Can I optimize budget by geo?

[alexanderhe88]

Hi Meridian experts, is it possible for me to optimize budget by geo?

[dattatreyam23]

Hi @alexanderhe88,

Thank you for reaching out to the Google Meridian support team.

I would like to inform you that Meridian currently doesn’t support optimizing budget by geo. For now the recommended approach would be to internally allocate the budget across geos, based on historical flighting patterns, as Meridian assumes historical flighting patterns in budget optimization. See example.

We appreciate your feedback and shall take it into consideration for future updates. You may track the changes to Meridian in the Change Log.

Feel free to reach out if you have any questions or suggestions regarding Meridian.

Thank you,

Google Meridian Support Team

[tgeria]

Hi Meridian expert,
follow the thread since the topic is the same. I read the feedback.

I’m working with a client who has media budgets allocated across multiple geographic areas. Meridian is presented as a platform that supports multi-geo analysis, which is one of the reasons we chose it.

However, based on my experience so far, even when we load geo-specific variables into the model, the output always appears to be cumulative — we don’t see the contribution broken down by individual country or region.

Is there currently a way to visualize the contribution by single country within the output? Or is the model designed to only return aggregated results across all geos?

Thanks in advance for your help!

Best

[dattatreyam23]

Hi @tgeria ,

Thank you for using Meridian!

Currently the plotting API (e.g., the contribution waterfall chart) does not support geo-specific contribution breakdown. However, Meridian is able to return geo-specific incremental outcome and total (expected) outcome by setting aggregate_geos = False. You can also select a subset of geos of interest via the optional selected_geos argument. Using the contribution's definition (incremental outcome as a percentage of total expected outcome), you can then compute geo-specific contribution.

Hope this helps!

Feel free to reach out if you have any questions or suggestions regarding Meridian.

Thank you,

Google Meridian Support Team

[tgeria]

Thank you for the feedback Meridian expert,

The point remains that if I enter both Italy and Germany into the data set, the model returns a total outcome instead results for each market. is it right?

Thanks you in advance.

[gousesb]

Hello @tgeria ,

Thank you for contacting us again!

No, as mentioned previously,

• Meridian is able to return geo-specific total (expected) outcome by setting aggregate_geos = False.
• Meridian is able to return geo-specific incremental outcome by setting aggregate_geos = False.

You could also retrieve geo-level summary metrics by setting aggregate_geos = False.

Hope these help.

Feel free to reach out for any further queries or feedback regarding Meridian.

Google Meridian Support Team

[tgeria]

Hi @gousesb
Thanks again for your feedback. Just to confirm, it is possible to obtain incremental outcomes per country, as well as metrics.
However, these are computed using the same underlying model (regardless of the country). Based on this, the budget optimization (function: optimizer.BudgetOptimizer(mmm) and optimize) follows the same logic across countries. In other words, given a total available budget of 100, and assuming other constraints are equal, the allocation among countries would follow the same mix.
Finally, is it possible to obtain the HTML results (functions: summarizer.Summarizer(mmm) and output_model_results_summary) specifically for each country?
KR

[gousesb]

Hello @tgeria ,

Yes, the incremental outcomes and metrics for each geo are still based on the same Meridian model, which is a geo-level hierarchical model.

For budget optimization, as pointed in our previous responses, it is also important to note that

• Meridian does not optimize budgets by geo.
• Meridian assumes historical flighting patterns in budget optimization. For example, see this documentation page for related explanation.

So, you must manually break down this allocation by geo guided by historical flighting patterns.

The HTML summary generated by summarizer.Summarizer(mmm).output_model_results_summary() is an aggregated report and does not provide a geo-specific breakdown. For geo-level results, you must retrieve the data using the Analyzer class and create your own visualizations.

Feel free to reach out for any further queries or feedback regarding Meridian.

Thank you,

Google Meridian Support Team

[twolf89]

Will this be an option in the coming future? It seems nonsensical to be able to build a by geo model and then not be able to optimize by geo, especially as that is what clients WANT to do.

[gousesb]

Hello @twolf89 ,

Thank you for contacting us!

Meridian allows for the selected_geos argument inside the budget optimization routine.

Users can run the following code:

budget_optimizer = optimizer.BudgetOptimizer(<your_meridian_object>)
optimization_results = budget_optimizer.optimize(
    selected_geos = <geo_of_interest>
)

This will calculate the optimized budget and response curve for the specific geo_of_interest. Users can then loop through each geo to obtain the geo-level breakdown of the optimized budget.

Feel free to reach out for any further queries or feedback regarding Meridian.

Google Meridian Support Team