This message was deleted.

Hi @Ginger Holt. The easiest way to include them is to balance the dataset. For example, if you have new entities, you can impute with zeros (or another default value) the past of those entities, train your models and then reconcile them. The only drawback is that you have to be extra careful when producing forecasts for that new entities to avoid a misleading influence of the imputed values. Does that make sense? 🙂

Yes, that makes sense, thanks. If I have an estimate of how much I expect the new entities to represent, let's say 20% of the total, then this would definitely effect the reconciliation weights of the other entities. My approach will be to include a dummy account which includes 20% of the total so as to not get biased reconciliation weights for the existing entities (I know the existing entities should only represent 80% of the total).

Happy to help if you need anything else.