Incrementality#

class pymc_marketing.mmm.incrementality.Incrementality(model, idata=None, data=None)[source]#

Incrementality and counterfactual analysis for MMM models.

Computes incremental channel contributions by comparing predictions with actual spend vs. counterfactual (perturbed) spend, accounting for adstock carryover effects. See the module docstring for the full mathematical formulation and design rationale.

Parameters:

modelMMM: Fitted MMM model instance.
idataaz.InferenceData, optional: InferenceData containing posterior samples and fit data. Mutually exclusive with data.
dataMMMIDataWrapper, optional: Pre-built data wrapper. When provided, idata is taken from data.idata and no re-wrapping occurs. Mutually exclusive with idata.

Attributes:

modelMMM: The fitted MMM model.
idataaz.InferenceData: Posterior samples and fit data.
dataMMMIDataWrapper: Data wrapper for accessing model data.

Raises:

ValueError: If both idata and data are provided, or neither is.

Examples

>>> incr = mmm.incrementality
>>> roas = incr.contribution_over_spend(frequency="quarterly")
>>> cac = incr.spend_over_contribution(frequency="monthly")

Methods

`Incrementality.__init__`(model[, idata, data])
`Incrementality.compute_incremental_contribution`(...)	Compute incremental channel contributions using counterfactual analysis.
`Incrementality.contribution_over_spend`(frequency)	Compute incremental contribution per unit of spend.
`Incrementality.marginal_contribution_over_spend`(...)	Compute marginal contribution per additional unit of spend.
`Incrementality.spend_over_contribution`(frequency)	Compute spend per unit of incremental contribution.