Inventory Demand Forecasting: Formulas, Weeks of Supply, and the Buy

Inventory demand forecasting is the practice of predicting how many units of each SKU you will sell over a set period, then converting that number into how much stock to hold and when to reorder. Demand forecasting stops at the units you expect to sell. Inventory demand forecasting carries it through to the purchase order and the cash that order commits.

That second half is where consumer brands lose money. The forecast lives in one spreadsheet, inventory positions live in a 3PL portal, and the buy gets made from whichever number someone trusted that week. Slumber Cloud ran that way until it moved SKU-level forecasting onto connected data and reached 90% inventory forecast accuracy, while cutting financial management costs 30%.

Below are the formulas, a worked SKU example, how to calculate weeks of supply, and an honest read on whether your ERP planning module is the right tool for this job.

Inventory demand forecasting vs demand forecasting

The two terms get used interchangeably. They are not the same thing, and the difference is the reason so many brands forecast well and still stock out.

Demand forecasting answers one question: how many units will we sell? Inventory forecasting answers three: how much do we need on hand, in which location, and by when? Research firm Simon-Kucher draws the same line, noting that demand forecasting estimates customer demand while inventory forecasting estimates the stock levels required to serve it.

A complete inventory demand forecast produces three outputs:

• Unit demand by SKU by week
• A target stock position, expressed as weeks of cover
• A dated purchase order with a quantity

Most brands produce the first one and improvise the other two. That works until it does not. Consumer brands carry complications the textbook case ignores: lead times differ by channel, promotions distort history, sell-in and sell-through move at different speeds, and every buy commits cash months before the revenue arrives. Inventory is usually the largest use of cash on a consumer brand balance sheet. A forecast that stops at units is a forecast that cannot tell you whether you can afford your own plan.

The inventory demand forecasting formula

Three steps: forecast the demand, convert it to a buy, then check the buy against cash.

Step 1: forecast unit demand

Net unit demand = (average weekly velocity x weeks in period x seasonal factor) + promo lift - expected returns

Worked example. A hero SKU sells 400 units a week on average. You are planning a 13 week period that historically runs 1.2x the annual average. You have a promotion planned that has added about 900 units in past runs. Returns run 4%.

• Base: 400 x 13 = 5,200 units
• Seasonally adjusted: 5,200 x 1.2 = 6,240 units
• Plus promo lift: 6,240 + 900 = 7,140 units
• Less 4% returns: 7,140 x 0.96 = 6,854 units of net demand

Step 2: turn demand into a buy

Order quantity = net demand + safety stock - on-hand units - inbound units

Continuing the example. You hold 2,100 units on hand with 1,500 already inbound. Safety stock is set at three weeks of cover, or roughly 1,440 units at the seasonal run rate.

• 6,854 + 1,440 - 2,100 - 1,500 = 4,694 units to order
• Supplier MOQ is 5,000, so the order rounds up to 5,000

Reorder date = date stock falls below (lead time x weekly demand) + safety stock

At a seasonal run rate of 480 units a week, a 10 week lead time means you need 4,800 units of cover plus safety stock on the shelf when you place the order. Work backward from the week your on-hand plus inbound position crosses that line and you have a date, not a feeling.

Step 3: put the buy in the cash flow

This is the step almost every guide skips. A 5,000 unit order at a landed cost of $9 is $45,000 committed. If terms are 30% on deposit and 70% on shipment, that is $13,500 leaving this month and $31,500 leaving roughly ten weeks later, both well before the units sell through.

Put those two payments into your 13 week cash flow next to payroll and ad spend, and the order stops being an ops decision and becomes a finance decision. That is the whole point of connecting the two. Our demand planning template has these calculations already built if you would rather start from something than from a blank sheet.

How to calculate weeks of supply

Weeks of supply = current on-hand units / average weekly unit demand

If you hold 2,100 units of a SKU selling 480 units a week, you have 4.4 weeks of supply.

