Views: 0 Author: Site Editor Publish Time: 2026-09-10 Origin: Site
Deciding how many garments to produce is one of the most important decisions a clothing brand makes.
Order too few, and popular products can sell out before the next shipment arrives. Order too many, and the brand may be left with excess inventory, markdowns, and money tied up in unsold products.
For this reason, production quantities should be based on more than last year’s sales or a buyer’s intuition.
A practical apparel forecasting process looks at actual demand, inventory availability, seasonality, promotions, lead time, safety stock, size distribution, and available budget.
The goal is to estimate how much customers are likely to buy during the period when the new inventory will be available, then determine how much inventory the brand needs to support that demand.
For an OEM clothing manufacturer, this forecast eventually becomes a production order broken down by style, color, and size.
At first, forecasting seems simple.
If a brand sold 1,000 pieces of a style last year, it might order another 1,000 pieces this year.
The problem is that historical sales do not always represent actual customer demand.
Imagine a T-shirt sold 100 pieces per week. After four weeks, the Medium size sells out. Customers continue looking for Medium, but the brand cannot sell what it no longer has.
The sales record might show that demand fell after the fourth week.
In reality, the product became unavailable.
The supplied research describes this as demand censoring. When a product is out of stock, observed sales can fall to zero even though customers may still want to purchase it. Using these raw numbers directly can cause you to underestimate future production quantities.
This creates a dangerous cycle:
Popular product → sells out → sales data appears weaker → next order is too small → product sells out again
Meanwhile, slow-moving products can remain in stock and continue generating sales data.
Good forecasting therefore starts by asking:
What did customers actually want to buy when the product was available?
The first step is to collect historical sales data for the style, color, and size being forecast.
Useful information can include:
Units sold
Selling dates
Inventory available
Stockout periods
Price changes
Promotions
Product launches
Seasonal periods
Sales by size
Sales by channel
The more detailed the data, the more useful the forecast can become.
A brand with several years of sales history can identify recurring patterns more easily than a brand with only a few months of data.
However, more data does not automatically mean better forecasting.
The data needs to be interpreted correctly.
Suppose a brand had a 30-day sales period.
The product was available for 20 days and sold 400 pieces.
If the brand simply calculates:
400 ÷ 30 = 13.3 pieces per day
it assumes the product could have sold at that rate throughout the entire period.
However, if the product was unavailable for 10 days, the observed sales rate during the available period was:
400 ÷ 20 = 20 pieces per available day
The second figure may provide a better starting point for estimating demand.
The research recommends calculating the true Rate of Sale using the period when the SKU was actually available rather than treating out-of-stock periods as zero demand.
This process is often called demand unconstraining.
The basic idea is straightforward:
If customers could not buy the product, those days should not automatically be interpreted as days when customers did not want it.
Once historical sales have been cleaned up, the brand needs a baseline forecast.
The baseline is an estimate of normal demand before special events such as major promotions or unusual spikes are added.
Different products can require different forecasting methods.
For a basic product with relatively stable demand, a moving average can provide a useful starting point.
For example, a basic black T-shirt that sells consistently throughout the year may not require an elaborate forecasting system.
The brand can examine recent sales and establish a representative average.
If sales have been consistently increasing or decreasing, the forecast should account for that trend.
For example, if a core underwear style has grown steadily over several seasons, simply using an old average could underestimate future demand.
Linear trend models can be useful for this type of situation.
Some products follow predictable seasonal patterns.
Swimwear, winter underwear, thermal clothing, and other seasonal products can have very different sales patterns throughout the year.
The forecast therefore needs to distinguish between:
Normal growth or decline
and
Recurring seasonal demand
The research identifies exponential smoothing and ETS models as useful forecasting approaches because they can account for error, trend, and seasonality.
For most smaller brands, understanding the principle is more important than choosing a complicated forecasting algorithm.
A major sale can make a product appear much more popular than it normally is.
Imagine a brand usually sells 100 pieces of a style per week.
During a 30% discount campaign, it sells 250 pieces.
Using 250 as the new normal would produce a very aggressive production forecast.
The same issue can occur with:
Flash sales
Influencer campaigns
Major advertising campaigns
Holiday promotions
New-product launches
Temporary price reductions
The research recommends separating promotional lift from the underlying baseline demand, then adding planned promotional effects back into the future forecast when appropriate.
For example:
Normal demand: 100 units/week
Expected promotional lift: +40%
Forecast during promotion: approximately 140 units/week
The exact lift should come from the brand’s own historical data whenever possible.
Clothing demand is strongly connected to time of year.
Customer demand can change because of:
Temperature
Climate
Holidays
Back-to-school periods
Travel seasons
Promotional calendars
Fashion seasons
A brand should therefore compare similar periods when building a forecast.
For example, using December sales to forecast July demand for a seasonal product could produce a very misleading result.
The brand should ask:
What did this product, or similar products, sell during the same season in previous years?
Seasonality becomes especially important when production lead times are long.
