Demand forecasting is the practice of predicting how much of a product customers will buy in a future period, using historical sales, trends, seasonality, and market signals. The forecast drives almost every operational decision downstream: how much to reorder, how much safety stock to hold, how much warehouse space to reserve, and how much cash gets tied up in inventory.
Get it wrong in one direction and shelves sit empty during peak demand; get it wrong in the other and cash sits frozen in fulfillment center storage, aging toward markdown. Forecasting is never about being exactly right — it's about being wrong by the smallest, most manageable margin.
What methods are used to forecast demand?
| Method |
How it works |
Best suited for |
| Time-series / trend analysis |
Projects forward from historical sales patterns and seasonality |
Established SKUs with a real sales history |
| Causal / driver-based forecasting |
Models demand against external drivers — price changes, ad spend, promotions, weather |
Products where a known lever (a promo, a price cut) predictably moves volume |
| Qualitative / judgmental |
Combines sales team input, market research, and expert judgment where data is thin |
New product launches with no sales history at all |
| Machine learning models |
Learns patterns across many SKUs and variables simultaneously, updating as new data arrives |
Large catalogs where manual forecasting per SKU doesn't scale |
Most operators run a blend: statistical models as the baseline, adjusted by human judgment for promotions, launches, and known anomalies the model hasn't seen before.
Why does cross-border selling make forecasting harder?
A single global demand number hides more than it reveals. Seasonality shifts by hemisphere and by holiday calendar; a product that peaks in November in the US may peak differently around a market's own retail calendar in Europe. Lead times differ by market too — ordering into a market with a longer import and customs pipeline needs a longer forecast horizon than reordering into an established domestic warehouse, or the safety stock cushion has to grow to cover the extra uncertainty. And demand itself isn't uniform: the same SKU can be a bestseller on one marketplace and a slow mover on another, which is why forecasting needs to run per SKU, per market — not once for the whole catalog.
Which metrics matter?
- Forecast accuracy (or its inverse, forecast error). How closely predicted demand matched actual sales, usually tracked per SKU and reviewed on a rolling basis rather than judged from any single period.
- Inventory turnover. How fast stock actually sells through — the downstream result of how good the forecast was.
- Stockout rate. How often in-demand SKUs go unavailable — the most visible and costly failure mode of underforecasting.
- Excess and aged inventory. Stock sitting unsold past its useful selling window — the cost of overforecasting, often hidden in storage fees until a markdown forces it into view.
What do brands get wrong with demand forecasting?
- Forecasting the catalog, not the SKU. Aggregate demand can look stable while individual SKUs swing wildly in opposite directions — the average hides both the stockouts and the excess stock happening underneath it.
- Ignoring lead time in the forecast horizon. A forecast is only as useful as the time you have left to act on it. Ordering inventory that needs weeks of ocean freight and customs clearance requires forecasting much further ahead than reordering from a domestic supplier.
- Never feeding actuals back into the model. A forecast that isn't compared against what actually happened, and adjusted, drifts further from reality every cycle — forecasting is a loop, not a one-time projection.
- Treating every market the same. Applying a US demand curve to a European launch (or vice versa) ignores real differences in seasonality, competition, and marketplace dynamics per country.
FAQ
What is the difference between demand forecasting and demand planning?
Demand forecasting is the prediction itself — the number. Demand planning is the broader process built around it: turning the forecast into purchase orders, inventory targets, and production or replenishment plans, and adjusting them as new data comes in.
How accurate should a demand forecast be?
There's no universal target — accuracy expectations vary by product type, demand volatility, and lead time. What matters more than hitting a specific number is tracking accuracy consistently over time and improving the forecasting process where the errors are largest and most costly.
Why is demand forecasting harder for new products?
Because there's no sales history to project from. New launches typically rely on qualitative methods — comparisons to similar products, market research, sales team input — until enough real sales data accumulates to shift to statistical forecasting.
Forecasting demand accurately across multiple countries, marketplaces, and lead times is exactly the kind of operating discipline that determines whether expansion is profitable. eBrands builds this per SKU, per market, for every brand we operate — see how it works for physical-goods brands.
