Demand forecasting means estimating future sales of a product or service in order to size production, inventory, and procurement in advance. It isn't fortune-telling: it's a method that combines sales history, market knowledge, and the judgment of the teams. A good forecast is never perfectly accurate. It is simply reliable enough to support better decisions than intuition alone.
The main families of demand forecasting methods
Qualitative methods
These rely on expert judgment: sales teams, market opinion, and panels. They are used when there is little or no history (a new product, a new market) or to bring in information the numbers don't yet capture.
Quantitative methods (time series)
These project the sales history forward: moving averages and exponential smoothing, accounting for trend and seasonality. They suit established products with regular behavior, where the past is a reasonable guide to the future.
Causal methods
These link demand to explanatory factors (price, promotions, weather, and economic indicators). They are richer, but they also require more data and more rigor to avoid confusing correlation with causation.
Choosing a demand forecasting method in an SME
The right choice depends on the data available and what is at stake. A regular product with several years of history can be forecast very well with a simple quantitative method; a launch with no history relies on judgment. The useful rule in an SME: start simple, measure the forecast error, and add complexity only if it genuinely improves accuracy. The forecast then feeds decisions: it is the first building block of the “decision” layer of the Two-Layer Model, the one that guides execution rather than being driven by it. Because forecast quality directly drives inventory levels, it is also central to our inventory optimization work.
The common mistake
The most frequent mistake is chasing the perfect forecast instead of the useful one. Models get refined while the real question (what decision does this number drive?) goes unanswered. The other trap is never measuring forecast error: without tracking it, with a metric such as MAPE or WMAPE together with bias, there's no way to know whether the method is improving or deteriorating. One last pitfall awaits SMEs: confusing the statistical forecast with the sales target. The former estimates what will probably happen; the latter expresses what the company wants to achieve. Mixing the two often introduces an upward bias, and overstocking follows. Keeping the two separate, then reconciling them deliberately, is one of the hallmarks of a healthy forecasting process.
Assess your forecast
Want to know whether your demand forecasting is reliable enough to drive your inventory decisions? Our forty-five-minute free 45-minute session gives you an answer, with no commitment.
Frequently asked questions
What is demand forecasting?
It's the estimation of future sales of a product or service, in order to size production, inventory, and procurement in advance.
What are the main demand forecasting methods?
Qualitative methods (expert judgment), quantitative methods (time series such as moving average or exponential smoothing), and causal methods (explanatory factors).
Which demand forecasting method should an SME choose?
The one that matches the data available: judgment for a launch with no history, a simple quantitative method for a regular product.