How do you forecast supplier orders?

A supplier order is not simply about replacing what was just sold. To order the right quantity, you need to anticipate what will be needed until the next restocking, while taking into account what is already available or on its way.

Ordering too little exposes you to stockouts. Ordering too much ties up cash and increases the risk of overstock.

The goal is therefore to determine, for each item, when to order and how much to order.

What data should you take into account?

Preparing a supplier order means cross-checking several pieces of information.

It's the combination of this data that lets you estimate the real need.

Don't rely only on current stock

Knowing that 15 units are left isn't enough to decide on a supplier order.

You also need to look at:

• the sales rate;
• the supplier's lead time;
• orders already in progress;
• the safety stock you want;
• seasonal variations.

The available quantity is therefore a starting point, not an ordering decision.

The reorder point: when to act

The reorder point, or restocking threshold, is the stock level at which a new order must be placed to cover sales until the next delivery.

A simple formula is:

Reorder point = (sales rate × supplier lead time) + safety stock

The reorder point answers the question “when to order?” The quantity to order answers a different question: “how much to order?”

This distinction matters. Reaching a threshold tells you that you need to act, but it doesn't mean you should automatically reorder a predefined quantity.

Estimating needs

Let's take a simple example.

A product sells 5 times a day on average and the supplier delivers within 10 days.

You therefore need to plan for about:

5 × 10 = 50 units

to cover sales during the restocking lead time.

If the merchant wants to keep 10 units of safety stock, the target need becomes:

60 units

If 20 units are still in stock, the quantity needed to reach this level becomes:

40 units

This example assumes steady demand and no other order in progress. In practice, both of these should also be checked before validating the restocking.

Calculating an appropriate safety stock

Arbitrarily setting the same safety stock for every item easily leads to holding too much stock on some and not enough on others.

The level needed mainly depends on two uncertainties: demand variation and supplier lead time reliability.

A commonly used formula is:

Safety stock = Z × standard deviation of demand × √(supplier lead time)

The Z factor corresponds to the service level you are aiming for. The higher this level, the larger the safety margin — at the cost of more stock tied up. As an example, a 95% service level corresponds to a Z factor of about 1.65.

You don't need to apply this formula systematically to every item to grasp the principle: a product with irregular sales, or whose supplier often misses lead times, generally needs more margin than a stable product that is restocked quickly.

Factoring in sales trends

A historical average can hide a recent change.

If sales have been rising for several weeks, future demand may be higher than the average calculated over a long period. Conversely, an item whose sales are slowing risks being over-ordered if you simply repeat the usual quantities.

Two simple approaches help refine the analysis:

• the moving average, suited to relatively stable demand, which smooths out one-off variations;
• recency weighting, which gives more weight to the latest periods and reacts faster to a change in trend.

The goal isn't to always pick one or the other, but to prevent an outdated average from hiding what is actually happening.

Taking seasonality into account

For some products, the last few weeks aren't the best reference.

December sales, for example, can be very different from February sales. In that case, comparing the period with the same period last year may be more relevant than a simple recent average.

A seasonality index can also be used. It compares sales for a given period with the annual average and lets you adjust the forecast when a peak or a dip is usually observed.

Seasonality thus prevents you from underestimating a need before a busy period or, conversely, from keeping a high order level when demand is normally about to slow down.

The case of minimum order quantities

The calculated need can't always be ordered as is.

A supplier may require:

• a minimum order quantity (MOQ);
• full-carton case packs;
• a minimum amount;
• certain delivery conditions.

If the real need is below the required minimum, you have to make a trade-off.

It may make sense to group several upcoming needs together, to delay the order slightly when stock allows it or, in some cases, to accept a temporary overstock if it remains economically preferable.

A theoretically optimal quantity must therefore always be checked against the supplier's actual conditions.

Not all products need the same follow-up

There is no need to spend the same time on every item.

Fast-moving, high-value products, or those with long supplier lead times, generally deserve closer monitoring.

