"Inventory forecasting" sounds like data science, but for most Shopify stores it comes down to a few practical methods. Here's what each one is, its trade-offs, and which to use — without any black-box maths you can't question.
1. Moving average
Average the last N periods of sales (e.g. the last 4 weeks) to estimate demand. Simple, stable, and a fine default for steady sellers. Downside: it reacts slowly to a genuine change in demand.
2. Weighted moving average
Same idea, but recent periods count more than older ones — so the forecast responds faster to trends while still smoothing noise. A good upgrade from a plain average for products whose demand is drifting up or down.
3. Exponential smoothing
A weighted average where the weighting decays smoothly into the past, controlled by one factor. It reacts to change without overreacting to a single odd day. Popular because it's accurate and cheap to compute.
4. Seasonal forecasting
Layers a seasonal pattern on top of the trend — essential if you have Black Friday, summer or holiday peaks. It needs a year or more of history to learn the pattern, so it's for established products, not new ones.
5. The reorder-point method
Rather than predicting a curve, this method answers the operational question directly: when do I reorder, and how much? It combines three numbers you can measure today:
- Sales velocity — units sold per day.
- Lead time — days from order to shelf.
- Safety stock — your buffer.
Reorder point = (sales velocity × lead time) + safety stock
Try it with the reorder point calculator. For most small and medium stores, this is the method that actually prevents stockouts — see the full walk-through in how to forecast inventory on Shopify.
Which method should you use?
- Most stores: the reorder-point method, on top of a moving or weighted average of velocity.
- Strong seasonality: add a seasonal layer for your peak products.
- Everyone: insist on explainable numbers — you should always see why a recommendation says what it says.
Foreshelf uses the transparent, reorder-point approach: it computes each product's velocity, applies your lead time and safety buffer, and shows the reasoning behind every recommendation — so you forecast confidently without trusting a black box.