AI Explained | How often is the model retrained?
Lauren Lovvorn
We do not retrain a model after it is initially stood up unless we have a request. If a model is working well for you, and you have no major changes in your data, a model is valid for about a year before it needs to be retrained, so we don't want to break what works!
While the model will not be retrained, we do run recommendations monthly. Learn more here: AI Explained | How often do you run new recommendations on my model?
When do I need to request a retraining?
If your data has had major changes, it might be time to retrain your model.
Some example reasons to retrain a model:
- A data source has been added/removed from the OneView
- The data logic in the OneView has been changed in some way
- Previously missing data has been added to ChannelMix
- You have less than one year of data. If this is the case, we may need a set schedule for the first year so we can develop the seasonality component.
Model Summary tab has the most recent model training date in the top section so we can see how old the model being used is.
How do I request a retraining?
To get a model retrained, submit a ticket to support specifying which dashboard contains the model you want to retrain.