Pulse reporting FAQ
This FAQ is designed to answer frequently asked questions regarding Pulse reporting.
How is view-through rate calculated in Custom Reporting?
View-through rate is a complex calculation based on quartile view-through rates in comparison to the amount of impressions.
List of the different quartile events that are fired:
- AD_START: The start of the video portion of an ad, meaning the video ad played at least between 0% and 25% of its length.
- AD_FIRST_QUARTILE: The user has reached the first quartile of the video portion of an ad, meaning the video ad played at least between 25% and 50% of its length.
- AD_MIDPOINT: The user has reached the midpoint of the video portion of an ad, meaning the video ad played at least between 50% and 75% of its length.
- AD_THIRD_QUARTILE: The user has reached the third quartile of the video portion of an ad, meaning the video ad played at least between 75% and 100% of its length.
- AD_COMPLETE: The user has reached the end of the video portion of an ad, meaning the video ad played to its completion.
Custom Reporting view-through rate is calculated as:
(AD_COMPLETE + 0.75 * (AD_THIRD_QUARTILE - AD_COMPLETE) + 0.5 * (AD_MIDPOINT -
AD_THIRD_QUARTILE) + 0.25 * (AD_FIRST_QUARTILE - AD_MIDPOINT)) / AD_START
- Number of AD_START = 1000
- Number of AD_FIRST_QUARTILE = 900
- Number of AD_MIDPOINT = 800
- Number of AD_THIRD_QUARTILE = 750
- Number of AD_COMPLETE = 700
(700 + 0.75 * (750-700) + 0.5 * (800-750) + 0.25 * (900-800))/1000 = 0.7875 =
Why are my unique impressions totals different than I expected?
When looking at reporting data for unique impressions, you might find situations where the total unique figure is different compared to the sum of unique impressions for the individual ads contributing to the total. How can this be?
- Ad A was seen by users Alice and Bob = 2 unique impressions
- Ad B was seen by users Alice, Bob, and Charlie = 3 unique impressions
Ad A + Ad B = 2 + 3 = 5
Total UNIQUE impressions
Ad A + Ad B = Alice + Bob + Charlie = 3
The same logic applies, for example, when looking at total unique impression for the duration of a campaign compared to the sum of unique impressions for each day of the campaign.
- Day 1: The ad was seen by users Alice and Bob = 2 unique impressions
- Day 2: The ad was seen by users Alice, Bob, and Charlie = 3 unique impressions
- Day 3: The ad was seen by users Alice, Bob, Charlie, and Dave = 4 unique impressions
- Day 4: The ad was seen by users Alice and Dave = 2 unique impressions
Day 1 + Day 2 + Day 3 + Day 4 = 2 + 3 + 4 + 2 = 11
Total UNIQUE impressions
Alice + Bob + Charlie + Dave = 4
Why is there a discrepancy in unique impression data between Custom Reporting and Performance Metrics Explorer?
Both Custom Reporting and Performance Metrics Explorer use the same HyperLogLog algorithm to provide unique impressions. The numbers provided are estimates of the real number of unique impressions, and there is always some error margin involved when producing the values.
Both reporting tools have the same error ranges, where the expected difference from distinct counts is only 2 to 3%, but they may still provide slightly different estimates due to the different implementation details and configuration parameters used in the two reporting tools.
For more information, see Using HLL++ to speed up count-distinct in massive datasets.
Why is there a discrepancy between Pulse and third-party tracking partner data?
- Are all tracking events fired correctly from the clients integration?
- Does the third party have the exact same definition of an “impression” as Pulse?
- Is the third party tracking all impressions fired by the integration?
- Is the third-party tracker correct in Pulse?
- Is cachebuster macro correctly used?