PySiMMMulator is an open source Python adaption of the R-package siMMMulator, which allows users to generate simulated data to use in testing Marketing Mix Models (MMMs). While this package cointains a full pipeline for data generation (configurable via .conf file) it can also be utilized in parts to generate select portions of MMM input data (ie campaign/channel spend)
Accessable via PyPI
pip install pysimmmulatorPySiMMMulator's simulator can either be run on a step-by-step basis, or can be run single-shot by passing a config file.
Run using this method, you'll be returned both a dataframe of for MMM input as well as the "True ROI" values for each of your channels. These true values are critical to validating your MMM model.
cfg = load_config(config_path="./my_config.yaml")
sim = simmm()
mmm_input_df, channel_roi = sim.run_with_config(config=cfg)Alternatively you may run each of the stages independently, which allows for easier debugging and in-run adjustments based on the results of each stage. The order of the stages is reflected below (without their inputs). Once you've run through every stage, results are available by calling the sim.final_df object (channel ROI results are stored as sim.channel_roi).
sim = simmm()
sim.simulate_baseline()
sim.simulate_ad_spend()
sim.simulate_media()
sim.simulate_cvr()
sim.simulate_decay_returns()
sim.calculate_conversions()
sim.consolidate_dataframe()
sim.calculate_channel_roi()
sim.finalize_output()A configuration file is required as input for this and should be passed as seen below. An output path can also be passed via -o, however when not passed the current working directory will be used.
pysimmm -i example_config.yaml -o .Setting up a dev environment
conda create --prefix=./env python=3.11
conda activate ./env
pip install -e '.[dev]'
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent