Hi
@David Wheeler
Without looking at your data, it’s hard to say what will work best. I recommend trying both. In my experience working with large retailers, if you have a clear hierarchical structure, it usually helps using information from different levels of the hierarchy (for example, client-store and store-region). Here’s a tutorial for the
hierarchical forecast
Since you seem to have a large number of time series, I strongly recommend using the multi-series handling of TimeGPT since it will allow you to do the forecasts for all the series in one call. For the time being, the tutorial that you mentioned is what we have.