You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi Meridian Team, I've been a Robyn MMM user for a long time now and have recently started testing Meridian. From my experiences with Robyn,I found that organic search data was often a channel that robyn struggled with to model. One of the solutions I found was to do use a nested model solution where we would train one model with organic search as the output and determine the true contributions of organic search using the intercept of that model.
From my understanding of the documentation, it looks like this is being handled using GQV since its meant to be a measure of 'pure' organic search. However would GQV also not be partly driven by media spent on awareness campaigns ?
The text was updated successfully, but these errors were encountered:
Thank you for reaching out to the Google Meridian support team.
It sounds like your approach is to estimate the GQV that would have occurred in the absence of advertising, and then to plug the point estimate in as a control variable in the final MMM. The first model is effectively estimating the causal effect of media on GQV, but might this estimate be biased if GQV also has a causal effect on media?
In a longitudinal data setting like MMM, it is possible to have a causal feedback loop where both GQV has a causal effect on media and media has a causal effect on GQV. Meridian was not designed to handle this type of confounding. Instead, our advice is to make a hypothesis about whether GQV acts more as a confounding variable or a mediator. GQV should be included as a control variable if you hypothesize that it is a confounding variable (or a proxy for an unmeasured confounder). GQV should not be included as a control if you hypothesize that it acts more as a mediator.
Some decision-making considerations are given in Including query volume as a control variable. The other information about Control Variables on this page may also be of interest.
Feel free to reach out if you have any questions or suggestions regarding Meridian.
Hi Meridian Team, I've been a Robyn MMM user for a long time now and have recently started testing Meridian. From my experiences with Robyn,I found that organic search data was often a channel that robyn struggled with to model. One of the solutions I found was to do use a nested model solution where we would train one model with organic search as the output and determine the true contributions of organic search using the intercept of that model.
From my understanding of the documentation, it looks like this is being handled using GQV since its meant to be a measure of 'pure' organic search. However would GQV also not be partly driven by media spent on awareness campaigns ?
The text was updated successfully, but these errors were encountered: