When a product almost never sells: intermittent demand.
A spare part that goes out three times a month, a niche medicine, a special screw. Many days at zero and then suddenly a large sale. This chapter explains why the average misleads and how these products are forecast.
Many zeros
StockAI calls a product intermittent when 40% or more of its periods have zero sales. It is more common than it seems: in a hardware or auto-parts catalog it can be half of the item codes.
The problem is that most models expect a line that rises and falls, not a row of zeros with jumps. They try to follow every jump and end up forecasting sales on days when there will be none, or zeros right before a large order.
The Croston idea: two questions instead of one
Instead of asking "how much sells per day?", the Croston method separates two questions: when something sells, how much does the customer take? And how many days pass between sales?
Each answer is updated with an average that gives a bit more weight to recent data (10% to the new data point). The forecast is the ratio: typical size divided by typical interval, with a small correction (the SBA one) because the original Croston tends to overshoot.
The forecast is a rate, not a calendar
The result is a flat line: "on average, 0.4 units per day". It does not say which day the customer will come, because that cannot be known. For buying that is not needed: what matters is how much will sell during the supplier lead time, and a rate answers that well.
Move the simulator. With few sales, the simple average jumps with each sale; Croston moves much less, because it separates the size from the interval.
A product that almost never sells
SimulatorSixty days of a sporadic product. Compare the simple average with the Croston rate.
Simulation with invented numbers. Croston with the SBA correction and the same weight StockAI uses (10% to the new data point).
When a product stops selling
Croston has a weak point: if a product stops selling, it keeps forecasting the same rate, because it only updates when a sale happens. TSB fixes that: each period it updates the probability that a sale happens, and if weeks pass at zero the forecast drops on its own.
Today, for an intermittent product, Croston and ETS compete automatically. TSB is available in the model list: if you turn it on and neither Croston nor ETS is in your selection, it competes alongside the other models you chose.
An honest warning
With intermittent demand, the service level you set (chapter 6) is less reliable: the safety stock math assumes sales that are more even. That is why StockAI flags these products with a warning, so you review them with your own judgment.
What you take away
- A product is intermittent if 40% or more of its periods have zero sales.
- Croston separates how much the customer takes from how often sales happen, and forecasts a rate.
- TSB, available as an option, lowers the forecast when a product stops selling.
- For these products the service level is less reliable, and StockAI warns you about it.
Where you see it in StockAI
Intermittent products carry their label on the forecast and a warning in Inventory. If a product has stopped selling, also check its signal: it is probably at OVERSTOCK.
Keep reading
- The forecast rangeWhere the band around the forecast comes from and why it widens.
- LearnA guide for non-technical readers: what a time series is, how a forecast is measured, what each model does, intermittent demand, the forecast range, and how it turns into a purchase.
- How it is calculated, in briefThe best way to understand it is to see it with your own products.
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.