Fundamentals

Time series

A time series is a sequence of measurements recorded in order — one value per timestamp — so yesterday can inform tomorrow in a way a shuffled table of the same numbers cannot.

Businesses live on series even when they pretend they live on spreadsheets. Daily orders, hourly load, weekly claims. The order is the information. Shuffle the rows and you still have a mean; you no longer have a Monday.

A series can be regular (every hour, no gaps) or messy (missing days, duplicated stamps, a timezone change nobody documented). Forecasting software spends more time fixing the clock than it spends fitting a model. That is not a bug.

When you have many series that share a clock — every store, every SKU — people say panel or hierarchical data. The forecasting job is the same: predict the next ticks, at the grain someone will actually buy against.