What is time-series forecasting in fashion?
Predicting future values from data recorded over time, such as weekly sales, returns or web visits.
In short
Time-series forecasting predicts future values from data recorded over time, such as weekly sales, returns or web visits. In fashion it is used to project demand for continuing items, plan stock and anticipate seasonal peaks.
How does it work in practice?
Time-series models look for trends, seasonality and recurring patterns in historical data. For a NOS item, such as a basic white shirt, they can project weekly unit sales and account for seasonal peaks around holidays or back-to-work periods. The forecasts then inform purchasing, production and replenishment decisions.
Typical fashion applications include:
- Replenishment planning for never-out-of-stock and continuing styles.
- Return volume forecasts to plan warehouse capacity.
- Traffic forecasts for stores and e-commerce.
- Category-level planning where history is more stable than at style level.
Why does it matter?
Accurate forecasts reduce both stockouts and overstock, which affect margin, markdowns and sustainability. For wholesale brands, better forecasts on continuing items also improve the ability to fulfil re-orders reliably, which strengthens relationships with retail partners.
How does AI use it?
Classical statistical methods remain common, while machine learning models can incorporate additional factors such as price, promotions, weather and product attributes. Time-series forecasting often works alongside demand sensing, which refines short-term estimates with the latest signals.
Common pitfalls
- New products. Many fashion items are new each season and have little or no history of their own, so forecasts must borrow from similar past styles.
- Lost sales. Historical sales understate demand when items were out of stock.
- Distorted history. Promotions, store closures or one-off events skew patterns unless flagged.
- Too much detail. Forecasting at size and store level can be very noisy, so many teams forecast higher up and allocate down.
Frequently asked questions
How do you forecast demand for new fashion products without sales history?
By using similar past styles as proxies, based on attributes such as category, price band and colour, and by forecasting at category level before allocating to styles.
What is seasonality in fashion forecasting?
It is the recurring pattern of demand over the year, such as peaks for outerwear in autumn or swimwear in early summer, which models must capture to forecast accurately.