The time scale of your historical and related datasets are taken from the calendar types you selected in your source data.
Time intervals for forecast horizons depend on the time scales given in the data collection. The time scales from the historical and related datasets determines supported time intervals for the forecast horizon. See the table below:
|Historical data time scale||Related data time scale||Possible time intervals for a forecast horizon|
Time scale aggregation
Historical data can be aggregated from lower to higher time scales, like days to weeks, or months. If your historical data is missing values towards the end of the timeline, for instance you have data for the first three days out of seven and you want to aggregate the data from days to weeks, PlanIQ uses internal logic to fill in the gaps. The limit for missing values is two days per week.
However, if you're missing data from the beginning of the timeline, for instance you have data for the last three days out of seven, PlanIQ does not fill in the missing values automatically. This may result in lower actuals for the first aggregated period.
Related data cannot be aggregated from lower to higher time scales, as it contains different types of data.
For example, for historical data, you can aggregate historical actuals like units sold. You cannot aggregate related data like daily prices into weekly prices, or store closures by day into store closures by week.
Time intervals for forecast models
The possible time intervals for a forecast model are dependent on the related data time scale.
If you only provide historical data, the time intervals of the forecast model must be equal to or greater than the time scale in the historical data.
If you provide both historical and related data, the time intervals for the forecast model must be:
- Equal to the related data time scale
- Equal or greater than the historical data time scale