reallift 0.3.4

Causal Inference Library for Lift Measurement

RealLift

Causal Inference Framework for Geo Experiments & Marketing Science

RealLift is an advanced Python library engineered to measure the true incremental impact (Lift) of marketing interventions through rigorous Causal Inference, Synthetic Control methodologies, and Robust Significance Testing.

---

Framework Pillars

RealLift is built upon three layers of defense against noise and volatility:

1. Auditable Planning (Design of Experiments): Algorithmically identifies the optimal geographic clusters and projects statistical power (MDE) prior to any campaign investment.
2. Causal Inference (Synthetic Control): Formulates robust counterfactuals via Convex Optimization with Intercept, ensuring behavioral alignment even across differing baseline levels.
3. Confidence Validation (Placebo & Significance): Defends analytical conclusions through permutation tests based on the MSPE Ratio and non-parametric Bootstrap confidence intervals.

---

Installation

pip install reallift

---

Quick Start Guide

This end-to-end example demonstrates how to simulate geographic data, plan the optimal experiment design, inject a simulated marketing effect, and precisely measure the impact.

1. Generate Historical Data

Simulate daily performance for 27 candidate geos over a 90-day baseline.

from reallift import generate_geo_data

geo_data = generate_geo_data(
    start_date="2025-01-01",
    end_date="2025-03-31",
    n_geos=27,
    pre_only=True,
    trend_slope=0.01,
    seasonality_amplitude=3,
    seasonality_period=7,
    noise_std=[1, 1.5],
    base_value=[50, 100],
    random_seed=42,
    save_csv=True,
    pre_file_name="demo_geodata_pre_test.csv"
)

2. Design the Experiment (Pre-Test Phase)

Use the DoE pipeline to rigorously select the best treatment and control geometries based on historical correlations.

from reallift import design_of_experiments

doe = design_of_experiments(
    filepath="demo_geodata_pre_test.csv",
    date_col="date",
    start_date="2025-01-01",
    end_date="2025-03-31",
    pct_treatment=None,
    fixed_treatment=None,
    experiment_days=[21, 28, 30, 35, 60]
)

3. Inject Simulated Marketing Campaign

Simulate a 21-day campaign with an artificial lift applied strictly to the optimal treatment geos assigned in the DoE.

from reallift import generate_simulated_intervention

geo_data_intervention = generate_simulated_intervention(
    filepath="demo_geodata_pre_test.csv",
    days=21,
    treatment_geos=doe['scenarios'][1]['treatment_pool'],
    lift=[0.05, 0.10],
    trend_slope=0.01,
    seasonality_amplitude=3,
    seasonality_period=7,
    noise_std=[1, 1.5],
    random_seed=42,
    save_csv=True,
    as_integer=True,
    file_name="demo_geodata_post_test.csv"
)

4. Evaluate Incremental Impact (Post-Test Phase)

Measure the precise financial and percentual lift utilizing the Causal Inference engine against the final dataset.

from reallift import run_geo_experiment

results = run_geo_experiment(
    filepath="demo_geodata_post_test.csv",
    date_col="date",
    treatment_start_date="2025-04-01",
    treatment_end_date="2025-04-22",
    doe=doe, 
    scenario=1,
    plot=False
)

---

Technical Differentiators

• SER Engine (Synthetic Error Ratio): Proactive volatility filtering during the design phase to eliminate "Zombie Controls."
• Convex Intercept: Intelligent baseline shift absorption that preserves the interpretability of synthetic weights ($\sum w = 1$).
• MSPE Ratio Strategy: A placebo methodology resilient to the natural variance of high-frequency volatile markets.
• Operational Freedom: Aggressive donor curatorship via ElasticNet that often frees up to 50% of geographic regions for other commercial operations.

---

Systems & Dependencies

• Mathematics: cvxpy, scipy, numpy
• Data Engineering: pandas, scikit-learn
• Visualization: matplotlib

---

Developed by Roberto Junior.