This page provides definitions for key terms used throughout the Meridian GeoX documentation.
Alpha
Significance level of the test, representing the probability of rejecting the null hypothesis when it's true, which is referred to as a Type I error.
Conversion
The primary KPI variable to be measured. It can be revenue, sales units, upper-funnel engagement metrics, or anything else that the marketing intervention may have a causal effect upon.
Cooldown
The period after the test period, used to evaluate if there's a delayed impact of the marketing intervention.
Cost per incremental conversion (CpIC)
CpIC is calculated as the inverse of the expected incremental conversion per dollar (1 / expected ICPD). When revenue data serves as the conversion metric, this value is equivalent to 1 / expected iROAS. This metric is primarily utilized to project budget needs for holdback experiments or specific holdback cells within multi-cell experimental designs. It is not required (and is ignored ) for go-dark or heavy-up experiments (or cells) since those tests measure existing live campaigns and experiment budget is projected using historical spend data.
Experiment budget
Projected total marketing spend changes due to the marketing intervention during the test period. When testing with your existing campaigns, the budget is determined by the size of the treatment group. For example, in go-dark experiments, a higher budget implies a larger percentage of market holdouts and larger media disruption. You may input a maximum allowed experiment budget per geo study design.
Geo unit
The geographic level at which the experiment is designed and executed, defined by regional market boundaries. While Meridian GeoX provides the flexibility to bring your own geo units, our recommendation is to use designated market area (DMA) for the US and geographic market area (GMA) for non-US regions. If these are not feasible, use aggregated local units such as administrative geographies (for example, provinces, states, or regions) in non-US regions.
Recommendation for market selection (US and International): Select your geographic units to minimize spillover effects between regions. Utilize market areas constructed from aggregated local units, such as cities and postal codes, where boundaries are intentionally drawn to avoid major commuting corridors. Ensure boundaries don't cross well-known commuting patterns to prevent daily travel from causing activity to spill over between treatment and control geographies.
Control geos
The group of geographic units that doesn't receive the marketing intervention, serving as the baseline to estimate the counterfactual or what happens if no marketing intervention is added of treatment geos.
Excluded geos
These geos are similar to control geos that don't receive the marketing intervention. However, they're completely excluded from the test and don't contribute to the counterfactual prediction.
Treatment geos
Group of geographic units that receives the marketing intervention being tested, such as a heavy-up or go-dark.
Incremental conversion
The lift in business results, such as transactions or revenue, that results from the marketing intervention, calculated using counterfactual modeling.
Incremental conversion per dollar (ICPD)
Indicates the efficiency of gaining incremental conversion with your marketing. Numerically, equals incremental conversion normalized by ads spend. It's equivalent to incremental return on ads spend (iROAS) if revenue data is used as conversion.
Minimum detectable effect (MDE)
The smallest percentage uplift that the experiment is statistically capable of detecting within the chosen statistical significance level, with the given treatment and control geo assignment, taking into consideration the experiment configuration—for example, the period of experiment or size of the treatment group.
Power
The probability that the experiment will correctly detect an incremental conversion if one truly exists, which is the ability to avoid an inconclusive readout.
Pretest
The period before the experiment begins, used to gather historical geo-level time series data to find study designs.
Test
The active period of the experiment during which the marketing intervention is applied to the treatment geos.