Hi there,
Our company currently is utilizing segmentation to build out marketing channel based on tracking code. We currently are unable to use the marketing channel processing rules due to issues in our tracking code implementation by our media partners. However, we still would like to utilize Analytics as we can to report on marketing channel utilizing the Tracking Code dimension.
My question is how can we ensure that segmentation is mutually exclusive? We have tried many different segmentation techniques, however, our sum of Visits and Revenue for these segments has a variance versus the total site performance (Visits +17%, Revenue -4%). Additionally, we have built out the logic to the best of our ability to ensure we have no overlaps between segments. Is there a way we can view hit level segments and compare for an overlap in Visits & Revenue?
We have tried setting up the segments as follows, and all still had a variance vs total site performance.
I appreciate the help!
Solved! Go to Solution.
Topics help categorize Community content and increase your ability to discover relevant content.
Views
Replies
Total Likes
I am trying to understand fully what discrepancies you are seeing, since I can't really see your data.
First, while you can use Tracking Code and a custom Visit Level eVar for analysis (I do, as another dimension to our data), I should point out that this will differ from Marketing Channels for a few reasons:
Now, I am interested in how you built your segments?
Keep in mind that one visit may have multiple tracking codes...
Example:
The same visit touched X, Y and Z
Now, if you created Visit Level Segments, you will have overlap between these three, since all three were in the same visit...
Example:
Orders | |
3 | |
Segment: X (Visit) | 1 |
Segment: Y (Visit) | 1 |
Segment: Z (Visit) | 1 |
Now, when you use segments in table, the totals are not de-duplicated, so if this is what you are seeing, yes, your orders and revenue will be over-counted.
It's a little strange, but Segments and "locked" dimensions (i.e. dimension values that are dragged over one at a time and are locked act like a Quick Segments) don't de-duplicate, but pulling over a Dimension that will auto adjust based on all values (standard drag of the dimension item, i.e. eVarX), will de-duplicate.
You really should be using a Hit level Segment, when paired with the Order, you will see the last touch value "Z" associated to the Order...
Orders | |
1 | |
Segment: X (Hit) | 0 |
Segment: Y (Hit) | 0 |
Segment: Z (Hit) | 1 |
The Hit shouldn't have multiple values, the last carried forward value (whether for the week, or for the visit) should be the only item that pairs with your Order/Revenue values. Keep in mind that both Tracking Code and your eVar have attribution... all hits will have the value.. you don't need to build your segments to Visit level attribution, you really want to see what specific value was tracked at the time of the Order... hence using "Hit".
You can still use order attribution, changing the model to "Participation" or another custom model, but the totals will likely go wonky again.. mostly because of using segments... if you were applying the model based on the dimension alone, the totals should de-duplicate...
We really need more controls on our totals, to allow us to select whether or not to apply de-duplication when dealing with segments as breakdowns
I am trying to understand fully what discrepancies you are seeing, since I can't really see your data.
First, while you can use Tracking Code and a custom Visit Level eVar for analysis (I do, as another dimension to our data), I should point out that this will differ from Marketing Channels for a few reasons:
Now, I am interested in how you built your segments?
Keep in mind that one visit may have multiple tracking codes...
Example:
The same visit touched X, Y and Z
Now, if you created Visit Level Segments, you will have overlap between these three, since all three were in the same visit...
Example:
Orders | |
3 | |
Segment: X (Visit) | 1 |
Segment: Y (Visit) | 1 |
Segment: Z (Visit) | 1 |
Now, when you use segments in table, the totals are not de-duplicated, so if this is what you are seeing, yes, your orders and revenue will be over-counted.
It's a little strange, but Segments and "locked" dimensions (i.e. dimension values that are dragged over one at a time and are locked act like a Quick Segments) don't de-duplicate, but pulling over a Dimension that will auto adjust based on all values (standard drag of the dimension item, i.e. eVarX), will de-duplicate.
You really should be using a Hit level Segment, when paired with the Order, you will see the last touch value "Z" associated to the Order...
Orders | |
1 | |
Segment: X (Hit) | 0 |
Segment: Y (Hit) | 0 |
Segment: Z (Hit) | 1 |
The Hit shouldn't have multiple values, the last carried forward value (whether for the week, or for the visit) should be the only item that pairs with your Order/Revenue values. Keep in mind that both Tracking Code and your eVar have attribution... all hits will have the value.. you don't need to build your segments to Visit level attribution, you really want to see what specific value was tracked at the time of the Order... hence using "Hit".
You can still use order attribution, changing the model to "Participation" or another custom model, but the totals will likely go wonky again.. mostly because of using segments... if you were applying the model based on the dimension alone, the totals should de-duplicate...
We really need more controls on our totals, to allow us to select whether or not to apply de-duplication when dealing with segments as breakdowns
Views
Likes
Replies