Guide

How to forecast demand for a new product

Forecast a new product with no sales history from a launch plan or a like item's launch, turn a dollar target into weekly units, and borrow a season after eight weeks of sales.

A new product has no sales history, so there is nothing for a statistical forecast to learn from. Skuvelo forecasts it from what you know instead: your own launch plan, the first weeks of an item that looks like it and, once it has some weeks of its own, the season other items in its range already follow. This guide sets out each one and how they fit together.

Start with a weekly launch plan

When you create the item in Products, enter Expected sales per week. It is saved as a launch plan, and while the item is Launching with fewer than eight weeks of history, that plan is its forecast. The next planning run recommends a first order on Replenishment that covers the plan through the supplier's lead time plus 14 days, rounded to the supplier's order rules. Until the item has enough history of its own, you can change the weekly figure from the item's page.

A first-ever launch carries no statistical buffer, because there is no record yet of how far launches miss their plan; the recommendation says so. Once three or more past launches that had a launch plan have moved on from Launching, the buffer comes from how far their first 12 weeks landed from their plans.

Use a like item's launch curve

Skuvelo also looks for an item that resembles the new one. If you have not typed a plan, the match is the forecast. If you have, the match is shown beside your plan and never changes it.

  • Replaces. If the new item replaces an older one, the older item's last eight clean weeks (in stock, listed and synced) seed the forecast at its own level.
  • Looks like. Otherwise Skuvelo takes the items in the same category, priced within 25% of the new item when it has a selling price. The weekly level is the average launch rate across all of them. The week-by-week shape comes from one of them: it prefers the same supplier, then the launch shape closest to your typed plan or, with no plan, the most recent launch. An item's launch curve is its first eight clean weeks in its first 12 on sale; an item with fewer is not used, and an item with no category gets no match.

The item page names the match. If it is the wrong product, choose Not this product and the next match, or none, is used.

Borrow a seasonal shape

A new item cannot show its own season. While it is Launching on a plan or a like item's weeks, the forecast is those weeks, with no season added. Once it has more than eight weeks of history and Skuvelo forecasts it from its own sales, it can borrow a season from like items. Skuvelo tests a shape only on weeks after the item's first eight, so at exactly eight weeks nothing is tested and nothing is borrowed. Skuvelo looks at the closest group first: its style, then its family, then its brand within its category, then its category. It uses the first group where at least two other items have a full season. If those items share a strong seasonal shape, Skuvelo tests that shape against the item's weeks. It uses the shape unless its error on those weeks is more than 2% above that of the forecast Skuvelo would otherwise pick, or it misses a service level that forecast reaches. A few weeks of sales cannot show a season the item has not lived through, so the shape gets that small margin. If the group's shape is not strong, nothing is borrowed, even when a wider group has a strong shape. Nothing is borrowed either when no group has two such items. In both cases, set a seasonal swing yourself on the demand grid.

Turn a dollar target into a baseline

Buyers often know a launch as revenue: "this should do about $6,000 a month." The launch plan is entered in units per week, so convert it: divide the monthly dollars by the selling price, then by 4.33 weeks. At a $30 selling price, $6,000 a month is 200 units a month, or about 46 a week. Enter that as Expected sales per week, and the first order is sized from it.

Watch the first weeks

On the item's page, the forecast's Confidence reads Not scored until the item has enough weeks of sales to check the forecast against, and names what the forecast is running on: your launch plan, the like item's weeks, or the weeks of sales so far. If sales run well above or below the plan, the item shows "Selling N% above your launch plan" (or below); the plan stays as you set it until you change it. After eight weeks of sales that track the plan, Skuvelo suggests moving the item from Launching to Selling, and its own history takes over.

To check how far forecasts are missing once sales arrive, use the forecast accuracy calculator.

See also What Skuvelo uses to plan and Understanding planning and replenishment.

Try the plan on your own stock.

Start a trial with your own sales and stock, or look around the live demo first.