Replies: 1 comment 2 replies
-
|
Hello @vsamaga , Thank you for contacting us! Meridian provides the baseline contribution as an aggregate figure. We don't offer a granular breakdown of the baseline by individual control variables in the standard output because it isn't possible to identify the causal effects for control variables. The baseline should only be analyzed as a whole, and you shouldn't assign specific numerical attributions to control variable components. For more information, you can check our documentation on Non-Media Variables and Reasons Controls don't have Causal Inference or Baseline Breakdown. If you need to, you can try reclassifying a control variable as a non-media treatment variable. This should only be done if you can assume the variable belongs to the treatment node in the causal graph, and it's only practically useful if the control is something you can intervene on. When reclassifying, you can use the Feel free to reach out for any further queries or feedback regarding Meridian. Google Meridian Support Team |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for your response. Can you provide recommendations on if you would consider distribution as a non-media variable, as the retailer would have the ability to intervene on inventory? Variables like competitor sales, seasonality, etc. would likely categorize as a control, I suppose. |
Beta Was this translation helpful? Give feedback.
-
|
Hello @vsamaga , Thank you for contacting us again! The classification of distribution as a non-media treatment variable or a control variable is based on its intervenability. Control variables, on the other hand, are external factors that influence the outcome but are not intervenable. They are included in the analysis to avoid inaccurate estimation of marketing impacts. We recommend including all the confounding variables and strong predictor variables as controls. Please find detailed discussion on Control Variables in our documentation. You're right to categorize competitor sales and seasonal changes as controls as they are non-intervenable (beyond your control). See our documentation for considerations while deciding if a variable is a non-media treatment or a control. Feel free to reach out for any further queries or feedback regarding Meridian. Google Meridian Support Team |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Hello, is there a way to show the contribution of each control variable in any of the channel contribution charts? In its current config, it seems as though the effects of the controls are being baked into the baseline contribution.
Beta Was this translation helpful? Give feedback.
All reactions