Meridian GeoX lets you compare different methodologies during experiment design, helping you identify the best approach for your unique marketing objectives and study setups.
To compare among multiple methodologies such as
time-based regression (TBR) and pairing different geo assignment rules
such as RANDOM versus STRATIFIED_SAMPLING, you can use the
geox.compare_designs() function. This lets you to evaluate which approach
yields better metrics, such as a higher R-square score or a lower
minimum detectable effect (MDE), on your specific dataset.
Comparison of TBR under different assignment rules
The following snippet sets up two different configurations
- Option A: TBR with Random Assignment
- Option B: TBR with Stratified Sampling
and compares them side-by-side. Although in general, Stratified Sampling is recommended since it works nicely with TBR and it tends to yield designs with lower variance.
import datetime
import meridian_geox as geox
# 1. Define the configurations you want to compare
comparison_results = geox.compare_designs(
design_data,
[
# Option A: TBR with Random Assignment
(
geox.DesignConfig(
experiment_duration=datetime.timedelta(days=28),
methodology=geox.Methodology.TBR,
geo_assignment_rule=geox.GeoAssignmentRule.RANDOM,
design_output_count=5
),
geox.Constraints()
),
# Option B: TBR with Stratified Sampling
(
geox.DesignConfig(
experiment_duration=datetime.timedelta(days=28),
methodology=geox.Methodology.TBR,
geo_assignment_rule=geox.GeoAssignmentRule.STRATIFIED_SAMPLING,
design_output_count=5
),
geox.Constraints()
),
]
)
# 2. View the combined metrics table to see which performed better
print("Comparing Methodology Metrics:")
comparison_results.design_metrics