How to Forecast Demand for New Product Launches (With Zero Historical Data)

The Goal Is Not Forecast Accuracy. It's Inventory Risk Management

Most advice on new-product forecasting starts with the same observation: you don't have historical sales data.

That's true. But it isn't the real problem.

The real problem is timing.

Retailers have to commit inventory long before they know whether customers actually want the product. Purchase orders get approved. Factory capacity gets reserved. Raw materials are sourced. Containers are booked. Cash is tied up.

By the time the first meaningful sales data arrives, most of the inventory decision has already been made.

That's why experienced planners rarely obsess over finding the "right" forecast number.

The better question is: how much inventory risk are we willing to take before we learn anything?

A new-product forecast is not a prediction machine. It's a framework for making inventory decisions under uncertainty. The goal is to reduce uncertainty enough to make smart commitments, not eliminate uncertainty entirely.

The retailers that consistently launch products well are not forecasting the future with magical accuracy. They're using analog products, market signals, scenario planning, and rapid post-launch learning to manage risk better than everyone else.

Why Traditional Demand Forecasting Breaks Down for New Product Launches

Most forecasting systems are built on history.

They analyze sales trends, seasonality, replenishment patterns, promotions, and customer behavior. When a SKU has years of sales history, those methods work reasonably well.

A new product launch has none of those advantages.

There is no SKU-level demand history. No sell-through curve. No replenishment pattern. No proven relationship between price and conversion. The statistical models that perform well on replenishment items suddenly have very little to work with.

This is the classic cold-start problem.

Planners are expected to make buying decisions despite having limited evidence and significant uncertainty around customer adoption.

At that point, forecasting becomes less about mathematics and more about structured judgment.

The forecast is built from category knowledge, comparable products, market signals, customer insights, and business assumptions. The process becomes assumption-driven rather than history-driven.

This is especially common in apparel, footwear, and seasonal retail.

A retailer may technically launch hundreds of "new" products every season. A new color, silhouette, fabric, fit, or collection may have no direct sales history even though it sits within a familiar category.

A women's fashion brand launching a new wide-leg pant can't simply pull historical demand from that exact SKU because it has never existed before. Instead, planners must evaluate the product's attributes and ask what similar products have done in comparable situations.

That distinction matters.

The objective isn't to remove uncertainty. Uncertainty is part of the job.

The objective is to create a disciplined process for managing it.

Start With Analog Products, Not Gut Feel

When retailers forecast a new launch successfully, there's usually some form of analog forecasting behind it.

The idea is straightforward.

Find products that behaved similarly in the past and use them as a starting point.

The mistake many teams make is choosing analogs based only on category.

A footwear retailer launching a new running shoe may look at historical performance from all running shoes. That sounds logical, but it often creates noise rather than clarity.

Not all running shoes serve the same customer.

A premium performance shoe sold through specialty channels behaves differently from an entry-level shoe sold through broader distribution. Different price points attract different customers. Different marketing support creates different demand patterns.

Strong analog selection looks across multiple dimensions:

• Price point
• Product category
• Customer segment
• Sales channel
• Launch season
• Promotional support
• Product positioning

The closer these variables align, the more useful the comparison becomes.

For example, imagine a footwear brand introducing a new trail-running shoe.

Using last year's best-selling road-running model as the forecast benchmark may seem reasonable, but it could produce wildly misleading expectations. A previous trail-running launch with a similar price architecture, target audience, and distribution strategy is likely a much stronger analog, even if its total sales volume was lower.

Another mistake is relying on a single analog.

One reference point creates bias.

Three or four reference points create perspective.

Instead of inheriting one product's sales curve, planners can build a range based on multiple comparable launches. That range often reflects reality better than a single forecast number ever could.

What Makes a Strong Analog?

A useful analog typically shares several characteristics with the new launch:

• Similar customer audience
• Similar purchase frequency
• Similar price architecture
• Similar channel mix
• Similar launch timing

The more overlap you have, the more confidence you can place in the comparison.

This isn't guessing.

It's structured estimation based on observable demand behavior.

There's still judgment involved. There always will be. But it's far better than forecasting from instinct alone.

Build Demand Signals Before the Product Even Launches

The strongest new-product forecasts often begin before inventory arrives.

Many retailers treat forecasting as a one-time planning exercise completed months before launch.

In practice, forecasting should be an ongoing process of collecting evidence.

Every customer interaction creates information.

The question is whether you're capturing it.

Useful pre-launch signals include:

• Pre-orders
• Waitlists
• Product launch signups
• Wholesale commitments
• Search demand
• Product page traffic
• Customer surveys
• Retail buyer feedback

None of these signals are perfect.

A waitlist doesn't guarantee purchases. Product page visits don't guarantee conversion. Customer surveys often overstate intent.

But forecasting isn't about finding one perfect signal.

It's about combining multiple imperfect signals into a more informed view of demand.

Think of it as demand triangulation.

If wholesale buyers are increasing commitments, search volume is growing, and waitlists are filling up, those signals reinforce each other.

Confidence increases.

If search demand is strong but wholesale interest is weak and customer signups are stagnant, uncertainty remains high.

That's useful information too.

One apparel retailer, for example, may see strong social engagement around a new seasonal collection. Internally, excitement builds and buying teams become optimistic.

