Is it possible to estimate KPI under new control variable values (e.g., inflation) in the near future?

[nam13092003]

hi, are there ways to calculate sales (KPI) in the near future when I change the value of the control variables (or non media variables). For example in the past the inflation rate was 7% and I want to calculate sales (KPI) with inflation rate of 10%. Is it possible to get the trained parameters to do this?

[gousesb]

Hello @nam13092003 ,

Thank you for contacting us!

Meridian utilizes Bayesian Causal Inference to estimate causal marketing effects, rather than serving as a traditional forecasting model for future predictions. However, it allows for the estimation of ROI under various hypothetical media scenarios, such as adjusting advertising spend on specific channels or reallocating budgets or even changing values of control variables as is your case.

The new_data argument within the analyzer class method facilitates the incorporation of new data to generate outputs such as expected or incremental outcomes. It also enables "what-if" scenario analysis post-model training. This functionality allows for the simulation of outcomes based on hypothetical changes to input variables, for instance, a different inflation rate, without necessitating a complete model retraining. It's crucial to note that the dimensions of these new tensors must align with those of the original tensors in input_data.

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

Google Meridian Support Team