pie

Predicted Incrementality by Experimentation (PIE).

The recommended entry point is PIEModel, which wraps the model in a RegressionModelBuilder interface with standard .fit(), .save(), and .load() methods.

Examples

Fit on a corpus of past RCTs, then predict incrementality for new campaigns:

import pandas as pd
from pymc_marketing.pie import PIEModel

X = pd.DataFrame({...})
y = pd.Series([...])

model = PIEModel(
    pre_determined_features=["objective", "vertical", "budget"],
    post_determined_features=["exposure_rate"],
)
model.fit(X, y, random_seed=42)
predictions = model.predict(X_new)

Modules

model

Predicted Incrementality by Experimentation (PIE) model.