In the settings menu, acceptable ranges are set for each ad attribute measurement. Values are assigned to the fair, poor and critical categories. 

When an ad is scanned, the data is captured, analyzed and assigned a rating based on the thresholds in the settings menu.  Scores are indicated by a dot in the Ad Grid. In this example, the CPU and size were flagged as “poor” (red).  Network requests were marked as “fair” (yellow).

The overall ad score is based on the lowest individual attribute rating for a given ad. In the example above, because one or more attributes are labeled as “poor”, the overall score falls in the “poor” category, indicated by the red circle.

Each ad is assigned a numeric value based on a combination of the overall ad rating and the severity of the issues detected in the ad.

For each G/F/P/C color rating, there is a max total point score. Poor ads for example, will never have a score above 55.  The lower the score, the greater the cumulative effect an ad has to the overall page load and user experience.

To calculate the numeric value, the total available points for a given G/F/P/C rating are split evenly between the ad dimensions that we track.  An ad earn up to “x” points based on scores within each dimension. 

An ad earns points based on how close its data is to the attribute “average”.  The “average” is a proprietary threshold that allows Ad Lightning increase the penalty for ads that egregiously violate a given threshold.  For example, while both 1MB and 5MB ads are out of spec, the 5MB ad score will be lower than the 1MB ad score because it is more impactful to the overall performance of an ad.

You can see in the example below that even though both ads are “poor”, the first ad has a lower score because CPU, network requests, and trackers are slightly higher.

Currently, core dimensions like size, network requests, trackers, CPU and waterfall contribute equally to the overall score.  Special calculations are assigned to malware (automatic 0) and asset detection like flash, audio etc.