Digithoughts

Analytics reports often contain calculated data such as 4 week or quarterly averages to provide bases for comparison. When calculating  values such as 4 week averages for calculated metrics and/or KPI (e.g. bounce rate, page views per visit), it is important to calculate the averages from the individually calculated KPI per the date range's granularity (i.e. day, week, month). Averages calculated from the averages of the individual numerator and denominator values may often give you different results. Why might this be important? Consider the examples above.

Let's say you were managing a site that recently launched a new campaign to drive more traffic to their website. Your campaign was wildly successful and during the last week of July, entries to the site increased dramatically; however most people tended to leave after landing on the home page. When analyzing the bounce rate during the first week of your campaign, it may be helpful to have an average bounce rate for comparison. 

In Example 1, the 4 week average Bounce Rate is calculated using the 4 week average for Single Access / 4 week average for Entries. While the final value is listed under the 4wk Avg. column, it doesn't actually represent the average Bounce Rate of the 4 individual weeks. This value calculated in H10 (69.3%) is more representative of the Monthly Bounce Rate, rather than the 4 week average Bounce Rate. The latter should be calculated by averaging the individual bounce rates per each week, as shown in the second example. The difference between the two calculations is nearly 20%, and with smaller samples and more variable data, the difference can be even greater.

While the first method is more susceptible to outliers, the second method tends to mask extreme values and buffers calculated averages against outliers. Both types of calculations can be helpful when analyzing calculated metrics, the important part is being clear and consistent about what you are calculating and how.

Up next Logan will be back to discuss Excel 2003 compatability.

Christine Wu Web Analyst

Read more from the Analytics category. If you would like to leave a comment, click here: Comment or stay up to date with this post via RSS, or you can Trackback from your site.