Google’s developer documentation for Meridian Scenario Planner now carries an “Open Beta” label, an August 6 last-updated stamp, and the tool’s first full user guide. The page defines it as a tool that “automates the creation of comprehensive reports that analyze the effectiveness of various marketing channels and their impact on sales or KPIs.” The feature itself first arrived in February; the guide is where the news sits, because the two ways to share the tool’s dashboard carry sharply different credential risk.
A Colab notebook, not a GA4 report, is the entry point
Meridian Scenario Planner is not part of Google Analytics 4, and the docs describe no GA4 integration. It runs from a Colab notebook, and that notebook requires at least one Meridian model already trained and stored in Google Drive or Google Cloud Storage. Running the notebook produces a unique link to a dashboard built in Data Studio, now branded Looker Studio, that opens after a Google account authorizes it. The dashboard itself has three tabs: Analyze, Plan, and Optimize.
Two link types put different people in control
The docs distinguish a temporary link from a persisted one, and the difference decides who ends up viewing whose data. A temporary creation link points to an unsaved dashboard instance; the recipient cannot see anything until they save it themselves, at which point they become the owner and authenticate with their own credentials and access rights. A saved link exists only once someone clicks “Edit and Save” on a temporary dashboard. From that point the credential model flips: per the docs, “Viewers effectively ‘impersonate’ the Owner, viewing data using the Owner’s credentials without needing direct access to the underlying data source.”
Which Meridian dashboard link is safe to share?
Google’s own recommendation is explicit: “To maximize data security, share the Temporary Dashboard link. This ensures the recipient must own the report and possess the necessary underlying data access rights.” A saved link is the more convenient one to forward, since the recipient needs no access to the underlying data source. It is also the one where every viewer reads the data on the owner’s credentials rather than their own.
Fit metrics come with an honesty clause
The Analyze tab reports R-squared, MAPE, and weighted MAPE. The docs favor the weighted version because it is less influenced by geographies and weeks that carry relatively small revenue. The same docs caveat that goodness-of-fit metrics “don’t give a complete picture of how good a model is for causal inference,” since the best-fitting model and the best model for causal inference can differ once the need to estimate an unobserved baseline enters the picture. Jeff, a developer relations engineer working on Meridian, frames R-squared in the walkthrough, as quoted by PPC Land, as “an important confidence check before we start spending money, rather than the ultimate proof of success.”
Budget planning has a fixed mode and a flexible one
The Plan tab optimizes a fixed budget for the highest overall ROI, or runs a flexible budget, which needs a target type: Target Total ROI or Target channel mROI. Depending on that choice, the flexible run either maximizes total incremental outcome or determines the maximum spend per channel. Constraints follow a stated formula: minimum spend equals one minus the min shift ratio, times historical spend, and maximum spend equals one plus the max shift ratio, times historical spend. In the demonstration figures, on an initial budget of 150 million, television at roughly $31 million gets a recommended increase of $1.6 million, and social at roughly $23 million gets $4.0 million more. Jeff describes the underlying shift as the point “where we transform our marketing budget from guesswork to decisive action,” and calls the tool a bridge “between your data science team and leadership.”
Accuracy keeps depending on what happens outside the tool
The shape is familiar from earlier coverage on this site. The GA4 diagnostic for missing GBRAID parameters showed that campaign-reporting accuracy depends on inputs assembled outside GA4’s own interface, in the ad click URL. Meridian’s credential asymmetry sits in a similar spot: Google’s guidance about it lives on a documentation page most sharers will never open. GA4’s own currency field for cost imports sat in the same family: a required input assembled outside the reporting UI decides whether the report is right.
The beta label does not remove the tool’s prerequisites. The model still has to be trained and stored before the notebook runs, and the docs still say goodness-of-fit gives no complete picture of causal validity. What changed on August 6 is that the sharing behavior is finally written down, in one sentence a data science team can forward along with the dashboard link itself.