Sales forecasting: definition, methods and example

Sales forecasting involves estimating the quantities a product, category, or store will sell over a given period: a week, a month, or a season. It starts with historical sales data, adjusted for stockouts and past promotions, and then incorporates seasonality, planned promotions, and external factors such as weather or holidays. In retail, it's used to size orders, prepare sales campaigns, and set prices that reflect available stock. It's not a budget: a forecast describes what is likely to happen, while a budget describes the target (see sales forecasting and budgeting ).

Sales forecasting methods fall into three categories: qualitative methods, statistical time-series methods, and machine learning models. The first relies on the judgment of buyers or salespeople, useful for a launch without historical data. The second, such as moving averages or exponential smoothing, extrapolate past trends and seasonality. The third combines numerous variables (price, promotions, weather, competitor pricing) and captures their interactions. A robust system often blends all three, depending on the specific product. The step-by-step method is detailed in the section on AI-powered sales forecasting .

The accuracy of a sales forecast depends primarily on its level of detail: it is always more reliable at the category and monthly level than at the SKU, store, and weekly level. High-turnover products are easy to forecast; new products, seasonal items, and long-tail products are much less so. Accuracy is measured using the MAPE (Mean Percentage Error) and bias, which reveals a tendency to overestimate or underestimate. Rather than a universal target figure, the goal is to track these indicators over time and improve them category by category.

Artificial intelligence improves sales forecasting by incorporating far more variables than a traditional statistical model and learning their interplay. For example, a machine learning model can estimate that a soda promotion sells more in warm weather and takes sales away from a competing brand. It retrains itself on recent sales data and adapts more quickly to changing trends. Its limitations are well-known: it requires a specific historical record and cannot predict events that have never been observed before, hence the importance of retaining the team's adjustments. To learn more: the 8-step method and KPIs .

Sales forecasting and pricing are linked in both directions: price influences volumes, and expected volumes guide pricing. A good forecast incorporates price elasticity , that is, how sales react to an increase or decrease in price. Conversely, anticipating overstock at the end of the season allows for gradual markdowns rather than a late liquidation. This is the principle of predictive pricing : simulating the effect of a price on sales before setting it. Booper offers AI-powered sales forecasting linked to pricing decisions.

Sales forecasting is best managed by a single team, tasked with producing a benchmark figure shared by sales, purchasing, supply chain, and pricing. Without this common reference point, each department builds its own forecast, and orders, prices, and sales targets are based on different assumptions. Implementation often involves a streamlined S&OP process: a monthly review where each department contributes its information, such as planned promotions or supplier constraints, before the figure is validated. See who should lead the sales forecast .