An ABC classification can help set this priority. A items account for a large share of revenue or stock value and therefore justify frequent checks. Less critical items can be tracked with simpler rules.

The goal is to focus the analysis where an ordering mistake would have the most impact.

The most common mistakes

Several mistakes can skew a supplier order:

• applying the same safety stock to every item;
• forgetting orders already placed but not yet received;
• using an average calculated over too long a period while sales are changing;
• repeating an exceptional increase caused by a promotion or an event;
• not correcting the forecast after a stockout.

This last case is particularly misleading: when a product has been unavailable, the sales recorded during that period don't represent the full potential demand. Using them as they are can lead you to underestimate the next need.

The full workflow, from need to stock update

A supplier order is part of a broader process:

1. NeedAn item is approaching its threshold or is at risk of a stockout.
2. Quantity proposalSales, available stock, expected needs and supplier constraints are used to estimate the quantity needed.
3. OrderThe purchase order lists the items, quantities and planned terms.
4. SendingThe order is sent to the supplier after validation.
5. ReceivingDelivered products are checked against the order.
6. Stock updateThe quantities actually received are added to available stock.

Restocking therefore doesn't stop at calculating a quantity: the quality of your inventory also depends on following up the order and its receipt.

Complete, partial and discrepant receipts

A delivery doesn't always match exactly what was ordered.

Complete receipt: all expected items and quantities have been delivered. Stock can be updated with the quantities received.

Partial receipt: part of the order is missing. The remainder must stay identified so it isn't mistaken for a new restocking need.

Receipt with discrepancies: the items or quantities delivered differ from the order. The discrepancy must be checked before updating stock to avoid skewing the available data.

This check is essential: your next restocking decisions will themselves be based on this data.

Importing a delivery note or supplier invoice

Receiving can become particularly time-consuming when the lines of a delivery note or an invoice have to be re-entered manually.

With Gillia's AI Vision, a delivery note or supplier invoice can be imported from a photo, a PDF or a spreadsheet.

Gillia reads the items, quantities and prices in the document and matches them to the existing catalog, even when the supplier's description differs slightly from the internal name.

Any discrepancies with the original order can thus be identified before validation.

“Import this delivery note and match the lines with my catalog.”

After checking, the validated receipt can be used to update the quantities actually received without re-entering each line manually.

How can AI prepare a supplier order?

Restocking preparation is where AI can save a large share of repetitive checks. The merchant can, for example, ask Gillia:

“Which products should I reorder this week?”

Then add:

“Prepare my supplier order based on the last few weeks of sales and the remaining stock.”

Or focus the analysis on priorities:

“Show me only the urgent orders.”

Gillia can then cross-check the available information — the actual sales rate comes straight from the sales analysis, while available stock and supplier lead time come from inventory management — to suggest the items and quantities to review.

The proposed order still has to be validated before it is sent.

The point is therefore not to replace the merchant's decision, but to avoid manually rebuilding the situation of each item before every order.

The multi-location dimension

In a network of several points of sale, a supplier order isn't always the first solution.

Before reordering an item, it can be useful to check whether another location has a surplus instead.

An internal transfer can sometimes cover the need faster than a new delivery, while avoiding buying more stock.

From advice to action

The value of an AI agent doesn't stop at displaying a recommended quantity.

When the necessary tools are available and connected, Gillia can support the different steps of the process: analyzing the need, preparing an order proposal, submitting it for validation, then using the receiving information.

This is the understand → analyze → execute logic: the analysis leads to a concrete operation, while keeping a validation step when the action genuinely commits the business — the principle behind an AI agent for retail.

When the same restocking check comes up regularly, it can also become a routine instead of an entirely manual check.

To round out this topic, supplier orders naturally fit into overall inventory management: for the complete logic, inbound and outbound, see Store inventory management; to avoid waiting for a stockout before acting, see Anticipating a stockout, and to avoid the opposite excess, Overstock and Inventory turnover. Gillia use cases also let you go further: create a supplier order, receive a partial order or an order with discrepancies or prepare targeted restocking based on thresholds.