Hi, I am trying to run a Lookalike modeling using a first party segment as base and Acxiom 3rd party traits. When AAM return the results, I saw two things kind of weird:
1. The horizon of chart 'Model Reach&Accuracy' chart started from 84% instead of 0%;
2. The top selected traits' relative weight is just 0.5%. I wonder if that's due to using 3rd party data. But I want to see if anyone has suggestion?
Based on the current result, I am not sure if this Lookalike model is good to use. Any thought?
Thanks in advance!
Solved! Go to Solution.
1) How large is the first party base segment? If it’s on the smaller side, it could be causing issues. In addition, I would look at the overlap between this segment and the Acxiom 3rd party traits via an Overlap report to determine how good the overlap is for modeling purposes. If possible, you may want to consider trying a different 3rd party data provider for comparison.
2) Please see above. How large is the algorithmic trait that you can generate from the model? If it’s not very large, I’d look at some of my suggestions in (1).
1) How large is the first party base segment? If it’s on the smaller side, it could be causing issues. In addition, I would look at the overlap between this segment and the Acxiom 3rd party traits via an Overlap report to determine how good the overlap is for modeling purposes. If possible, you may want to consider trying a different 3rd party data provider for comparison.
2) Please see above. How large is the algorithmic trait that you can generate from the model? If it’s not very large, I’d look at some of my suggestions in (1).