Comparison with Other MMM Libraries¶
Given the popularity of the Media Mix Modelling (MMM) approach, there are many packages available to perform MMM. Here’s a high-level overview of how Abacus compares to some of the most popular packages.
Criteria |
Abacus |
PyMC-Marketing |
Lightweight-MMM |
Robyn |
Orbit KTR / Karpiu |
Recast |
|---|---|---|---|---|---|---|
Language |
Python |
Python |
Python |
R |
Python |
RStan |
Approach |
Bayesian |
Bayesian |
Bayesian |
ML (Ridge) |
Bayesian |
Bayesian |
Inference Type |
Estimation |
Estimation |
Estimation |
Predictive |
Predictive |
Estimation |
Foundation |
PyMC / PyTensor |
PyMC |
NumPyro/JAX |
GlmNet + Nevergrad |
STAN/Pyro |
STAN |
Company |
T&Pm |
PyMC Labs |
Meta |
Uber |
Recast |
|
Open Source |
Own model |
Yes |
Yes |
Yes |
Yes |
No |
Model |
Build |
FOSS / Adopt |
FOSS / Adopt |
FOSS / Adopt |
FOSS / Adopt |
Buy |
Budget Optimizer |
Yes |
Yes |
Yes |
Yes |
No |
Yes |
Time-Varying Intercept |
No |
Yes |
No |
No |
Yes |
Yes |
Time-Varying Coefficients |
No |
Coming Soon* |
No |
No |
Yes |
Yes |
Custom Priors |
Yes |
Yes |
Yes |
No |
No |
Yes |
Lift-Test Calibration |
Yes |
Yes |
No |
Yes |
No |
Yes |
Out-of-Sample Predictions |
Yes |
Yes |
Yes |
No |
Yes |
Yes |
Unit-Tested |
Yes |
Yes |
Yes |
No |
Yes |
? |
Forecasting Module |
Yes |
No |
No |
No |
No |
Yes |
Scenario Planning Module |
Yes |
No |
No |
No |
No |
Yes |
Optimisation period |
Flexible |
Immediate |
Immediate |
Future |
None |
Unknown |
Note: “FOSS / Adopt” indicates the library is Free and Open Source Software, allowing users to adopt and potentially modify it. “Build” indicates a model primarily developed in-house (like Abacus). “Buy” indicates a commercial product.
* Feature status may change; check library documentation.