Bridge modeling with business needs for combined small media channels

[victoriatascau]

Naturally there can be situations with media channels with lower spendings (and resulted ROIs) compared to other bigger media channels. In case such small media channels have already been obtained by summing even smaller ones (as advised in the documentation), what can be done in this case? Is it advised to try to further combine with other media channels, or can the resulted small ROIs be leveraged somehow? Is it possible to deduce information on the possible ROIs of the smaller components? For example, in case the budget allocator advises budget increase/decrease for the media channel made out of combined smaller media channels, what is the recommended way of proceeding for the smaller components? In an effort to bridge the modeling decision with the business needs, the latter being per media channel, no matter how small?

[dattatreyam23]

Hi @victoriatascau,

Thank you for contacting us!

Here is how to balance MMM statistical requirements with your granular business needs in Meridian:

• Combining further vs. using current ROIs: This depends on model confidence. Check your posterior credible intervals for the combined channel. If they are very wide, the model lacks a reliable signal, and further combination is advised. If narrow, the signal is reliable, and you can leverage the current setup. (See Data Requirements).

• Deducing ROI of smaller components: Strictly speaking, no. Once aggregated, the model evaluates combined variance, yielding a spend-weighted average ROI. Disaggregating this post-modeling requires arbitrary assumptions that the MMM cannot statistically validate.

• Executing Budget Allocator recommendations: When the Budget Allocator recommends adjusting a combined channel, bridge this macro-recommendation with micro-level execution using:

  1. Proportional Allocation: Distribute the budget proportionally based on historical sub-channel spend splits.
  2. Platform Metrics: Use bottom-up data (platform CPA, CPC, or MTA) to allocate the bulk of the new budget to the most efficient sub-channel.
  3. Experimentation: Use the MMM recommendation as a hypothesis and run geo-experiments on specific sub-channels to validate individual efficiency before scaling.

In short, let Meridian govern macro-budgeting (the combined bucket) and use your platform metrics to govern micro-execution.

Hope this helps. Feel free to reach out to us with any further queries regarding the same.

Thank you,

Google Meridian Support Team