Data availability decides how to forecast demand for a new product. Time series and regression need sales history, so a SKU with none relies on comparable products, customer signals and early sales.
In Prediko's own forecasts for 27000+ new SKUs, only 21% had an error under 50% across their months, and the middle half missed by 55% to 122%.
The method note below shows what these figures do and do not measure.
Why is new product demand forecasting tough?
New product demand forecasting is hard because there is no sales history to anchor the first inventory order.
- Cannibalization and portfolio effects: New products don’t always create new demand. Some of it shifts from similar products in your catalog, so you have to estimate the new item’s sales and its effect on the rest of the range.
How accurate are new-SKU forecasts?
New SKU forecasts in Prediko under actual sales: Across 27000+ new SKUs, forecast units ran 12% below actual units in total, and only 21% of SKUs had an error under 50%.
April to September 2026
A new size or colorway also counts as a new SKU in this data even when its parent product has history, so start those from the parent’s sales.
Method note: This takes into Prediko’s forecasts only, not other tools or vendor accuracy claims.
Demand Forecasting Methods For New Products
Demand forecasting methods for new products depend on one question: Does the SKU have sales history? Time series and regression need it and without it judgmental methods are usually the only option.
1. Analogy based forecasting
Analogy-based forecasting uses the performance of a similar product as a reference point.
If you’ve launched something comparable in the past, same category, similar price, same target customer, you use that product’s launch data to estimate demand for the new one.
It’s useful when you’re releasing a line extension or entering a category with clear benchmarks.
This method gives you an anchor when there’s no direct history. It works best if
- You’ve launched something similar before
- You have access to detailed launch data
- The market context hasn’t changed too much
The forecast is only as accurate as the analogy you choose. If the reference product is off, the numbers will be too.
2. Customer survey & intent analysis
This method uses direct input from potential customers before launch. You ask how likely they are to buy, what they’d pay, or how they’d use the product. You can run surveys, interviews, or a pre-launch landing page with sign-up prompts.

Works best if
- You target the right customer segment
- You pair surveys with behavioral tests like pre-orders or waitlists
- You need fast feedback during product development
On its own, this method is directional. It’s stronger when combined with actions like pre-orders that show real commitment.
Learn how to properly setup pre-orders for your Shopify D2C store.
3. AI & ML models
AI and machine learning models forecast a new SKU from the history of other SKUs, so they need clean sales data across your catalog and channels, plus product attributes.
A SKU with only a few months of its own sales still carries wide error, as the accuracy section above shows.
To compare tools, see our guide to AI demand forecasting software.

Process to Do Demand Forecasting for New Products
Demand forecasting for new products takes five steps, from finding comparable products to tracking early sales against the forecast.
Step 1: Identify comparable products
Look for products that share key traits with the one you’re about to launch, same price range, customer segment, use case, and channel. The goal isn’t to find an exact match. It’s to find products that faced similar market conditions at launch.
Look for these data points in your comparable products.
- Sales patterns in the first 30, 60, 90 days
- Seasonality or timing of the product launch
- Marketing or channel strategy used
- How actual demand compared to initial expectations
Step 2: Conduct market research
Market research anchors the forecast to category demand, not just comparable launches. Use industry reports, similar-product sales or third-party retail data, then estimate the share you can reach at your price point and distribution plan.
Step 3: Adjust for external factors
External factors such as seasonality and competitor launches shape demand outside your business, so adjust the baseline for them.
- Timing: is the launch in a peak or a slow period?
- Competitors: are similar products entering the market at the same time?
Step 4: Build multiple variations
With new products, there’s too much uncertainty to work off a single forecast.
Create at least three versions: a base case, a high case, and a low case. Each should reflect different assumptions around sell-through, customer response, and availability.
This gives your team clear boundaries to work within and helps identify where the biggest risks or gaps might be. It also forces you to stress-test decisions against different demand levels before you commit to production.

Step 5: Monitor early signals and refine in real-time
The first few weeks after launch give you real sales to check the pre-launch forecast against. Track how the product is performing, not just by total sales, but by channel, region, and customer segment.
Look for patterns
- Are sell-through rates matching expectations?
- Are repeat purchases happening, or is it all first-time buys?
To check the forecast against actuals, track two numbers. Error is the total absolute gap between forecast and actual units, divided by actual units. Bias shows which way you miss.
To adjust inventory after a big launch, compare weekly sell-through with your base, high and low cases and reorder toward the case sales are tracking. If sales follow the high case, reorder early because supplier lead time sets how soon stock lands; if they follow the low case, hold further orders.

Common Mistakes to Avoid in Demand Forecasting of New Products
Common mistakes in demand forecasting for new products come down to two gaps: ignoring supply timing and starting from weak data.
1. Ignoring lead times and reorder cadence
Forecasting sales is only half the equation. You also need a clear supply plan to meet that demand. New products often come with uncertain lead times, especially if you're working with unfamiliar or new suppliers.
Learn how to manage and deal with suppliers in this webinar recap.
If your forecast assumes steady availability without factoring in production or shipping times, stockouts are likely. You also need to define how often reorders will happen. Without that, it’s hard to match supply to actual demand patterns.
2. Poor data foundation
With new products, that often means relying on proxy data, past launches, market research, or early customer feedback. But if those inputs are weak or misaligned, the forecast breaks before it starts.
Common issues include using outdated sales data, pulling comparables from the wrong product category, or relying on survey responses that don’t reflect your actual buyer segment.
Frequently Asked Questions
Which forecasting method is best for new product launches?
It depends on the product and data available. Judgmental methods such as analogy-based forecasting, expert panels and customer intent surveys are the usual starting points, because a new SKU has no history for time series or regression.
How do you forecast demand for a product with less than a year of sales data?
Use analogous product forecasting. Find a similar product by category, price point, or customer profile, then use its sales pattern as a proxy. Update the forecast monthly as real sales data builds. Avoid relying too heavily on the first few weeks, as launch effects can distort demand.
For a new Shopify store launch with no sales history at all, plan the first inventory order as a range. Build base, high and low cases from comparable products, pre-launch signals such as waitlists or pre-orders, and market research, buy toward the base case for the first cycle, and keep budget back to reorder once real sales arrive.
How do I forecast a purchase order quantity for a supplier when I order every four months?
Estimate demand over the four-month order cycle, then add safety stock for demand variability and supplier lead time.
Formula: (average daily sales × days in order cycle) + safety stock buffer. Increase the quantity if a seasonal peak falls within that window.








.avif)