Control and non media values

[mmiikkoo]

Hello, I went through the documentation but haven't found answer for these 3 questions:

1. How exactly should I create control or non media values? Example:
   The example input CSV contains GQV as control. That column contains positive or negative decimal values like "0.9442244". But what exactly it is? Is it some normalized value? Is it change over time? I found no explanation for that.

2. Let's say I want to include non media variable like weather or competitor sales. In what form should I use these numbers? In the example file I see promo as an example for non media variable but again it contains some decimal numbers but I am not sure what exactly is the definition of such variable.

3. Every website has some direct non paid traffic. To make the picture complete, this channel should be included as well. I obviously wanted to include it into organic media, but this answer suggests that it should be put under controls which is rather confusing for me. So I would like to ask if putting it under organic media is okay.

Thank you in advance and have a great day :)

[dattatreyam23]

Hi @yongqiangzhangzazzle,

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

To answer your queries:

1. Control and non media values aren't created, the data is obtained from your marketing initiatives. The primary purpose of control variables is to improve inference on the causal effect of treatments on the KPI. Also, you don't need to bother yourself with the scale of the sample dataset provided because scaling is done by Meridian and it is also reversed to give you results in the same scale you input in.
   If you are interested in knowing about the steps in selecting control variables or treatments properly, see here

2. Non-media variables, like weather, competitor sales, or promotions for that matter, should be represented numerically using a meaningful scale that reflects their impact. For example, price can be represented by its actual value, weather by temperature or rainfall amounts, and promotions by a binary input or a discount percentage (which might come in decimal values as witnessed). The key is to choose a scale that accurately captures the variable's effect and to apply it consistently throughout your data. Unlike paid or organic media, the counterfactual for non-media variables is not zero, but rather the minimum, maximum, or a user-defined value, as zero may not be a realistic scenario. You may also checkout Discussion #524 that talks about managing weather data in Meridian.

3. Organic media variables are media activities that have no direct cost. These can include, but are not limited to, impressions from newsletters, a blog post, social media activity, or email campaigns. Direct traffic is not an activity itself, nor is it directly intervenable. Rather, it is the result of brand awareness and other marketing efforts. Organic media variables behave like paid media variables with no associated cost, and direct traffic doesn’t fit that criteria. Hence it makes sense to include it rather as a control variable. See here for more info on organic media and non-media treatment variables.

Also, I would like to inform you that Discussion #139 that created the confusion for you, was written before organic media options were released for Meridian. Hence, you may proceed with the above mentioned suggestion.

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

Thank you,

Google Meridian Support Team