schema.processors.marketing_processor.MarketingProcessor

Generates MarketingAnalysis protos for a given trained Meridian model.

Inherits From: ModelProcessor

schema.processors.marketing_processor.MarketingProcessor(
    trained_model: model_processor.ModelType
)

A MarketingAnalysis proto is generated for each spec supplied to execute(). Within each MarketingAnalysis proto, a MediaAnalysis proto is created for each channel in the model. One NonMediaAnalysis proto is also created for the model's baseline data.

Methods

execute

View source

execute(
    marketing_analysis_specs: Sequence[schema.processors.marketing_processor.MarketingAnalysisSpec]
) -> marketing_analysis_pb2.MarketingAnalysisList

Runs a marketing analysis on the model based on the given specs.

A MarketingAnalysis proto is created for each of the given specs. Each MarketingAnalysis proto contains a list of MediaAnalysis protos and a singleton NonMediaAnalysis proto for the baseline analysis. The analysis covers the time period bounded by the spec's start and end dates.

The singleton non-media analysis is performed on the model's baseline data, and contains metrics such as incremental outcome and baseline percent of contribution across media and non-media.

A media analysis is performed for each channel in the model, plus an "All Channels" synthetic channel. The media analysis contains metrics such as spend, percent of spend, incremental outcome, percent of contribution, and effectiveness. Depending on the type of data (revenue-based or non-revenue-based) in the model, the analysis also contains CPIK (non-revenue-based) or ROI and MROI (revenue-based).

Args
marketing_analysis_specs	A sequence of MarketingAnalysisSpec objects.

Returns
A MarketingAnalysisList proto containing the results of the marketing analysis for each spec.

output_type

View source

@classmethod
output_type() -> type[marketing_analysis_pb2.MarketingAnalysisList]

Returns the concrete output type that this ModelProcessor produces.

spec_type

View source

@classmethod
spec_type() -> type[MarketingAnalysisSpec]

Returns the concrete Spec type that this ModelProcessor operates on.

__call__

View source

__call__(
    specs: Sequence[S], output: pb.Mmm
)

Runs an analysis/optimization on the model using the given specs.

This also sets the appropriate output field in the given MmmOutput proto.

Args
specs	Sequence of Specs containing parameters needed for the analysis/optimization. The specs must all be of the same type as self.spec_type() for this processor
output	The output proto to which the results of the analysis/optimization should be attached.

Raises
ValueError	If any spec is not of the same type as self.spec_type().