Hello @kenleeyx,

Thank you for contacting us with your queries and detailed background information. I shall answer all your questions in the same order.

1. The initial part of your understanding is correct. That is,

   When kpi_type is non_revenue and revenue_per_kpi is not provided, ROI is interpreted as incremental KPI units per monetary unit spent. In this case, the default value for roi_m and roi_rf will be ignored and a common ROI prior will be assigned to all channels to achieve a target mean and standard deviation on the total media contribution.

   Organic media and non-media treatments only support contribution priors and coefficient priors. The prior type used for each of these is controlled by organic_media_prior_type and non_media_treatment_prior_type parameters respectively which can be set while defining ModelSpec. The corresponding defaults for each parameter’s prior distribution is listed in our API documentation on PriorDistribution.

2. When the priors aren’t specified for the non-media treatments and/ or the organic media channels, the model will use default values for them. Both of these will default to using contribution priors and the corresponding default contribution prior for the organic media channels is Beta(1.0, 99.0) and for non-media treatments is TruncatedNormal(0.0, 0.1, -1.0, 1.0). This behaviour is detailed in the documentation links shared above.

   The code you provided only affects the paid media channels. It has no impact on the non-media and organic channels, which will then use the defaults we've described

3. Paid media priors are considered separately from the priors for non-media treatments. So they can’t have a single total contribution prior. This is consequent from the fact that you can have different prior types for paid media and non-media treatments as mentioned earlier.

Coming to your main question of actionable steps to experiment with priors and ensure ROI alignment with your expectations, I’d suggest starting with customizing ROI priors. You may refer to our documentation on How to choose treatment prior types and the pages that follow it under the “Custom Priors” tab. There is detailed guidance on customizing ROI priors, using past experiment data to guide prior setting, setting custom priors with non-revenue metrics, etc.

Do reach out if you have any further questions or suggestions regarding Meridian.

Google Meridian Support Team