Demand forecasting for purchasing: from past sales to the purchase order.
Forecasting demand is not guessing the future: it is estimating how much of each product will sell during the time an order takes to arrive, with the margin of error in view. This guide explains how that forecast is used to buy better, and how StockAI does it.
What forecasting is for in purchasing
The forecast is the input to almost every purchasing decision: the reorder point, the safety stock and the quantity to order all depend on how much is expected to sell. A flat average works while demand is stable; it fails with seasons, trends, promotions or sales that come in bursts.
Why one method is not enough
Not every product sells the same way. A stable one forecasts well with a simple model; a seasonal one needs a model that understands the calendar; an intermittent one, with many zero periods, needs models built for that. That is why StockAI classifies each product by how it sells (stable, seasonal, intermittent or volatile) and makes up to nine models compete, product by product: LightGBM, XGBoost, ARIMA, SARIMAX, Prophet, ETS, Croston, an LSTM neural network and a global model that learns from the whole catalogue.
How the winning model is chosen
- Each model has a stretch from the end of your history hidden and is asked to forecast it, in several cuts. That measures it against data it did not see.
- The winner is chosen with a cost in which running out weighs three times more than having extra, because in purchasing a stockout and an excess do not cost the same.
- It has to beat two naive forecasts: repeating the last value and repeating the previous season. If it does not, a complex model is not justified.
- The winner is retrained on the full history and projects demand with a probable range around it, not a single number.
From forecast demand to the purchase
The forecast is not the end. It is crossed with stock, what is on its way and the supplier’s lead time to give the reorder point, the stock signal and the suggested quantity. That part is explained in how to calculate the reorder point and safety stock. The full walkthrough, with screenshots, is in how StockAI works.
The limits, said up front
- A product with fewer than 20 periods of history is left out of the forecast and shown as SIN DATOS (no data): no number is invented for it.
- The range is not a guarantee: it is the band where sales will probably fall.
- It cannot guess what never happened. A new promotion or a new large customer is tried in the scenario simulator.
- The colour of the stock signal is decided by a published rule, not the model: AI forecasts and the rule decides.
- It does not buy for you: it suggests, and you send the order.
How to measure whether a forecast is good
On dates the model did not see, product by product, and against what actually sold afterwards. StockAI shows each product’s error, not one percentage for the whole catalogue, and does not publish a marketing accuracy figure: the one you see is your data’s.
Frequently asked questions about demand forecasting
What history do I need to forecast?
Which models does StockAI use?
What about a new promotion?
How often is the forecast updated?
And to see the product
Upload your history and see each product’s forecast and error.
Keep reading
- How to calculate the reorder pointWhat the reorder point is, the formula (demand during lead time plus safety stock), a worked example and how StockAI computes it for every product.
- Safety stockWhat safety stock is, the formula with service level, a worked example, and how sales variability and supplier delays change it.
- Sales forecasting for distributorsWhat decides a distributor’s purchase is how much it will sell: a sales forecast per product and, from it, what to order, how much and from whom. What to demand and how to choose.
Put your own AI to work forecasting your sales. Today.
Three ways to start, depending on how much you want to commit right now. All three lead to the same product.
Create your account
With the whole engine. Upload your sales file and see your first forecast, and with it your first list of what to order.
Create my accountLook before giving your email
An instant trial account with sample data. It lasts 24 hours and is then erased.
Try it without signing upTalk to us
For the corporate plan, the source code, or to see it with your data and someone beside you. We answer within 24 hours.
Sales and contact: contacto@stockai.es. Phone: +506 7186 2820.
Sales forecasting made in Costa Rica, for companies across Latin America.