Expand my Community achievements bar.

The Analytics Community Lens news letter for February 2024 is now here! Come see upcoming events and what's been going on in the community!

Plot success events against the conversion date when using a date dimension and a non default attribution model (Attribution IQ)

Avatar

Level 2

9/13/23

By default, all dimensions are included when using Attribution IQ in Adobe Workspace. This includes the Date dimensions. However, I would like there to be an option to disable the modelling for the Date dimensions.

 

This is because the current behaviour when adding a Date dimension (e.g. day) to a metric with a First Touch model is that conversions are attributed to the first day within the lookback window that a touchpoint occurred.

 

For example, if my organisation uses a first touch attribution model, and I want to show the daily trend for conversions attributed to the Organic Search channel in the past week, then I cannot do this. Instead, they are plotted in the previous 37 days (depending on when the first touch for the channel occurred.

image (1).png

 

I believe a visualisation which plotted the success events on the day they occurred, whilst still using a First Touch model, would be more inutitive.

5 Comments

Avatar

Community Advisor

9/13/23

If your intention is to see the First Touch Marketing Channel against your conversions, why aren't you just using the "First Touch Channel" dimension against your conversion metric?

 

You don't really need the attribution model in this case.

 

This would get you the last week view and properly see the "first touch channel" instead of the "first touch of the last touch channel" as it is now.

 

 

Avatar

Level 1

9/13/23

As a potential work around, if you breakdown your "Natural Search" by date (instead of just using the line chart), then you can override the allocation model on the inner report to use "Same Touch".  As I'm trying it out, it seems like that only works if you override the allocation model on the outer level too.  So, that might not be an option for you since it may not match exactly what you currently have—unless you are already using report time processing for this (virtual) report suite.

Avatar

Level 2

9/21/23

Hi Jennifer

 

That's a fair point for the use case that I showed, although this doesn't exactly return the same thing because of how the marketing channel expiration window works. When I compare for a given week, it differs by about 5%.

 

Also, for different attrubution models (e.g. Linear) then I think you will still have the same problem that I posed.

 

Edward

Avatar

Community Advisor

9/21/23

@edwardg5146728 you are still trying to apply a "first touch" attribution to a dimension that is configured to have a last touch attribution. What you are doing doesn't make much sense... you are mixing attributions and getting a muddy mess... of course what I suggested won't match, I am trying to help you get actual first touch attribution.

Avatar

Level 2

9/25/23

Hi Jennifer

 

I think because we're speaking about workarounds that we've moved away from my original use case. We want to use Attribution IQ, which is supposed to progress things beyond there being "First Touch" and "Last Touch" channels

 

In fact, the business I am working with would like to use Attribution IQ to use a 30 Day Linear model to define standard attribution metrics to be used throughout the business. (I chose FT in my example because it I felt that it better demonstrated the issue with the current behaviour).

For them, they want these to be the business-wide metrics when comparing channels and measuring the return from a channel.

But with the current behaviour, it is hard to get consistent reporting when taking this approach. Chiefly because the date selector in a panel works differently to the date dimensions.

 

For example, if I have one panel which shows the contibutions from each channel last week, then Programmatic receives 888 conversions:

edwardg5146728_0-1695635284689.png

 

But if I create another panel to look at the  over the past 4 weeks with the same metric, Programmatic only has 430 (more than a 50% dropoff)

edwardg5146728_1-1695635455707.png

 

 

I feel that the actual problem is that in the current behaviour treats the Date dimensions as if credit should be attributed to them. Which causes these problems.

 

And which I don't think makes sense. I can't change my marketing strategy so that I got more traffic three weeks ago.