Impact metric in RUM smartboard

Overview

The Impact metric in the RUM Smartboard takes whatever performance metric is selected and normalizes it against the number of pageviews for each segment in a given dimension, providing a weighted average. The problem it solves is that if you just sort by average performance, that dimension may have a skewed average for a very small amount of traffic, so the overall effect of that segment on user experience might not be significant. By looking at Impact, you can more easily see which performance problems are significant.  

How Impact is calculated

Assume we have seven pageviews from the New York region and four pageviews from Los Angeles.

Let’s say the pageviews have the DNS times given below:

Let T be the total number of page views:

Let A be the average of the DNS times accross both the regions:

Now,
% Impact of New York  = Sum of DNS times from New York / (A * T)
                                       
                                        

Similarly,
% Impact of Los Angeles =
                                       

If we just consider the average of the DNS performance, we get
New York DNS(Avg) = 49.86ms
Los Angeles DNS(Avg) = 73.5ms

You can see that even though the average for LA is higher, NY has greater impact since it generated more traffic.