Competitor_Sales_Control/Sentiment_Score_Control Inputs + Organic Volume

[0degreelongitude]

Hi all! Our team has encountered a few obstacles while setting up the model:

• Could someone provide insight into recommended sources/inputs for competitor_sales_control and sentiment_score_control in the CSV data? I am uncertain about where to acquire the data needed for these fields, but it seems that at least one of them is necessary to execute the model.
• Are individuals including non-paid search volume (e.g., SEO) impressions in their organic volume, or is this factor considered in other aspects of the model?

Any insight would be much appreciated - thank you!

[Jon77Ruler]

I'm not sure you have to use one of those specifically, but I agree I get an error unless I include at least one control.
To that extent, I just create a column "t" (trend) in my data and populate it with sequential numbers. Then throw that in as the control variable. Or something marginally more sophisticated from a time series decomp.

SEO impressions; curious to know also. I've shied away as if there's a high Brand-term mix in there then that can significantly over-claim credit. Probably needs sub-modelling, if it's worth it? My guess anyway!

[cpulavarthi]

Hello @0degreelongitude,

Thank you for contacting us!

I would like to inform you that the controls argument while loading data is an optional argument (Ref - API Documentation). If you have no controls to be included for your use case, then don't define the argument. Passing an empty list will return errors because the data loader is trying to map controls data that isn't there to the InputData object. Obtaining the data for competitor sales and sentiment scores will be business specific. Using dummy values for the same isn't necessary because the controls argument is optional in the first place.

Non-paid search volume can either be considered as a control or an organic channel depending on your business use case. The correct choice for its assignment will be dependent on the causal graph (i.e., an expected model of the cause and effect relationship between the non-paid search volume and your target KPI). You may use it as an organic channel if it is intervenable. You may consider it a confounding variable and add it as a control if it causally impacts both your treatment variables (I.e., media execution, etc) and your target KPI. Check our documentation on Control Variables for more information to guide your decision.

Feel free to reach out if you have any further queries!

Google Meridian Support Team