Table of contents:
Author: 
Bani Kaur
0
Last Updated:
November 9, 2026

What Is SKU Forecasting? Definition, Models, and Workflow Explained

SKU forecasting projects demand for each size, color and variant. See how accuracy varies by demand pattern, best practices per SKU, and a 4 step workflow.

What is SKU Forecasting & Why Is It Important?

SKU forecasting is the process of projecting future demand for each individual SKU, meaning each size, color or variant you sell.

Aggregate demand forecasting, on the other hand, looks at total demand across categories, brands, or even an entire store. While useful for high-level planning, it lacks the detail needed to manage inventory at the product level.

SKU forecasting is important because it balances customer demand with financial efficiency, keeping the right products available without wasting money on excess stock. It also factors in seasonality, promotional spikes, and regional demand differences.

The difference is simple: Aggregate forecasting shows the big picture, SKU forecasting gives the granularity needed to avoid inventory mistakes.

How accurate is SKU-level forecasting?

SKU level forecasting is far more accurate for steady SKUs than for intermittent ones. In sample of Prediko's own data for April to September 2026, 69% of in-stock SKUs had a forecast error under 50% across these months, and that share ranges from 80% for steady SKUs to 43% for intermittent ones.

Share of in stock SKUs with forecast error under 50% from April to September 2026. The table shows steady, variable and intermittent separately.

Demand patternSKUs under 50% errorSKUs scoredBrands
Steady80%500+50+
Variable54%7,000+100+
Intermittent13%49,000+120+
All classified SKUs19%94,000+130+

All classified row also includes new, no-prior-sales and holiday-peaked SKUs, which the rows above do not break out.
For those groups, see demand forecasting for new products and seasonal demand forecasting.

The analysis is based on sample of SKUs, so read it as a direction rather than a rule.

This only takes into account Prediko's own forecasts. It is not a benchmark of other tools, and month-level SKU error is not comparable with vendor claims measured at category or weekly level.

Method note: April to September 2026. Random sample of brands with an active paid subscription in the window, scored on their selected forecasting engine. SKU-months counted only when the SKU was in stock all month, and each SKU needs 3 or more qualifying months.

Error is WAPE: Sum of absolute differences between forecast and actual units divided by actual units. Each SKU is placed in the first group that applies, in this order: new (first sale less than 6 months before April 2026), no prior sales (no order sales in the prior 26 weeks), intermittent (30% or more of those weeks with no sale), holiday-peaked (35% or more of April 2025 to March 2026 sales in the late-October to December peak), steady (weekly sales coefficient of variation of 0.5 or below over those weeks), and variable (everything else).

Best Practices for SKU Forecasting Accuracy

1. Segment your SKUs

Segmenting SKUs means grouping products by characteristics such as demand patterns, sales volume, margin contribution, or lifecycle stage. Each segment can then be forecasted with the method that fits it best.

Here’s how to approach it

  • High-volume, steady-demand SKUs: Need consistent replenishment
  • Seasonal or trend-driven SKUs: Show spikes in certain months and need flexible forecasting
  • Long tail SKUs with unpredictable demand: Benefit more from safety stock buffers than aggressive forecasts

Demand pattern is the segment to start with because the table above shows steady and intermittent SKUs forecast very differently. Sort each SKU by how steadily it sells and how many weeks pass with no sale, then forecast each group on its own terms.

For long-tail SKUs, calculate safety stock per SKU based on demand variability instead of pushing for a tighter forecast.

To estimate demand by size and color from Shopify sales data, forecast the parent product first, then split that forecast by each variant's share of past sales.

Leave out months when a size or color was out of stock, because those sales understate demand, and cover variants that sell intermittently with safety stock rather than a tight forecast.

2. Use AI-powered forecasting apps

AI-powered demand planning software helps automate the entire workflow of SKU forecasting.

Whichever tool you choose, ask it to report error by demand pattern, as in the table above, rather than as one overall number.

3. Set safety stock and reorder points per SKU

Safety stock provides a buffer when demand spikes or suppliers are delayed, while reorder points signal the exact moment to replenish before stock runs out.

To be effective, both must be calculated at the SKU level to reflect real demand patterns.

Economic Order Quantity (EOQ) adds another layer by identifying the most cost-efficient order size, balancing holding costs with ordering costs.

Together, these calculations reduce the risk of overstocking while avoiding last-minute emergency reorders.

For example, a fast-selling T-shirt line might need higher safety stock and tighter reorder points than a niche accessory that sells slowly. Treating them the same risks tying up capital in the wrong SKUs or missing sales altogether.

