Google Meridian | Treatment Prior Types

— Before we configure priors, we need to categorize our inputs. In Meridian, treatments typically fall in two key buckets. First is paid media. These are marketing channels like TV, video, or display where you have direct spend data. The model usually links the outcome directly to that spend. Second, organic media and non-media treatments. These include organic marketing efforts or business decisions like price changes that can affect sales volume. But the key difference is these variables don't have an associated spend component. You have business metrics like ROI, but regression models often speak in coefficients. To help translate that business knowledge into the language that the model understands, Meridian's research team has developed three specific prior types while also offering coefficient priors. First is ROI. This is the default for paid media. Defined as the ratio of incremental outcome driven by a channel to the cost of that channel. One typically uses ROI priors to calibrate with lift studies. Next is MROI or marginal ROI. This looks at the return of one additional monetary unit. So if this is ROI, this would be MROI. It targets a slope at the margin. This is essential for budget optimization. Last is contribution. This is the proportion of the total outcome attributable to a channel. This is your go-to for organic and non-media when spend data is missing. So, how do you choose? If it's organic or non-media treatment, use the contribution prior type. Since you lack spend, this is the most intuitive quantity for setting an informed prior. However, for paid media, you often have more options to choose from. Let's use a simple decision tree to illustrate this logic for paid media. For paid media, first ask if you want conservative budget optimization in line with historical allocation. If you're in a scenario like this, MROI will work best. If not, ask which metric best aligns with your domain knowledge. To start, experiment results are often expressed in terms of ROI. In cases like these, choose ROI. On the other hand, if you have a better understanding of your business metrics in terms of percentages, use contribution. Most users often default to ROI priors. If ROI is what you care about, the next question to ask is the outcome revenue? If yes, use the ROI prior type. It typically aligns best with business knowledge. If no, say you have some a metric like app installs, and you don't have knowledge of your incremental KPI per channel, you'll want to use total paid media contribution prior. However, if the answer to that question is yes, you'll still use the ROI prior. The math handles the unit conversion. A final note to consider when making your selection, prior parity on one metric can create a lack of parity on another. For this reason, it's important to align your prior type with what your business primarily cares about. If you care most about ROI, set ROI priors. If MROI is what you care about, select that. Choosing the type that matches your goal will help ensure consistency. Thank you. —