How forecasting works

For every product whose inventory Shopify tracks, Shelfcast estimates its daily demand — how many it expects you to sell each day over the coming weeks. That one number drives your reorder suggestions.

Forecast-based suggestions are part of the Growth and Pro plans. On Starter, Shelfcast still shows daily demand, but suggestions come from the reorder points you set.

What goes into a forecast

• What sold each day, across every sales channel and every location — online store, POS and any others. Test orders and cancelled orders are left out, and refunds and order edits are taken off.
• Which days the product was out of stock, so they can be skipped.
• When the product was created in Shopify, which is where its history starts (or its first sale, if that’s earlier).

Today doesn’t count until it’s over, so a day that’s still going never looks like a slow one.

Forecasts cover your whole store. Shelfcast adds up sales from every location and plans one suggestion per product, not one per location.

Out-of-stock days don’t count

A product can’t sell when you have none, so counting those days as zero sales would make it look less popular than it is. Shelfcast leaves them out of every average, both the units and the days.

Example: a product that sells 2 a day was out of stock for 10 days last month. Its forecast is still 2 a day, not less.

Shelfcast knows a day was out of stock when:

• its hourly stock record showed nothing available at the end of that day. These records start when you install Shelfcast; or
• the product is out of stock right now. Then every day since its last sale is treated as out of stock.

The forecast methods

Automatic is the default for every product. It looks at the product’s history and picks New item, Seasonal or Steady. You can also choose a method yourself on any product.

Steady

For products that sell at a fairly even pace. Shelfcast averages the last 90 days in stock, counting the most recent 30 for more:

0.6 × (the last 30 days’ rate) + 0.4 × (the rate over the 60 days before that)

Example: a product sold 3 a day over the last 30 days and 1 a day over the 60 before that. Its forecast is 0.6 × 3 + 0.4 × 1 = 2.2 a day.

If only one of the two stretches has days in stock, that one is used on its own.

Seasonal

For products whose sales rise and fall with the time of year. Shelfcast looks at what sold over the same stretch of time last year — the weeks you’re planning for now — and adjusts it for how this year is going.

The adjustment compares your last 90 days with the same 90 days a year before. If you’re selling twice as much as last year, last year’s figure is doubled. It never scales by more than double or less than half.

Example: last year, the coming 44 days sold 5 a day, against 1 a day for the rest of the year. This year the product is selling twice what it did last year. Its forecast is 5 × 2 = 10 a day.

Automatic picks Seasonal only when both are true:

• the product has at least a year plus your planning horizon of history (with the default settings, 365 + 51 = 416 days); and
• last year, the weeks you’re planning for sold at least 25% above or below that year’s average.

If you choose Seasonal for a product without that much history, Shelfcast uses Steady instead, and shows Steady so you know.

New item

For products with less than 60 days of history. Shelfcast takes the average per day in stock since the product was listed, but always divides by at least 7 days, so one early sale isn’t read as a rush.

Example: a product sells 7 the day after it’s listed. Its forecast is 7 ÷ 7 = 1 a day, not 7.

Reorder point

Not a forecast so much as a fixed rule: reorder a set quantity whenever stock falls to a set point. Automatic never picks it — you choose it per product, and on Starter every product uses it. The product still shows a daily demand (worked out the Steady way), but the suggestion ignores it. → Reorder points

No sales yet

A product that has never sold has nothing to forecast from. Its daily demand is 0, so nothing is suggested unless it uses a reorder point.

Your planning horizon

Forecasts look ahead over your planning horizon: the supplier’s lead time, plus Order enough for (days), plus Safety stock (days). With the default settings that’s 14 + 30 + 7 = 51 days. The horizon decides how far ahead Seasonal looks, and how much a suggestion covers. → Planning settings

Choosing a method for a product

Open the product from Products, find the Planning section for the variant, choose a Forecast method and press Save. That variant’s forecast is worked out again straight away.

The method shown on the product and on the Reorder page is the one actually used. If you chose Seasonal and it fell back, you’ll see Steady.

On Starter the Forecast method menu is turned off, with the note “Your plan reorders every product from its reorder point. Upgrade for forecasts.”

Just after you install

Shopify shares the last 60 days of your orders, so that’s where Shelfcast’s records begin. Your history then builds up every day Shelfcast runs.

Forecasts can read low at first

Shelfcast counts a product’s history from the day it was created in Shopify, and the days before its records begin count as days with no sales. Until your own history builds up:

• Steady forecasts can read a little low for about the first month, because part of the 90 days they average comes from before the records begin.
• Products created more than a year plus your planning horizon ago (about 14 months with the default settings) can be picked as Seasonal under Automatic. With no sales on record for the same weeks last year, they forecast 0 a day, and nothing is suggested.

If a product that sells shows Seasonal with a daily demand of 0, set its Forecast method to Steady. Switch it back to Automatic once Shelfcast has a year or more of your sales.

Stockouts before you installed aren’t known either, unless the product is out of stock right now, so a product that sold out often last month can look slower than it is for a while.

Which products are forecast

• Every product whose inventory Shopify tracks — including draft and archived products in Shopify.
• Not products marked Not tracked. Shopify doesn’t count their stock, so there’s nothing to plan against.

Forecasts refresh after each nightly sync and whenever you change something that affects them. → When forecasts refresh