Setting Priors
Overview
Rockerbox uses a Bayesian methodology in our marketing mix models. An important step in building models is setting priors.
Setting priors in Bayesian models allows for the incorporation of prior knowledge or beliefs about parameters before modeling. Carefully chosen priors enhance model robustness and improves estimation accuracy.For MMM, we recommend providing priors in the form of ROIs that match your best understanding of the ROI for a given channel or subchannel. This may be informed by:
- In-platform performance metrics, such as Facebook's reported ROAS
- ROIs from alternative measurement tools, including MTA
- Incrementality or lift studies
Settings Priors
If you have channel or subchannel-level priors, provide them to your Rockerbox CSM to include in modeling. If you don't have any priors, or for any channel where you do not provide a prior, then Rockerbox will use defaults based on the closest match to the channel or subchannel.
Prior Templates
Use this template to quickly document the set of priors for each of your channel or subchannels. Make a copy, replace the sample values with your own, and deliver it to your CSM.
Example
Channel | Subchannel | Prior |
Prospecting | 2.5 | |
Retargeting | 4.2 | |
The Trade Desk | 1.8 | |
3.1 | ||
AdWords | NonBrand | 2.2 |
AdWords | Performance Max | 3.5 |