--- title: How to read demand and set your forecast url: https://tideswell.xyz/docs/inventory/demand description: See how many units each product category is expected to sell over your buying horizon, and choose which past weeks the forecast learns from. --- > Documentation index: fetch https://tideswell.xyz/llms.txt to discover every page before exploring further. # How to read demand and set your forecast **Selling › Demand** shows how many units each product category is expected to sell over the time you are buying for. It sits beside the other stock screens in . An admin sets what the numbers are stated over, and which past weeks they ignore, in **Settings › Forecasting**. At Fairgreen Golf Co., Priya Raman opens Demand before she places the autumn order with Kaimen Knitwear, to see how many quarter-zips the season is likely to need. ## Before you start [#before-you-start] You need: * The Store and Forecasting modules on. Forecasting starts on in every workspace, and the **Demand** row and **Settings › Forecasting** show only while it is. See . * Your Shopify store connected. See . * **See sales, inventory, reports and the daily brief** on **Settings › Permissions**, which starts at lead. See . * For **Settings › Forecasting**, an admin. Saving a change there also needs **Change stock, receiving and store settings**, which starts at lead. ## Where do the demand numbers come from? [#where-do-the-demand-numbers-come-from] Tideswell works them out every night from your Shopify sales history, one category at a time. The work is the **Group forecast refresh**, one of the jobs in . A category is a Shopify product type, such as Quarter-zips. Each option value inside it, such as a size, also gets a row of its own, written as "Quarter-zips / Size M". Weeks when a category was out of stock are left out of the maths, so a sell-out never reads as a week nobody wanted it. Until the first night's run, the screen reads **No demand numbers yet**. Once it fills in, the line at the foot of the list counts the categories the nightly scan reached. That line also says **Wholesale orders are not counted**. Plan dealer orders from alongside these numbers. ## What do the columns mean? [#what-do-the-columns-mean] One row per category, with five columns: * **Category**: the product type, and the option value where there is one. * **Last 52 weeks**: a small chart of the units sold each week over the past year. * **Variants**: how many variants the category holds. * **Units to order**: what the category is expected to sell over your buying horizon, such as **420 over 30 days**. It is the expected sales, before any stock you already hold. * **Range**: the low and high ends of that estimate, such as **380 to 465**. The grey line under each category's name says how the number was worked out. It names the method, such as **seasonal forecast**, **trend forecast** or **average, not a forecast**, and how many weeks of history it averaged over. It adds how many sold-out or skipped weeks were left out, and how many variants could not be placed in the category. A category with no usable weeks reads **Not enough data** rather than a number. That happens when every week of its history was sold out or falls inside a period you told it to ignore. ## What do you see when you open a category? [#what-do-you-see-when-you-open-a-category] Click a row to open it, and click it again to close it. One category is open at a time. A chart shows this year's weekly units as **Demand**, with the **Forecast** running on as a dashed line. Where there is a year of history, **Same period last year** runs beside it, and weeks the category was out of stock are shaded and marked **Out of stock**. Rest on any week to see its units, or the units forecast for a week ahead. Under the chart, one sentence gives the answer, such as "420 units over 30 days, somewhere between 380 and 465, across 18 variants." Three cards follow: * **How this was worked out**: **Method**, **Averaging window**, **History span**, **Weeks not learned from** (stockouts and periods you excluded), **Stated over** (your buying horizon) and **Believed on** (the date of the run). * **Vs same period last year**: the change between the latest week and the same week a year ago, with how many variants were selling in each. It reads **Not comparable** when either week was out of stock, and needs a full year of weeks before it shows a figure. * **What other windows would have said**: the units per week each length of history would have given, with **(used)** beside the one Tideswell chose. It fills in once a category has 8 weeks in stock. When some variants could not be placed, a fourth card, **Not counted in this number**, says how many and why. That covers a variant whose option is blank, or spans two values, such as a size 8-10. ## How much history does each number use? [#how-much-history-does-each-number-use] Tideswell picks it for you, category by category. It tests each candidate length of history against what actually sold next and keeps whichever predicted best. That keeps the number tied to the evidence rather than to the order someone already had in mind. Each row names the window it used, and **What other windows would have said** shows what the shorter and longer ones would have given. ## How do you set how far ahead you are buying? [#how-do-you-set-how-far-ahead-you-are-buying] Set it on **Settings › Forecasting**, under **How far ahead you are buying**. Demand states every number over this many days. It starts at 30. 1. Open **Settings › Forecasting**. From Demand, the link at the foot of the list, **Change your buying horizon or excluded periods**, opens it too. 2. In **Days**, enter a number from 1 to 365. 3. Press **Save**. Set it to the time between placing an order and needing the stock on the shelf, plus however long that order has to last. Kaimen Knitwear delivers 60 days after Fairgreen orders, and each order has to last a month, so Fairgreen sets 90. Make the change the day before you plan from it. Demand fills in again over the new horizon after that night's run, and until then it shows no numbers. Past forecasts stay on record as they were, so you can later compare what was predicted with what sold. ## How do you stop an unusual week from skewing the forecast? [#how-do-you-stop-an-unusual-week-from-skewing-the-forecast] Add it under **Periods to ignore** on **Settings › Forecasting**. Those weeks are left out of the maths entirely, rather than counted as weeks with no demand. 1. In **Applies to**, choose **Everything**, **One category** or **One product variant**. 2. For one category, type its **Category name** as Shopify spells it. For one variant, enter its **Variant ID** in the form the box shows. 3. Choose the **From** and **To** dates. 4. In **Why**, say what happened. It is required, and it is the only record of why the number moved. 5. Press **Add period**. Fairgreen ran a warehouse clearance sale for one week in March, so Priya adds that week for **Everything** with the reason "Warehouse clearance, not normal demand". Each period is listed with its dates, what it applies to and its reason. Press the bin beside one (**Remove period**) to take it off. Most workspaces need none, so add a period only when you know a week misleads. ## Where else does this forecast show? [#where-else-does-this-forecast-show] The same nightly forecast, worked out for each product, drives the **Demand** line, **Cover** and **Reorder** on **Selling › Inventory**. See . To reorder up to stock limits you set yourself, rather than from the forecast, see .