Hello!
As has been mentioned by previous users, I'm interested in geo-level budget allocation optimization, but this is not currently supported by Meridian.
I'm curious whether there is a specific reason why geo-level optimization wasn't included. For example, could geo-level response curves have higher variance, making optimization less reliable?
At the moment, I'm using response_curves() to obtain response curves for each geography and then running my own optimizer to determine the optimal budget allocation across geos.
Would you consider this a valid approach for geo-level optimization? Or are there any methodological concerns or pitfalls that you would recommend accounting for when optimizing directly from the geo-level response curves?
Thank you!
Hello!
As has been mentioned by previous users, I'm interested in geo-level budget allocation optimization, but this is not currently supported by Meridian.
I'm curious whether there is a specific reason why geo-level optimization wasn't included. For example, could geo-level response curves have higher variance, making optimization less reliable?
At the moment, I'm using response_curves() to obtain response curves for each geography and then running my own optimizer to determine the optimal budget allocation across geos.
Would you consider this a valid approach for geo-level optimization? Or are there any methodological concerns or pitfalls that you would recommend accounting for when optimizing directly from the geo-level response curves?
Thank you!