Question about geo-level budget allocation optimization

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()](https://developers.google.com/meridian/reference/api/meridian/analysis/analyzer/Analyzer#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!

Start with the Analyzer.response_curves() entry point referenced in the issue and review how geo-level response curves are exposed. Determine whether the repository documents or implements geo-level budget allocation, and define what a supported approach or methodological guidance would need to cover before implementation.

Written by the indexing model from the issue text.