4. Leverage SKU relationships

SKUs often influence each other’s sales. Some are substitutes. If one runs out, customers switch to a similar product. Others are complements. They sell together, such as shoes and laces.

Forecasting them in isolation means you may overstock one item and understock another.

For example, if a retailer runs out of a best-selling sneaker, demand might spill over to a similar model. Without accounting for that link, the substitute could stock out too, leaving missed sales across both products.

Recognizing these relationships makes forecasts more realistic and ensures inventory is balanced across connected SKUs.

How to Implement A Proper SKU Forecasting Workflow

Step 1: Data collection and cleaning

Data should be captured at the most granular level, down to each SKU, location, and time period.

Duplicate entries, missing values, or outdated records can distort the actual demand patterns, so cleaning the dataset is critical to ensure only reliable information feeds the forecast.

Aim for real-time data as it further improves accuracy. Instead of relying on sales numbers that are weeks old, you can respond quickly when demand shifts.

Step 2: SKU Categorization

Once data is collected and cleaned, the next step is to categorize SKUs. Two common methods are:

  • ABC analysis: Classifies SKUs based on their contribution to revenue.
    • A items are high-value products that make up a small share of SKUs but a large share of sales
    • B items are mid-range contributors
    • C items are low-value products that form the majority of SKUs but generate less revenue
  • Velocity tiers: Groups SKUs by how quickly they sell.
    • Fast movers require frequent monitoring, replenishment, and safety stock levels
    • Medium movers show steady demand and require regular but less frequent replenishment
    • Slow movers benefit from longer review cycles, minimal safety stock, and alternative strategies like bundling

This step ensures forecasting models are matched to the right product type instead of treating every SKU the same.

Step 3: Accuracy tracking via dashboards and KPIs

Forecast accuracy metrics such as WAPE (weighted absolute percentage error, also written WMAPE) and SMAPE (Symmetric Mean Absolute Percentage Error) show how closely forecasts align with actual sales and highlight where adjustments are needed.

WAPE is the sum of the absolute differences between forecast and actual units divided by actual units. It still works in zero-sales weeks, where percentage errors are undefined.

Track bias alongside error, since bias shows whether forecasts run consistently above or below actual sales, and report both by demand pattern rather than as one overall number.

With the right inventory software, you also get customizable SKU forecasting report templates,  including sell-through rate, excess stock and ABC analysis.

‍

Step 4: Monthly reviews and adjustments

Demand shifts with promotions, seasonality, and market changes, so regular reviews are essential.

By checking forecasts against actuals every month, one can spot patterns, correct bias, and prevent small errors from snowballing into stockouts or excess. This way, your forecasts stay accurate, adaptive, and always aligned with real demand.

Frequently Asked Questions on SKU Forecasting

How should D2C brands set forecast accuracy targets by SKU category?

D2C brands should set forecast accuracy targets by demand pattern, then adjust for margin, stockout risk, and category importance, instead of using one target across all products.
Steady SKUs clear a given error threshold far more often than intermittent ones, so a flat target is too loose for the first group and out of reach for the second. For intermittent SKUs, cover the gap with safety stock.

What's a quick method to build size and color level forecasts?

A quick method is to use historical sales by SKU variant, split demand by size and color mix, adjust for seasonality and stockouts, then apply the forecast to upcoming inventory needs.

Can AI predict demand at the SKU and store level?

Yes. AI demand forecasting can predict demand at the SKU and store level by analyzing past sales, seasonality, inventory availability, promotions, and other demand signals.

Bani Kaur, Content Marketing Specialist at Prediko writing on inventory management & demand forecasting
Author Bio
Bani Kaur
Content Marketing Specialist
She brings over 6 years of SaaS and eCommerce experience to Prediko, turning complex topics like demand forecasting and inventory planning into practical, easy-to-follow content for merchants

What is Prediko?

Top Rated
Inventory Forecasting
Software (250+ Reviews)

Prediko rated 4.9 stars on the Shopify App Storefivestars
  • Sales forecasting & demand planning
  • Inbuilt AI agent to help you execute inventory workflows
  • Plans start from $49/month
Start Free Trial

Related articles on Inventory Forecasting

Demand Forecasting and Planning

Learn how Demand forecasting works for the best Shopify brands.
Read Guide

Shopify Inventory Forecasting

Learn the best practices to forecast inventory for your Shopify stores.
Read Guide

AI Inventory Management

The most up to date guide to learn everything about AI Inventory management.
Read Guide