable outcome, such as making a purchase at the online store. In both cases a higher the value indicates a more profitable lead from the engine.

Figure 9. Cluster Analysis can help group search
engines based on the traffic they send to your site.

By inputting Average Steps and Goals from the Search Engines report into a statistics program, the software can find distinct groups of engines. Figure 9 shows a sample distribution of search engines based on these two quantities. While the grouping seems obvious for most engines in this set of data, notice that WiseNut is hard to define. The clustering software, however, can tell, as the grey lines indicate, that it belongs in the lower left group.

By analyzing these engines on two quantities (and in most cluster applications, you would use more than two), we can interpret the value of each engine in ways that could not be seen before. The group at the top is good for generating advertising leads. Visitors from those engines seem to spend more time on the site and therefore see more ads. Pages on the site that carry advertisements, could be optimized and re-submitted to make them more prominent on these engines. The group on the lower right does not spend a lot of time on the site, but makes more purchases. Pages in the online store could then be optimized to make the store more efficient, if possible, and to make them list directly in these search engines to make purchases easier for these visitors. The group on the bottom left does not have much value in either instance. Perhaps listings in those engines should be removed.

Popular statistics packages, such as Statistica, SAS and SPSI, include clustering tools that you can use with the data from your web analytics reports. You may also be able to do your own analysis if your data is not too complicated or if it falls into obvious, discernable groups. Here are a few introductory resources on this topic:

Cluster Analysis: A textbook chapter from StatSoft Inc discussing Cluster Analysis algorithms and methods.
Cluster Analysis: What Is It?: A visual tutorial on Cluster Analysis from the Department of Biological Sciences at Manchester Metropolitan University.