This page introduces the analysis module, which is key to measuring the incremental impact of your campaigns. Learn how to prepare your data, understand counterfactual modeling, leverage design-aware inference, and use your reporting outputs for marketing mix model (MMM) calibration.
Prepare your analysis
To begin the analysis, prepare for pretest and test period time series data, such as new cost and conversion data. The analysis module seamlessly leverages provenance information—the specific design configurations, analysis methodology, and geo assignment generated during the earlier study design phase. Both the original design and the time series data are required during the pretest and test periods to ensure that the analysis of the incrementality effects and associated confidence intervals are calculated accurately. For more information, see Prepare your analysis data.
Counterfactual modeling
To calculate the actual incrementality of your campaigns, Meridian GeoX framework relies on counterfactual modeling, primarily time-based regression (TBR) methodology. For more information, see Counterfactual modeling.
Robust inference
To determine whether the observed lift is statistically significant, Meridian GeoX employs an advanced methodology called design-aware inference. This approach solves the methodological challenges of other standard inference frameworks, which often fail for GeoX. Our robust inference approach ensures precise p-values without inflating the type-I error, while also maximizing statistical power for your study. For more information, see Robust inference.
Moving forward: Outputs and MMM calibration
The final output of the analysis phase provides you with actionable metrics, including the point estimate of incremental conversions (lift) and percentage lift, along with associated confidence intervals and p-values for these metrics. You can also find metrics like the incremental conversion per dollar (iCPD)—equivalent to incremental return on ad spend (iROAS) if revenue data is used—accordingly. You can generate visualizations of these results for a deep-dive analysis. These outputs can be adjusted if needed and exported as custom, data-driven priors to calibrate your Meridian model for enhanced long-term measurement. For more information, see Analysis outputs and Intro to incrementality based calibration.