In real life you almost never get a perfect, clean history for all SKUs and periods.
You may have short history, gaps, changes in coding, or only POS data from one
channel. :contentReference[oaicite:4]{index=4}
Short history (new product or new market).
Partial history (only retail POS, no e-commerce). :contentReference[oaicite:5]{index=5}
Irregular data (missing weeks or months).
Format changes (SKU codes merged, brands restructured).
The goal is not to “wait until data is perfect”, but to reasonably restore missing
information and make your assumptions explicit. :contentReference[oaicite:6]{index=6}
Key Techniques for Incomplete Data
Interpolation – fill gaps using neighbouring periods (e.g. mean of previous and next month). :contentReference[oaicite:7]{index=7}
Extrapolation – extend a trend beyond available data (e.g. early launch months for a new SKU). :contentReference[oaicite:8]{index=8}
Proxy variables – use category, segment or analog SKU instead of missing data.
Analog data – borrow history from similar product, market or customer. :contentReference[oaicite:9]{index=9}
Задача — не ждать «идеальных» данных, а аккуратно восстановить недостающую
информацию и явно зафиксировать допущения. :contentReference[oaicite:15]{index=15}
Основные приёмы при неполных данных
Интерполяция (Interpolation) — заполнение пропусков по соседним периодам. :contentReference[oaicite:16]{index=16}
Экстраполяция (Extrapolation) — продление тренда на будущие периоды (например, для новых SKU).