Why are there duplicates in the dataset? Should we deduplicate before aggregating?
Good work.
The forecast includes a confidence band that's asymmetric, wider on the upside than the downside, which suggests the underlying model assumes a particular risk distribution. Should we document whether this asymmetry reflects real constraints (e.g., a ceiling or floor) or is an artifact of the forecasting method? (edited to fix a unit)
The interval bounds are exclusive on the right, which means the month-end cutoff excludes the 31st for months with 31 days — that asymmetry is causing an off-by-one error in the YoY comparison.