One nuance changes the answer materially. Use forward-looking forecast demand, not a trailing average. A SKU that sold 300 units a week for the last quarter and is about to run a promotion into peak season does not have the cover a trailing calculation suggests. The trailing number is most wrong exactly when the decision matters most.

Two related terms get used interchangeably on calls, so worth being precise. Weeks on hand usually counts only stock physically in the warehouse. Weeks of supply often includes inbound units already committed. Days of inventory is the same calculation on a daily denominator. Pick one definition and hold it across the team, because a plan built on two definitions produces two answers.

There is no universal target. Cover has to exceed your lead time plus a buffer, and lead time varies enormously by channel. A DTC SKU replenished domestically in three weeks can run lean. A retail SKU on a twelve week overseas lead time with a replenishment commitment cannot. As the finance lead at Oats Overnight put it, they needed to know the answer before their lead times made the decision for them.

Where the numbers actually come from

Every formula above needs four data feeds: sell-through by channel, current inventory positions, inbound orders with lead times, and landed cost. They almost never live in the same place.

• Shopify carries order and SKU detail, which gives you DTC velocity, discount behavior, and repeat purchase patterns.
• Amazon Seller Central carries orders by ASIN plus sellable and inbound FBA positions.
• Retailer point-of-sale data carries sell-through and door-level velocity, which is the only honest read on whether product is moving off shelf or just sitting in a DC.
• Your WMS or 3PL carries actual on-hand and inbound units.
• Your accounting system or ERP carries landed cost.

The important point is that none of these systems can forecast on its own, because each one sees a single slice. Shopify reports history. Amazon shows FBA cover but knows nothing about your retail stock. The 3PL knows units but not margin. Getting to one number means getting the feeds into one place, which is what Drivepoint's 100+ connectors exist to do. Laundry Sauce consolidated Shopify, Amazon, and QuickBooks into a single model and now forecasts at 98% accuracy, saving roughly 10 hours a month that used to go to pulling and pasting data.

Does NetSuite demand planning solve this?

Yes, partly. NetSuite has a Demand Planning module that forecasts item demand from historical data, seasonality, and open opportunities, then builds a supply plan and suggests purchase and work orders. If you already run NetSuite, it is worth understanding what it does before buying anything else.

What it is less suited to, for a multi-channel consumer brand:

• It is a separate module with its own implementation and configuration, not a switch you flip.
• It is organized around items and warehouses rather than sales channels, so channel-level economics sit outside it.
• DTC cohort behavior, marketplace fees, and retailer scan data generally are not in scope, which means the demand signal it works from is narrower than your actual demand.
• Planners regularly conflate it with MRP, and the setup work to make the two behave predictably is not trivial.

Here is the practical comparison for a brand deciding where planning should live.

None of this is an argument for ripping out your ERP. NetSuite stays your system of record. The planning layer sits on top of it, pulling GL detail and item costs as the actuals foundation and adding the channel and cash dimensions the module was never built to carry. We integrate with your ERP rather than compete with it. And because the model lives in Excel, the person who owns the buy can check the math themselves.

What accurate inventory forecasting is worth

Three levers, all measurable.

• Cash released. Overstock is cash sitting in a warehouse. Tightening cover on slow movers converts it back.
• Margin protected. Stockouts cost the sale and the ranking. The recovery usually costs expedited freight on top.
• Hours returned. Laundry Sauce got back 10 hours a month and $12,000 a year in financial management costs. Slumber Cloud cut financial management costs 30%.

Across Drivepoint customers, EBITDA margin improves by an average of 6.7 points in the first year, 75% increase EBITDA in year one, and brands save roughly $200,000 against hiring the equivalent finance team. Research from McKinsey on AI-driven forecasting in supply chains reports error reductions in the range of 20% to 50% where the data foundation is solid, which tracks with what we see: the accuracy gain comes less from a cleverer algorithm than from finally forecasting on complete data.

If you want to see the mechanics on your own numbers, the demand planning template is the fastest place to start, and SKU-level demand planning shows what it looks like connected. For a deeper read on method selection, see our breakdown of traditional vs AI-based demand forecasting.