If a factory needs several months to produce and ship a custom order, the brand has to forecast demand well before the actual selling season begins.
A new style creates a forecasting problem because there is no direct sales history.
A brand cannot calculate last year’s sales for a product that did not exist last year.
The research recommends using attribute-based clustering and like-item mapping for new products. Instead of guessing a number from nothing, planners can compare the new product with existing products that share important characteristics.
Relevant attributes can include:
Product category
Silhouette
Fabric construction
Fabric weight
Price
Target margin
Color
Pattern complexity
For example, suppose a brand is launching a new 280 GSM French Terry hoodie.
It can look at previous hoodies with similar:
Fabric weight
Construction
Price
Target customer
Fit
Color positioning
Those products can provide a more useful starting point than simply choosing an arbitrary quantity.
New products contain more uncertainty than established products.
A brand can therefore consider:
Smaller initial production
Pre-orders
Early customer interest
Historical performance of similar products
Flexible replenishment
Shorter production cycles where available
This can reduce the risk of committing too much inventory before actual demand becomes clear.
Forecasting demand is only half of the problem.
A brand also needs to know how long it will take to replace the inventory.
Suppose a brand expects to sell 500 pieces per month.
If replenishment takes two weeks, the brand has one inventory problem.
If replenishment takes four months, the brand has a very different problem.
The research emphasizes that factory production time should not be treated as the entire supply-chain lead time. Total lead time can include material procurement, sampling and approvals, bulk production, transportation, customs, and inbound processing.
A simplified apparel supply chain looks like:
Material procurement
↓
Sampling and approvals
↓
Bulk production
↓
Transportation
↓
Customs and inbound processing
↓
Warehouse availability
The brand should forecast inventory needs across this entire period.
This is particularly important for custom fabrics and custom colors, where material preparation can add significant time.
A forecast is still an estimate.
Actual demand can be higher than expected.
Production can also take longer than planned.
Safety stock provides a buffer against these uncertainties.
The research identifies two major sources of risk:
Demand variability
Lead-time variability
For example, suppose a brand expects to sell 1,000 pieces during the time it takes to replenish inventory.
If demand suddenly increases or production is delayed, 1,000 pieces may not be enough.
Safety stock provides additional protection.
A simple rule such as “always produce 10% extra” may seem convenient.
However, different products have different levels of risk.
A stable black T-shirt with predictable year-round demand may require a different buffer from a seasonal fashion style whose sales are highly unpredictable.
The research recommends using different forecasting and safety-stock strategies according to both revenue importance and demand predictability, using ABC-XYZ classification.
The basic concept is useful even without performing the full statistical analysis:
High-value + predictable products: protect availability carefully.
High-value + unpredictable products: monitor demand closely and adjust quickly.
Low-value + unpredictable products: avoid tying up too much inventory.
This helps brands spend inventory investment where it matters most.
Once a brand knows its expected demand and safety stock, it can determine when another production order should be placed.
This is the Reorder Point (ROP).
Conceptually:
Reorder Point = Expected demand during lead time + Safety stock
For example, if a brand expects to sell 1,500 units during its replenishment lead time and wants another 300 units as a safety buffer:
ROP = 1,800 units
When the available inventory position reaches that level, the brand should consider starting the next replenishment order.
The research describes the reorder point as the inventory threshold that triggers a new purchase order, based on expected demand throughout total lead time plus safety stock.
The important point is timing.
A brand should not wait until inventory reaches zero before contacting its manufacturer if the manufacturer requires weeks or months to replenish the product.
Knowing that a brand needs 5,000 pieces does not tell the factory how many pieces of each size to produce.
The production order might need to be:
XS: 350
S: 900
M: 1,900
L: 1,300
XL: 550
The exact ratio depends on the brand’s customer base and historical sales.
This is called a size curve.
The research highlights an important problem here: raw sales ratios can become distorted when popular sizes sell out early.
Suppose Medium repeatedly sells out.
Its historical sales may actually represent the number of Medium units the brand managed to sell, rather than the number customers wanted.
The brand should therefore analyze demand by size during periods when each size was actually available.
This is especially important for underwear because size is closely connected to fit, comfort, and customer satisfaction.
A production forecast should therefore answer two separate questions:
How many total pieces should we make?
and
How should those pieces be distributed across sizes?
A demand forecast can produce a very large production requirement.
The brand still needs to determine whether it can afford to manufacture and hold that inventory.
This is where financial planning enters the process.
The research describes Open-to-Buy (OTB) as a financial constraint that helps reconcile bottom-up SKU forecasts with the amount of inventory spending the business can support.
For a smaller clothing brand, the concept can be simplified:
Forecast the inventory you need, then check whether your available cash and inventory budget can support it.
If the forecast suggests ordering 20,000 pieces but the business can safely finance only 12,000, the brand needs to prioritize.
