Switch off saturation transformation in pymc marketing MMM

Questions version agnostic
1

Hi,

I have a slightly modified application of the pymc marketing MMM, for which I do not want to use the saturation transformation.

I have two kinds of purchases: With and without discount codes. For purchases with discount code, I know the marketing channel which triggered the purchase. For the others, I don’t. I now want to estimate which marketing channel caused the purchases without discount codes, assuming that every purchase must have been triggered by a channel (which is reasonable in my case).

My predictor variables are not costs, but purchases with adcodes. Therefore, the saturation transformation does not make sense.

How can I switch it off?

Thanks for any hint
Matthias

2

Hi @bayesian_padawan

You are able to define a custom Saturation transformation which the transformation y=x which would allow for an equivalent of no saturation transformation

import pandas as pd

from pymc_marketing.mmm import (
    MMM,
    GeometricAdstock,
    SaturationTransformation,
)


class Identity(SaturationTransformation):
    lookup_name = "identity"

    def function(self, x):
        return x

    default_priors = {}


mmm = MMM(
    adstock=GeometricAdstock(l_max=8),
    saturation=Identity(),
    date_column="date_week",
    channel_columns=["x1", "x2"],
    control_columns=[
        "event_1",
        "event_2",
        "t",
    ],
    yearly_seasonality=2,
)

data_url = "https://raw.githubusercontent.com/pymc-labs/pymc-marketing/main/data/mmm_example.csv"
data = pd.read_csv(data_url, parse_dates=["date_week"])

X = data.drop("y", axis=1)
y = data["y"]
mmm.fit(X, y)
1 Like