Does lifetimes package handles seasonality automatically?

[skagrawal10]

Hi,

I have a usecase, where we see unusual spikes in the months of Jan & Jun. In these months, we see, 4 times more sales from same customers. When we have used this package for this use case, it is projecting very high sales in subsequent months like Feb & Jul.

So, my question here is: should we remove seasonality before entering data into this package? or is it handled automatically in model?

Please share your thoughts and how we can handle these scenarios.

Thanks

[roestigraben]

Hi,

I have a very similar use case with the very high sales peaks at Xmas. It looks strange to me that your model reacts in such a way as you describe it.

At around minute 28 of this video, you'll see Peter Fader's comments on peaks' periods.

A good way to see your model fit is to use the cumulative transactions compare between model and actuals. Your model should give you a good baseline.

If you want to reflect the high seasonality, I would imagine that you modulate your model with some secondary model that tracks to these effects (maybe a machine learning approach).

I am looking into using a 2nd model which tries to follow seasonality but is also overall per year consistent with the baseline.