New Product Forecasting

Three routes for the borrowed data are standard: a predecessor product whose history the new item takes over in full or in part, an attribute model that estimates demand from category, brand, pack size and price tier, and a category or cluster aggregate that is disaggregated down to the item. Each route encodes an assumption that can fail in a specific direction, so a launch forecast is provisional by design and re‑estimated as observations arrive.

• Predecessor‑successor mapping: the new item takes over a defined percentage of the outgoing item’s history; that percentage states how comparable the planner considers the two.
• Attribute or analogue models: used where no direct predecessor exists – demand comes from items sharing attributes rather than from a single ancestor.
• Middle‑out hierarchy: the forecast is built at category or cluster level and disaggregated, which keeps a single item with three weeks of data from driving the number.
• Store‑profile benchmarks: for a launch across a network, demand per store is derived from the closest store profiles instead of the item’s own history.

New product forecasting in practice

The decisive input is rarely the model. It is the answer to two questions a category manager owns: which existing item the newcomer resembles, and what share of that item’s history it should inherit. A model propagates that judgment, it cannot correct it – a wrong predecessor produces a forecast that looks statistically sound and is off by a constant factor for the whole launch.

The first weeks of a launch are also the least usable part of the series. Where the launch is measured on deliveries to stores rather than on sell‑out, early volumes mix the initial shelf fill with consumer demand. Where the item enters with an introductory promotion, no clean observation of base‑price demand exists until the promotion ends. A baseline fitted on those weeks starts too high, and the correction usually arrives as an over‑order.

Veritico STOCK covers new items through predecessor‑successor mapping with a percentage takeover of history, and offers bottom‑up, top‑down and middle‑out approaches, the last being the useful one for new and strongly seasonal items. It cleans history of outliers, stockouts and promotional effects before calculating, and its exception‑based order management tags new items so a planner only opens the lines that need a decision – see demand forecasting and inventory optimization. Where the launch is part of an assortment change, Veritico RANGE simulates the effect of listing on category KPIs before the change is made; in one documented use case (400 products, 46 stores) that produced 12 newly listed products per store – a single use case, not an average across clients. More in assortment strategy and rationalization.

Related terms

• Demand Forecasting
• Baseline Demand
• Seasonality
• Assortment Optimization
• Dead Stock