It may choose to:
Protect core products
Reduce slow-moving styles
Reduce color depth
Launch fewer SKUs
Use smaller initial orders
Delay lower-priority products
Plan another production run later
This is where forecasting becomes a business decision rather than a purely mathematical exercise.
After demand, seasonality, lead time, safety stock, size curves, and budget have been reviewed, the brand can determine its production requirement.
A simplified workflow is:
Historical sales
↓
Remove stockout distortion
↓
Build baseline demand
↓
Adjust for seasonality and planned promotions
↓
Consider production and shipping lead time
↓
Add appropriate safety stock
↓
Determine reorder point
↓
Apply size curve
↓
Check inventory budget
↓
Issue production order
The research summarizes a similar seven-step planning protocol, beginning with unconstrained historical demand and ending with a locked purchase order issued to the manufacturer.
From the manufacturer’s perspective, a good production forecast eventually needs to become a clear purchase order.
The factory may receive:
Style
Color
Total quantity
Size breakdown
Material specification
Packaging requirements
Delivery date
Shipping requirements
Quality requirements
The quality of this information affects production planning.
For example, a manufacturer may be able to produce 10,000 pieces of a style, but the production schedule still depends on the specific size and color breakdown, fabric availability, trims, delivery deadline, and other requirements.
This is why forecasting and manufacturing are closely connected.
A brand’s demand planning determines how much it wants to sell and when.
The manufacturer’s production planning determines how to turn that requirement into physical garments.
No forecast is perfect.
The purpose of forecasting is to make better decisions under uncertainty.
If actual demand is higher than expected, the brand may experience:
Stockouts
Lost sales
Emergency production
Expedited shipping
Customer dissatisfaction
If demand is lower than expected, the brand may face:
Excess inventory
Discounts
Markdown losses
Warehouse costs
Cash tied up in unsold products
The best forecasting systems therefore monitor actual sales continuously.
The forecast should change as new information becomes available.
For example:
Initial forecast → actual sales → updated demand estimate → revised production plan
This is particularly useful for fashion products with short lifecycles.
Imagine an underwear brand has a core style that sells throughout the year.
Historical data shows that, after accounting for stockouts, the product normally sells about 800 pieces per month.
The brand expects a seasonal increase of approximately 20%.
That gives a projected demand of roughly:
800 × 1.20 = 960 pieces per month
The manufacturer and logistics process requires three months from order placement to inventory availability.
The brand therefore needs to plan for approximately:
960 × 3 = 2,880 pieces
It then adds an appropriate safety buffer based on its demand and supply risk.
Suppose the brand decides that another 320 pieces provide a suitable planning buffer.
The inventory requirement becomes approximately:
2,880 + 320 = 3,200 pieces
The final step is to divide those pieces according to the expected size curve.
This is a simplified example rather than a universal formula. Real forecasting should account for changes in price, promotions, inventory availability, seasonality, lead-time variability, and product lifecycle.
The important point is the sequence:
Demand → selling period → lead time → risk buffer → size distribution → production quantity
A brand does not need an advanced AI forecasting system to improve its production planning.
It can start with better data discipline.
Record when each style and size becomes unavailable.
Otherwise, low sales may be mistaken for low demand.
Record major discounts and campaigns so unusual sales spikes do not automatically become the new baseline.
A total style-level sales number can hide important differences between sizes.
Record how long actual orders take from purchase order to warehouse receipt.
After each selling period, compare predicted demand with actual results.
Use similar products and attributes as references rather than pretending a new product has historical data.
A forecast should change when new information changes the expected demand.
Once the production quantity has been determined, the manufacturer needs a clear order.
For a custom clothing project, useful information includes:
Style number
Colorways
Total quantity
Size breakdown
Fabric specification
Trim requirements
Packaging requirements
Quality standards
Target delivery date
Shipping destination
A clear production order reduces the chance of misunderstandings between the brand and factory.
It also allows the manufacturer to plan fabric, cutting, sewing capacity, and other production resources more effectively.
For a long-term OEM relationship, historical production data can become useful as well.
Actual production lead times, material consumption, quality issues, and order quantities can all help improve future planning.
There is no single number that can guarantee the correct production quantity.
A strong forecast combines several pieces of information.
Historical sales tell the brand what happened.
Stockout analysis helps reveal demand that sales data may have missed.
Seasonality and promotions explain unusual changes.
Lead time determines how far ahead inventory needs to be planned.
Safety stock protects against uncertainty.
Size curves turn total demand into a practical production breakdown.
Budget limits keep inventory plans financially realistic.
Together, these factors create a much stronger basis for deciding how many garments to manufacture.
For clothing brands, the most useful shift is often moving from:
“We sold this many last time, so let’s order the same amount.”
to:
“How much did customers actually demand, how will that demand change, and how much inventory do we need to cover the time until we can replenish it?”
That question leads to better production decisions.
And for an OEM manufacturer, it creates a clearer connection between customer demand and the physical production plan.
Good production planning starts before the purchase order. The better the demand forecast, the better the chance that the right products, sizes, and quantities are available when customers want them.
