A forecast is not one certain future number. It is a set of assumptions, a range of possible demand and a rule for measuring error.
Extending an average sales line forward treats stockouts as weak demand, promotions as normal days and one-off spikes as a new trend. A useful forecast starts with history preparation.
1. Define object and horizon
Choose SKU, warehouse or region, fulfilment model, history period and decision horizon. A one-week replenishment view and a quarterly purchasing plan need different assumptions.
Declare the unit: completed sales, orders or demand before cancellations. Do not mix them in one series.
2. Mark stockout periods
Zero sales with zero availability does not mean zero demand. In Product 360, compare sales with availability and mark periods when customers could not buy.
Preserve the observed fact and any lost-demand estimate separately, including method and confidence.
3. Separate recurring patterns from events
Label discounts, campaigns, holidays, price changes, deliveries and major listing changes. Compare similar weekdays and seasons.
For a new SKU, use a range and documented analogues without presenting another product’s history as fact.
4. Build three scenarios
Create base, low and high cases. Change one assumption at a time: sales rate, price, advertising, return rate or seasonality.
If cleaned history indicates 10 units a day, 30-day scenarios might be 240, 300 and 390 units. This is a decision range, not a promise.
5. Backtest
Hide the latest historical weeks, forecast them and compare with actual results. Store error by SKU and horizon. A store-wide average can hide one product with systematic over-forecasting.
Record direction as well as percentage: does the method consistently overstate or understate demand?
6. Connect demand to replenishment
Replenishment also needs available stock, confirmed inbound stock, full lead time, minimum order quantity and chosen safety stock.
The forecast answers what may be needed. It cannot determine an order automatically without business constraints.
Forecast passport
Preserve calculation date, horizon, history and exclusions, scenario assumptions, method, actual result, error and adjustment decision. Forecasting then becomes a reproducible process rather than an unexplained number.
Current ProfitVena status
ProfitVena’s live Wildberries workflow and Product 360 provide sales, returns and inventory history within connected-source coverage. These data support scenario preparation and validation.
Forecast Center is preparing for integration and is demonstrated as a preview. This article does not claim that production forecasting or automated purchasing is already available. Other marketplaces remain on the roadmap until live connections are confirmed.
Technical part updated in ProfitVena.
ProfitVena is developed by Wicsora LLC and is part of the Wicsora ecosystem.
Methodology: How to forecast marketplace sales.
Previous article: Five Seller Reports to Review Every Week.
Next: how to calculate a reorder point and replenishment plan.
