Demand Planning: The Order That Was Already in Your Forecast

Your forecast for April says 1,000 units. A distributor sends a firm order for 600. The planning system adds them up and plans 1,600. Six weeks later, 400 units sit in the warehouse, and everybody blames the forecast. The forecast was fine. The order was already in it.

Demand planning is the process of deciding how much of each product customers will need, and when, by combining forecasts with actual orders so every expected sale is counted once. Half of it is bookkeeping: matching each order to the forecast it belongs to.

Forecast consumption: how orders "eat" the forecast

A forecast is a placeholder for orders you haven't received yet. When the real order arrives, it should replace part of the placeholder, not sit on top of it. Planners call this forecast consumption.

Think of a restaurant on Saturday night. The manager expects 100 guests, and 60 have booked. The kitchen preps for 100, not 160. The bookings are part of the 100; the other 40 are expected walk-ins.

The tricky part is dates. Orders rarely land in the same week as the forecast they belong to, so you need a rule for how far an order may reach. That rule is the consumption window: how many periods backward and forward an order may look for forecast to consume.

Demand planning with dates: a worked example

April forecast: 1,000 units, spread as 250 per week over weeks 14–17. An order for 600 arrives, due in week 15.

No window (same week only). The order consumes the 250 in week 15. The other 350 has nowhere to go, so it is added on top. Total demand: 600 + 250 (week 14) + 250 (week 16) + 250 (week 17) \= 1,350. You just planned 350 units nobody asked for.

Window of one week back and one week forward. The order takes 250 from week 15, then 250 from week 14 (backward first), then the last 100 from week 16. Remaining forecast: 150 in week 16 and 250 in week 17. Total demand: 600 + 150 + 250 \= 1,000. Exactly what you expected to sell.

No consumption at all. 600 + 1,000 \= 1,600. The number from the opening paragraph.

Same forecast, same order, three answers: 1,000, 1,350 or 1,600. The difference is 600 units of finished stock, and it came from a setting, not from the market.

Two rules keep this honest. First, an order should consume forecast only for the same item and the same market; an order from Germany shouldn't wipe out the forecast for Spain. Second, someone must decide when an order is extra: a one-off project or a promotion the forecast never included. That call belongs in the S\&OP meeting, not in one planner's head.

Forecast bias vs forecast accuracy

Two products, both selling exactly 100 a week for four weeks.

• Product A is forecast at 110 every week. Error: 10% each week.
• Product B is forecast at 110, 90, 110, 90. Error: also 10% each week.

On an accuracy report they look identical: 90% accurate. They are not the same problem.

Product A is biased. It always misses in the same direction. After four weeks you have built 440 against 400 sold, and 40 extra units sit on the shelf. Next month there will be 80. Bias piles up.

Product B is noisy but unbiased: 400 forecast, 400 sold. The misses cancel out, and a small safety stock absorbs the weekly swings.

The fixes differ too. Bias is the easy one: find out why (sales padding numbers "to be safe" is the usual reason) and correct it. Noise can't be argued away; you cover it with a buffer or a shorter lead time. So always report bias next to accuracy. An accuracy number alone hides the one error you can actually fix.

Myth: a better forecast will fix late deliveries

It's tempting to think that if the forecast were right, orders would ship on time. Look at where late orders actually come from. A customer orders a custom pump with a three-week lead time and wants it in two. No forecast helps; that configuration was never forecast. A standard product is late because the paint line was overloaded, a casting arrived a week late, or the stock record said 40 and the shelf held 12.

Forecasts matter most for items you make to stock and for long-lead materials. For everything else, the promise depends on capacity and material at the moment the order arrives. That's why checking a sales order before you confirm the date matters more than another point of forecast accuracy.

And one thing demand planning should never touch: components. You forecast the pump; the seals, screws and housings follow from the bill of materials. That's dependent demand, and it's calculated, not guessed.

How it looks in factory.online

In factory.online, the 600-unit sales order creates manufacturing demand directly, and MRP explodes it through the BOM so every shortage stays linked to that order. The remaining 400 can be tested in a planning scenario before anything is bought, so the planner sees what the extra volume would do to materials and capacity instead of discovering it as stock in May.

If your planning numbers regularly add up to more than you sell, book a demo and we'll walk through it with your own orders.