Forecasting Demand for a Small Clothing Brand Without Guessing
By The Velocity Wear Team
Every apparel brand is a forecasting business whether it wants to be or not. Order too much and your margin is sitting in a stockroom; order too little and you sell out in the sizes that matter and disappoint the customers you just spent money acquiring. Forecasting tools have become genuinely affordable at small scale — but they solve one half of this problem and not the other.
What forecasting can do with thin history
Statistical forecasting needs history, and a brand six months old does not have much. What it does have is often enough for the questions that matter most.
- **Sell-through rate.** How fast a design moves once it is live. Two or three drops is enough to see a pattern, and it is the single most useful number you own.
- **Seasonality by proxy.** You may not have a year of your own data, but heavyweight fleece sells in autumn everywhere. Category seasonality is borrowable in a way brand-specific demand is not.
- **Reorder trigger points.** Given a sell-through rate and a lead time, the date at which you must reorder to avoid a gap is arithmetic, not prophecy.
The size curve is the real problem
Total quantity is the number people agonise over and it is rarely what hurts them. What hurts is the split. Sell out of M and L in three weeks while XS and 3XL sit for a year, and you have simultaneously lost sales and created deadstock from a single order.
A model cannot infer your size curve from three months of sales, because early sales are dominated by whoever happened to buy first. Your friends and your first hundred customers are not a representative sample of the market you are trying to reach.
This is the practical argument for a low minimum. At our 20-piece minimum, mixed sizes are included inside the run, so a first order can deliberately be a spread — two XS through two 3XL — bought to learn the curve rather than to maximise margin. The second order is then based on evidence.
Working backwards from lead time
The most common stockout is not a forecasting failure but an arithmetic one: the reorder was placed too late.
- 1**Establish your real lead time.** Production plus shipping plus the time it takes you to approve artwork. The last one is longer than people admit.
- 2**Work out your weekly run rate** on the sizes that actually move, not the average across all sizes.
- 3**Set the trigger at lead time plus a buffer.** If the whole cycle is six weeks and you sell twenty a week, you reorder when stock in the fast sizes hits around 150, not when you are nearly out.
- 4**Reorder the curve you observed,** not the curve you ordered last time.
Where the volume tiers change the maths
Forecasting in apparel is not purely about demand, because the unit price moves with quantity. Our tiers run from base pricing at 20–49 pieces to up to 40% off at 1,000+, which means a larger order is genuinely cheaper per piece.
The trap is treating that discount as free money. A 30% saving on stock that takes two years to sell is a loss dressed as a saving, once you count the cash tied up and the storage. The honest comparison is the discount against your sell-through rate: if the extra units clear inside a season, take the tier; if they will not, do not.
A method for a brand with no data
Order small and wide first. Sell it. Record what moved, in what size, how fast. Reorder the winners deep and the losers not at all. That is not sophisticated, but it is a forecast built on your own customers rather than an assumption, and it beats any model running on three months of noise.