Then pre-orders open.

Actual purchase intent turns out to be far lower than expected.

That doesn't mean the launch will fail. It simply means the retailer now has better information before placing larger inventory commitments.

The strongest signals are usually those connected to actual purchasing behavior.

A customer joining a waitlist is expressing interest.

A customer placing a pre-order is committing money.

Those two actions should not carry equal weight in the forecast.

For retailers operating with long lead times, these early signals may be the most valuable forecasting inputs available.

Forecast Scenarios, Then Build Inventory Plans Around Them

One of the most damaging habits in retail planning is treating forecasts as precise numbers.

A forecast of 10,000 units implies a level of certainty that rarely exists.

Especially for new products.

A more practical approach is scenario planning.

Instead of producing one forecast, create multiple demand outcomes.

For example:

The purpose isn't to identify which scenario will happen.

The purpose is to understand the consequences if any of them happen.

Once those scenarios exist, inventory decisions become more structured.

Questions shift from forecasting to risk management:

• How much inventory should be purchased initially?
• How much safety stock is justified?
• When should replenishment orders be triggered?
• How much markdown exposure exists if demand disappoints?
• What happens to WOS under each scenario?

These conversations are far more valuable than arguing over whether the forecast should be 9,800 units or 10,200 units.

Inventory outcomes drive profitability.

Forecast accuracy alone does not.

Many retailers survive mediocre forecasts because they maintain flexible inventory positions and responsive replenishment processes.

Others experience heavy markdown pressure despite relatively accurate forecasts because they committed too aggressively upfront.

I've seen planning teams celebrate a forecast that ended up remarkably close to actual sales, while simultaneously sitting on excess inventory because allocation assumptions were wrong and replenishment timing was poor.

The forecast wasn't the problem.

The inventory strategy was.

Forecasting and inventory planning are inseparable disciplines. The forecast exists to support inventory decisions, not to win an accuracy contest.

The First 30 Days Matter More Than the Original Forecast

Many launch forecasts fail because they are treated as finished plans.

The launch forecast should be considered Version 1.

The real learning starts after customers begin buying.

The first few weeks provide information that months of planning simply cannot.

This is when planners should closely monitor:

• Sell-through rates
• Weeks of supply (WOS)
• Regional demand differences
• Store performance
• Channel performance
• Inventory velocity
• Return rates
• Size-level performance

Size-level performance deserves particular attention.

In apparel and footwear, inventory problems often appear long before total sales numbers look concerning.

A launch may appear healthy overall while specific sizes are already stocking out.

A footwear retailer may discover that demand is concentrated heavily in sizes 8, 9, and 10 while larger sizes remain untouched. If planners continue replenishing based on aggregate demand, stock imbalances develop quickly.

Now the business has two problems.

Popular sizes are unavailable.

Slow-moving sizes accumulate excess inventory.

Neither issue is visible if you're only monitoring total unit sales.

Allocation decisions frequently require similar adjustments.

A new product might outperform expectations in urban stores while underperforming in suburban locations. Without rapid reallocation, some stores experience stockouts while others carry excess inventory for weeks.

The retailers that react fastest typically outperform those with the most sophisticated launch forecasts.

Learning speed matters.

The faster you update assumptions, the lower the inventory risk becomes.

This is where modern forecasting tools have become increasingly valuable. Rather than waiting for monthly or quarterly reviews, planners can continuously evaluate demand signals and update forecasts as new information arrives.

Platforms such as Flagship help merchandising teams monitor changing demand patterns at a much more granular level, allowing adjustments before inventory problems become expensive. The advantage is not perfect prediction. It's faster visibility when demand starts behaving differently than expected.

Where AI Is Changing New Product Forecasting

AI is becoming particularly useful for products with limited or no history.

Traditional forecasting asks:

"How did this SKU sell previously?"

That question breaks down when the SKU has never existed.

Attribute-based forecasting approaches the problem differently.

Instead of focusing on SKU history, the model evaluates characteristics such as:

• Price point
• Color
• Style
• Category
• Brand position
• Seasonality
• Distribution strategy

It then identifies patterns across similar products.

This approach is especially relevant in fashion and footwear, where large portions of the assortment are effectively new every season.

A planner launching a new style may not have direct historical demand, but they often have years of history on similar combinations of attributes.

AI can process those relationships at a scale that would be difficult to replicate manually.

That doesn't eliminate uncertainty.

It simply provides another input into the decision-making process.

The best planning teams still combine model outputs with merchant judgment, inventory constraints, supplier realities, and commercial objectives.

Retail remains both an art and a science.

Stop Forecasting Demand. Start Forecasting Risk

The most useful question in new-product forecasting is not:

"How many units will we sell?"

It's:

"How much inventory can we safely commit before we know?"

That shift changes the entire planning process.

Instead of chasing a perfect forecast, retailers focus on reducing uncertainty through analog products, pre-launch demand signals, scenario planning, and rapid post-launch learning.

The forecast becomes a tool for managing inventory exposure.

Nothing more.

Nothing less.

The retailers that consistently succeed with new launches are rarely the ones with flawless forecasts.

They're the ones that make better inventory decisions while uncertainty still exists.

That's the real skill.

Not predicting demand perfectly.

Managing inventory risk before demand reveals itself.