SKU demand forecast calculator

SKU demand forecasting calculator

SKU-level forecasting means predicting units for each product on its own, because that is the level you buy at. Paste your monthly sales and get the next three months per SKU, its trend and season, and what to order. It runs in your browser, with nothing uploaded and no email to give.

Your monthly sales

Nothing to hand? (made-up products, 34 months).

How much to order

How to forecast demand per SKU with this tool

  1. Paste or drop your sales. Each SKU needs a row or dated rows, with at least three months of sales for a useful result.
  2. Check the column picks. It guesses them from the header row.
  3. Set your lead time and how many days the order should cover, then pick a service level.
  4. Read the table. Click any SKU to see its chart and the arithmetic behind its numbers.
  5. Download the result as a CSV if you want it in a spreadsheet.

Where to get the sales file

Two layouts work. The tool tells them apart by the header row.

  • One column per month. A SKU column, then columns headed like Jan 2026 or 2026-01. A pivot table of your sales gives you this. Add an On hand column and the tool subtracts it from the order.
  • Dated rows. A SKU, a date and a quantity on each line, added up by month for you.
  • Shopify: Orders, then Export. It reads Lineitem sku, Created at and Lineitem quantity.
  • Amazon: Reports, then Fulfillment, then the All Orders report under Sales. It reads sku, purchase-date and quantity.
  • Anything else: a Walmart or eBay download, or a 3PL shipment file. Pick the columns if the guess is wrong. Put channels together in one file and repeated SKUs are added up.

How the forecast works

Each SKU gets the same four steps, and each step only applies when the data supports it.

  1. Season. With two full years of sales, it compares each month with that year's average, after taking out the growth. If the years agree (correlation of 0.6 or better) and the peak and trough are at least 30% of an average month apart, those twelve monthly indices are used.
  2. Trend. With six months of sales, it takes the last twelve months, season removed, and asks whether sales move consistently one way. If they do, the trend is the median slope of every pair of months, so one big month cannot tilt it. If they do not, the SKU is flat.
  3. Level. The average of the last six months with the season removed, moved forward along the trend. A SKU that barely sells in those six months uses a full year instead.
  4. Project. Level plus trend, times the index for that month. The trend keeps 90% of its strength each further month, because a fast start rarely holds at the same speed.

A worked example

The sample file's MUG-12-WHT has 34 months of sales ending October 2026. Its level comes out at 108.2 units a month, rising about 1.3 units a month. November sells 1.83 times an average month, and the two most recent full years agree on that shape with a correlation of 0.99. So November is (108.2 + 1.3 × 0.9) × 1.83, about 200 units.

With 45 days of lead time and 30 days of cover, the order has to carry 75 days, or 2.46 months. That is 486.8 units of forecast across November, December and part of January. The typical miss on this SKU is 4.6 units, so at a 95% service level safety stock is 1.645 × 4.6 × √2.46, about 11.9 units. Subtract the 140 on hand and the suggested order is 359, rounded up.

What it cannot see

A sales file records what shipped, so a month you were sold out looks like a month nobody wanted it. A one-off promotion looks like demand. Cancelled orders count if your export includes them. A new SKU with under six months is a rough guide, and the tool marks it. It does not know your minimum order or case pack, so round the order up to fit. It does not split stock by location or channel. When a figure looks wrong, you know something about that SKU the file does not.

If a forecast is going into a purchase order, the landed cost calculator works out what those units really cost. For more on the method, read inventory forecasting explained, or see how purchasing in sku.io turns a forecast into a purchase order.

Questions people ask

What is SKU-level forecasting?

SKU-level forecasting predicts how many units of each individual product you will sell, month by month, instead of forecasting one number for the whole business or a category. It is the level you buy at, so it is the level the forecast has to be right at.

What is the formula for a demand forecast?

This tool uses forecast = (level + trend) × seasonal index. Level is recent average monthly sales with the season taken out. Trend is the units per month that level is moving. The seasonal index is how a given calendar month compares with an average month. A SKU with no trend or season is simply its level.

How many months of sales do I need?

Three months gives a forecast, but only as an average. Six months lets it measure a trend. Twenty-four months, two full years, lets it look for a season, because one year cannot tell a season from a one-time spike.

How is the suggested order quantity worked out?

It adds up the forecast over your lead time plus the days the order should cover, adds safety stock, and subtracts what you have on hand. Safety stock is the service level factor times the size of the typical forecast miss times the square root of the months covered. The result is rounded up, and never goes below zero.

Is my sales data uploaded anywhere?

No. The forecast runs in your browser, and the download button builds the CSV in your browser too. The tool is excluded from the site's session recording, so your SKUs and sales are not in it.

Why does a SKU that sold out show a low forecast?

Sales are not demand. A month where you ran out records what you could ship, not what people wanted. The tool cannot see stockouts in a sales file, so if a SKU was out of stock for part of its history, raise the forecast by hand.

Want this on live numbers?

Bring the result to a call. I'll go through it with you and show how sku.io forecasts demand per SKU with the math on screen, and turns the forecast into supplier purchase orders.

Book a Demo

Founder-led demo · 2-day onboarding · no onboarding fees